Federal Subject-Matter Jurisdiction: Arising Under, Diversity, Supplemental Claims, and the Defect That Never Waives

In October 2000, after almost three years of motions and discovery and a six-day trial in Texas, a federal jury awarded a limited partnership called Atlas Global Group $750,000. Three weeks later, before the judgment was entered, the losing defendant filed a motion that had nothing to do with the evidence. When Atlas sued in 1997, the motion pointed out, two of its partners had been Mexican citizens. The defendant was a Mexican company. That put aliens on both sides of the case, and diversity jurisdiction does not allow that.

Atlas had a good answer, or what looked like one. The two Mexican partners had left the partnership a month before trial, so by the time anyone was picking a jury, the parties were fully diverse. The Fifth Circuit thought that was enough. By a 5-4 vote, the Supreme Court disagreed in Grupo Dataflux v. Atlas Global Group, L.P., 541 U.S. 567 (2004). Citizenship is measured when the suit is filed, the Court said, and it has adhered to that rule "regardless of the costs it imposes." The verdict was gone.

Very little else in American litigation works this way. Miss a statute of limitations defense and you lose it. Forget to object to venue or personal jurisdiction and you lose those too. Subject-matter jurisdiction, meaning the court's authority to decide this category of dispute at all, cannot be waived, forfeited, stipulated into existence, or excused because everyone behaved well and the jurors worked hard. The party who chose the court can challenge it after losing. The judge must raise it even if nobody asks. And in February 2026, a unanimous Supreme Court confirmed that even a district judge's good-faith ruling that a problem had been fixed will not save the judgment if that ruling turns out to be wrong.

The short answer

• Federal courts need two permissions. Article III sets the outer limit, and a statute must actually grant the power. Most jurisdictional fights are about the statute, and Congress has granted less than the Constitution allows.
• A case "arises under" federal law in one of two ways. Federal law creates the claim, or, in a small category defined by Grable & Sons Metal Products, Inc. v. Darue Engineering & Manufacturing, 545 U.S. 308 (2005), and Gunn v. Minton, 568 U.S. 251 (2013), a state claim necessarily turns on a disputed federal question that matters to the federal system as a whole. A federal defense does not count.
• Diversity means complete diversity plus more than $75,000. No plaintiff may share citizenship with any defendant. Citizenship of an LLC or partnership is the citizenship of every member, traced through every layer, as of the filing date.
• Removal gives defendants a one-way door with a hard clock. The 30-day deadline in 28 U.S.C. § 1446(b)(1) cannot be equitably tolled after Enbridge Energy, LP v. Nessel, No. 24-783 (U.S. Apr. 22, 2026). Remand orders are generally unreviewable. Procedural defects are waived after 30 days, and jurisdictional defects are never waived before final judgment.
• Plaintiffs control more than they used to. Under Royal Canin U.S.A., Inc. v. Wullschleger, 604 U.S. 22 (2025), a plaintiff in a removed case who deletes every federal claim strips the court of supplemental jurisdiction, and unless another basis such as diversity remains, the case goes back.
• A wrong ruling cannot manufacture jurisdiction. Under Hain Celestial Group, Inc. v. Palmquist, No. 24-724 (U.S. Feb. 24, 2026), when a court wrongly drops a nondiverse defendant as improperly joined, the later judgment falls on appeal, even a judgment won after trial.
• "Never waives" has an end point. Once direct review is over, a judgment is usually safe from collateral attack, even if jurisdiction was missing.

Why federal courts have to show their authority

State trial courts are usually courts of general jurisdiction. A state judge can presume the power to hear a contract case, a divorce, or a tort claim unless something says otherwise. Federal courts begin at the opposite end. They "possess only that power authorized by Constitution and statute," as Kokkonen v. Guardian Life Insurance Co. of America, 511 U.S. 375 (1994), put it, and that power "is not to be expanded by judicial decree." The party invoking federal authority carries the burden of proving it, a rule that goes back at least to McNutt v. General Motors Acceptance Corp., 298 U.S. 178 (1936).

The reason is structural. Article III lists the kinds of cases the federal judicial power may reach, and it leaves to Congress whether to create lower federal courts and how much of that power to give them. Every time a federal court decides a state-law dispute it has no authority over, it takes a case from a state court system that does have authority and has its own constitutional role. Jurisdictional limits are therefore a rule about the division of government power, and two private litigants have no standing to trade that away. Consent can waive a personal right. It cannot hand a court power that Congress withheld.

The cases that make this point are almost cruel. In Capron v. Van Noorden, 6 U.S. 126 (1804), a plaintiff sued in the federal circuit court in North Carolina, lost a jury verdict, and then assigned as error that the record never showed the parties were diverse. His own lawyer argued that "the consent of parties could not give" the court jurisdiction. The Supreme Court reversed the judgment in favor of the man who had picked that court in the first place. Eighty years later, in Mansfield, Coldwater & Lake Michigan Railway Co. v. Swan, 111 U.S. 379 (1884), the Court reversed a judgment against defendants who had removed the case themselves over the plaintiffs' objection. The Court admitted it looked like "an anomaly and a hardship" that the party at whose urging the error occurred could benefit. The rule, it said, "is inflexible and without exception." American Fire & Casualty Co. v. Finn, 341 U.S. 6 (1951), reached the same place: a defendant removed a Texas insurance suit, lost at trial, then moved to vacate its own loss because the removal had been improper, and won that motion in the Supreme Court. Jurisdiction, the Court explained, is guarded against expansion by "prior action or consent of the parties."

Judges bear this duty independently. In Louisville & Nashville Railroad Co. v. Mottley, 211 U.S. 149 (1908), neither side raised jurisdiction, and the Court decided the case on that ground anyway because "it is the duty of this court to see to it that the jurisdiction of the Circuit Court, which is defined and limited by statute, is not exceeded." Fed. R. Civ. P. 12(h)(3) codifies the point for trial courts: if the court determines at any time that it lacks subject-matter jurisdiction, it must dismiss. A court of appeals must check not only its own jurisdiction but also that of the court below. Mitchell v. Maurer, 293 U.S. 237 (1934). That can feel like an ambush to the parties. It is meant to.

How a court cannot duck the question

The Court closed off the tempting shortcut in Steel Co. v. Citizens for a Better Environment, 523 U.S. 83 (1998), rejecting "hypothetical jurisdiction," the practice of assuming jurisdiction to reach an easy merits ruling. A court with doubts about its power must resolve them before deciding the merits, however easy the merits look.

The ordering rule has some give among threshold questions. Ruhrgas AG v. Marathon Oil Co., 526 U.S. 574 (1999), found no unyielding hierarchy between subject-matter and personal jurisdiction, so a court may dispose of a straightforward personal jurisdiction objection first. Sinochem International Co. v. Malaysia International Shipping Corp., 549 U.S. 422 (2007), allows the same for forum non conveniens. For the sister doctrine that focuses on the defendant rather than the dispute, see Personal Jurisdiction Over Online and Foreign Defendants: From Zippo to Ford Motor.

What counts as "jurisdictional," and why the label matters

Courts used the word loosely for a long time. The modern rule comes from Arbaugh v. Y & H Corp., 546 U.S. 500 (2006): a statutory limit is jurisdictional only if Congress clearly says so. Otherwise it is an ordinary element or claim-processing rule that can be forfeited. Title VII's 15-employee minimum in Arbaugh was an element. Copyright registration under 17 U.S.C. § 411(a) is a precondition to suit, not a limit on power. Reed Elsevier, Inc. v. Muchnick, 559 U.S. 154 (2010). Title VII's charge-filing requirement can be forfeited. Fort Bend County v. Davis, 587 U.S. 541 (2019).

The removal statutes show the same sorting. A late notice of removal is a real defect, but not a jurisdictional one. That leads to an odd pairing in Enbridge: the 30-day deadline can be waived if the plaintiff fails to object in time, yet a court cannot extend it for equitable reasons when the plaintiff does object. Nonjurisdictional does not mean flexible.

The procedural vehicle

A challenge comes under Fed. R. Civ. P. 12(b)(1) or, in a removed case, by motion to remand under 28 U.S.C. § 1447(c). A facial attack accepts the allegations and says they fall short. A factual attack disputes the facts, and then the court may weigh evidence outside the pleadings, including through jurisdictional discovery, and need not presume the allegations true. The proponent of jurisdiction proves contested facts by a preponderance; Hertz Corp. v. Friend, 559 U.S. 77 (2010), confirms that "competent proof" is required once allegations are challenged. Appellate review of the legal question is de novo, and any factual findings are reviewed for clear error.

The standing rules are a separate Article III limit with their own literature. They are covered in Article III Standing After TransUnion: Concrete Injury, Traceability, and the End of No-Injury Claims. They also create a removal trap, discussed below.

Federal question jurisdiction: what "arising under" really means

Article III's phrase "arising under" reaches very far. Osborn v. Bank of the United States, 22 U.S. 738 (1824), read it to cover any case in which federal law forms an ingredient. The general federal question statute, 28 U.S.C. § 1331, uses nearly identical words and means much less. As Merrell Dow Pharmaceuticals Inc. v. Thompson, 478 U.S. 804 (1986), put it, the Court has "long construed the statutory grant of federal-question jurisdiction as conferring a more limited power." Same phrase, two meanings. Law students find this annoying, and they are right, but the gap has been the law for well over a century.

The well-pleaded complaint rule: the Mottleys' train passes

E.L. and Annie Mottley were hurt in a train collision in Kentucky in September 1871. They released their claims against the railroad in exchange for free passes for the rest of their lives. The railroad honored that deal for more than 35 years. Then Congress passed a 1906 statute restricting free passes, and the railroad stopped renewing them as of January 1, 1907.

The Mottleys sued in federal court. Their complaint anticipated the railroad's defense and said the statute did not apply or, if it did, was unconstitutional. They won below. The Supreme Court, raising the issue itself, sent them away. A suit arises under federal law only when the plaintiff's own statement of his cause of action shows that it is based on federal law, and a contract claim with a predicted federal defense is still a contract claim.

The rule sounds technical, but it does useful work. A court can decide jurisdiction on day one by reading one document. If jurisdiction depended on what the answer might say, nobody would know where the case belonged until much later. The cost is that genuinely federal disputes sometimes stay in state court, though the Supreme Court can still review the federal question from the state's highest court.

Four corollaries follow.

• A federal defense does not make a case removable, even a dispositive one the plaintiff concedes is the only real issue. Caterpillar Inc. v. Williams, 482 U.S. 386 (1987), describes the plaintiff as master of the claim, free to avoid federal jurisdiction by relying only on state law. See also Rivet v. Regions Bank, 522 U.S. 470 (1998).
• A federal counterclaim does not create jurisdiction. Holmes Group, Inc. v. Vornado Air Circulation Systems, Inc., 535 U.S. 826 (2002). Congress later altered this for patent and copyright cases, as described below.
• The Declaratory Judgment Act is no shortcut. 28 U.S.C. § 2201 adds a remedy but no jurisdiction. Under Skelly Oil Co. v. Phillips Petroleum Co., 339 U.S. 667 (1950), and Franchise Tax Board v. Construction Laborers Vacation Trust, 463 U.S. 1 (1983), courts look past the declaratory posture to the underlying coercive action.
• An insubstantial federal claim fails at the jurisdictional level only in extreme cases. Bell v. Hood, 327 U.S. 678 (1946), reserves dismissal for want of jurisdiction for claims that are immaterial, made only to obtain jurisdiction, or wholly frivolous. Most weak federal claims lose on the merits under Rule 12(b)(6), which means the court had power to decide them.

The one major exception to the anticipated-defense rule is complete preemption, a misnamed doctrine under which Congress has so thoroughly replaced state law that a state-law label is treated as federal from the start. The Court has recognized it for § 301 of the Labor Management Relations Act in Avco Corp. v. Aero Lodge No. 735, International Association of Machinists, 390 U.S. 557 (1968); § 502(a) of ERISA in Metropolitan Life Insurance Co. v. Taylor, 481 U.S. 58 (1987), and Aetna Health Inc. v. Davila, 542 U.S. 200 (2004); and usury claims against national banks in Beneficial National Bank v. Anderson, 539 U.S. 1 (2003). Ordinary preemption, which is what most preemption arguments are, remains a defense and supports no removal.

Merrell Dow, Grable, and Gunn: the small category of embedded federal issues

Justice Holmes offered a clean test in American Well Works Co. v. Layne & Bowler Co., 241 U.S. 257 (1916): a suit arises under the law that creates the cause of action. That test explains the vast majority of cases, but it has never been the whole story. In Smith v. Kansas City Title & Trust Co., 255 U.S. 180 (1921), a state-law suit to stop a trust company from buying federal farm loan bonds was federal because the state claim depended on whether the federal statute authorizing those bonds was constitutional.

Merrell Dow pushed back. Families alleged that the morning-sickness drug Bendectin caused birth defects, and they pleaded state-law negligence with misbranding under the Food, Drug, and Cosmetic Act as a presumption of negligence. Congress had provided no private right of action under that Act. By 5-4, the Court held that a federal standard embedded in this way did not create arising-under jurisdiction.

Grable then explained that the absence of a private right of action is evidence, but not dispositive. The IRS seized Michigan land from Grable to satisfy a tax debt and sold it to Darue. Grable sued in state court to quiet title, claiming the IRS's notice was defective because the agency used certified mail when the Internal Revenue Code called for personal service. Quiet title is pure state law. But Grable could win only by establishing the meaning of a federal tax statute, and the meaning of that statute was the central dispute. A unanimous Court held the case removable. The national interest in federal tax collection made a federal forum appropriate, and because quiet title actions turning on tax notices are rare, allowing them would barely change the state and federal workload.

Gunn restated the test in the form every practitioner now uses. A state-law claim arises under federal law only if the federal issue is (1) necessarily raised, (2) actually disputed, (3) substantial, and (4) capable of resolution in federal court without disrupting the federal-state balance Congress approved. Chief Justice Roberts acknowledged the mess he was trying to clean up, observing that the canvas in this area "looks like one that Jackson Pollock got to first."

The facts of Gunn are a patent lawyer's nightmare. Vernon Minton lost a patent infringement suit when his patent was held invalid under the on-sale bar. His lawyers raised an experimental-use argument too late. He sued them for malpractice in Texas state court, lost, and then argued that the state court had no jurisdiction because the claim turned on patent law, which belongs exclusively to federal courts. The Texas Supreme Court agreed with him. The U.S. Supreme Court unanimously did not.

The federal issue was necessarily raised and actually disputed. The case failed on substantiality, and the Court's explanation is the most useful thing in the opinion. Substantiality asks about importance to the federal system as a whole, not importance to the parties. Malpractice is backward-looking: "No matter how the state courts resolve that hypothetical 'case within a case,' it will not change the real-world result of the prior federal patent litigation." Minton's patent stayed invalid. State court rulings on hypothetical patent questions would not bind federal courts on real ones, and states have a special interest in regulating lawyers.

Here is how the four factors tend to play out.

A related decision, Merrill Lynch, Pierce, Fenner & Smith Inc. v. Manning, 578 U.S. 374 (2016), held that the exclusive jurisdiction clause of the Securities Exchange Act tracks the same "arising under" test as § 1331. Plaintiffs who plead state claims that happen to mention federal securities rules can stay in state court unless the Grable test is met.

IP cases: exclusive jurisdiction and its limits

This is where technology companies often trip. 28 U.S.C. § 1338(a) gives district courts original jurisdiction over civil actions arising under any Act of Congress relating to patents, plant variety protection, copyrights, and trademarks, and then says no state court has jurisdiction over any claim for relief arising under the patent, plant variety, or copyright statutes. Trademark is conspicuously absent from that second sentence, so Lanham Act claims can be heard in state or federal court. 15 U.S.C. § 1121(a) gives federal courts jurisdiction regardless of amount or diversity.

Three things follow, and they surprise people.

Patent and copyright cases can be in state court when the claim is really about a contract. A dispute over who owns a patent or whether a license was breached is generally a state-law claim even if the whole value of the case is the patent. Judge Friendly's classic opinion in T.B. Harms Co. v. Eliscu, 339 F.2d 823 (2d Cir. 1964), said a suit arises under copyright law when it seeks a remedy the Act expressly grants or requires construction of the Act, and ownership-by-contract disputes usually do neither. The circuits have refined the test, particularly for mixed ownership and infringement claims; compare Bassett v. Mashantucket Pequot Tribe, 204 F.3d 343 (2d Cir. 2000), with Scholastic Entertainment, Inc. v. Fox Entertainment Group, Inc., 336 F.3d 982 (9th Cir. 2003). Authorship and work-for-hire disputes that require applying the Act can stay federal, as in JustMed, Inc. v. Byce, 600 F.3d 1118 (9th Cir. 2010), a dispute over who owned a startup's source code. Ownership doctrine itself is covered in Copyright Ownership, Joint Authorship, and Termination of Transfers and Patent Ownership, Assignments, and Standing: Chain-of-Title Problems That Sink Cases.

Patent questions embedded in state claims now usually stay in state court after Gunn. Gunn itself removed malpractice from exclusive federal jurisdiction, and the reasoning has traveled. In Xitronix Corp. v. KLA-Tencor Corp., 882 F.3d 1075 (Fed. Cir. 2018), the Federal Circuit concluded that a monopolization claim premised on patent fraud did not arise under patent law and sent the appeal to the Fifth Circuit. And in Vermont v. MPHJ Technology Investments, LLC, 803 F.3d 635 (Fed. Cir. 2015), a state attorney general's consumer-protection suit over patent demand letters went back to state court, with the Federal Circuit affirming the remand (after finding its own appellate jurisdiction in a patent-preemption counterclaim). Patent lawyers who assume exclusive jurisdiction covers everything with a patent in it are working from an outdated map.

Counterclaims are different in patent and copyright cases. Holmes Group held that a patent counterclaim could not create Federal Circuit jurisdiction. The Leahy-Smith America Invents Act of 2011 responded three ways. It extended the exclusivity sentence of § 1338(a) to "any claim for relief." It added to 28 U.S.C. § 1295(a)(1) appellate jurisdiction over compulsory patent counterclaims. And it created 28 U.S.C. § 1454, which allows any party, including a plaintiff defending a counterclaim, to remove a civil action in which any party asserts a claim arising under the patent, plant variety, or copyright statutes, and allows a court to extend the removal deadline "for cause shown." That extension language matters more now that Enbridge has ruled out tolling for ordinary removals: the Court pointed to § 1454(b)(2) as evidence that Congress knows how to allow case-specific extensions when it wants them.

Suppose a Spokane developer named Tamarack Loop Software licenses code to a Boise distributor. Tamarack sues in Washington state court, alleging only breach of contract for unpaid royalties. The distributor answers and counterclaims for a declaration that it does not infringe Tamarack's copyright once the license ends. Before the AIA, the counterclaim would have been irrelevant to removal under Holmes. Today, § 1454 gives either party a route to federal court, and if the counterclaim truly arises under the Copyright Act, the state court cannot adjudicate it. Tamarack's own contract claim comes along only if it falls within supplemental jurisdiction, and § 1454(d) tells the court to remand whatever falls outside.

Declaratory judgment suits are the other common IP trap. An accused infringer seeking a declaration of non-infringement or invalidity relies on the coercive infringement suit the patentee could bring; that hypothetical suit supplies the arising-under basis, as Medtronic, Inc. v. Mirowski Family Ventures, LLC, 571 U.S. 191 (2014), confirmed in a licensee's action. Whether a live controversy exists is the separate Article III question from MedImmune, Inc. v. Genentech, Inc., 549 U.S. 118 (2007). Those two questions are treated together in Declaratory Judgment Actions: Jurisdiction, Discretion, and the Race to the Courthouse.

Federal claims in other technology disputes usually do the jurisdictional work on their own. A claim under the Defend Trade Secrets Act fits within 18 U.S.C. § 1836(c), so a competitor-hire dispute framed that way is federal, while one pleaded only under a state trade secret act needs diversity; see Hiring From a Competitor: Inevitable Disclosure, Employee Raiding, and the New Employer's Trade Secret Exposure. The same goes for a civil Computer Fraud and Abuse Act count, which is often the sole federal anchor for a bundle of state claims, as explained in The Computer Fraud and Abuse Act After Van Buren: Gates, Insiders, Scrapers, and the Loss Requirement. If that count is the only federal claim, Royal Canin means the plaintiff holds a key to the exit door.

Diversity: an old statute, and a very modern set of traps

28 U.S.C. § 1332(a) gives federal courts jurisdiction over civil actions between citizens of different states, and between citizens of a state and citizens or subjects of a foreign state, when the matter in controversy exceeds $75,000, exclusive of interest and costs. The traditional rationale is protecting out-of-state litigants from local bias. Whether that fear is still warranted is a fair debate. What is not up for debate is that the statute generates an unusual number of avoidable disasters.

Complete diversity

Strawbridge v. Curtiss, 7 U.S. 267 (1806), read the statute to require that every plaintiff be diverse from every defendant. One shared citizenship destroys the case. Strawbridge is a reading of the statute, not of Article III, as State Farm Fire & Casualty Co. v. Tashire, 386 U.S. 523 (1967), made clear when it upheld minimal diversity under the interpleader statute. That is why Congress could use minimal diversity in the Class Action Fairness Act, a regime treated in CAFA Jurisdiction and Class Action Removal: Amount in Controversy, Exceptions, and Mass Actions.

Aliens get special treatment. An alien can sue or be sued by a state citizen, but aliens on both sides with no state citizens on each side do not qualify, which was the Grupo Dataflux problem. A U.S. citizen domiciled abroad is "stateless," a citizen of no state, and cannot be sued in diversity at all; one such party destroyed jurisdiction in Newman-Green, Inc. v. Alfonzo-Larrain, 490 U.S. 826 (1989). A lawful permanent resident domiciled in the same state as an opposing citizen does not create diversity, under the proviso in § 1332(a)(2). Foreign corporations may be treated as foreign citizens even when organized in a British overseas territory. JPMorgan Chase Bank v. Traffic Stream (BVI) Infrastructure Ltd., 536 U.S. 88 (2002).

The stateless problem has teeth for modern partnerships. Because an LLC or partnership takes on the citizenship of each member, one member who is a U.S. citizen living overseas leaves the entity unable to be sued or to sue in diversity at all. Global law firms and investment partnerships with partners working abroad need to check for this before relying on § 1332.

Individuals: citizenship means domicile, and the residence allegation is a mistake

A person's citizenship is her domicile, meaning her true fixed home and the place she intends to return to. Residence is different, and pleading residence is a defect courts catch constantly. The Seventh Circuit has said so with some heat in Meyerson v. Harrah's East Chicago Casino, 299 F.3d 616 (7th Cir. 2002), and Heinen v. Northrop Grumman Corp., 671 F.3d 669 (7th Cir. 2012). Mas v. Perry, 489 F.2d 1396 (5th Cir. 1974), the casebook staple, involved a newlywed graduate student couple in Baton Rouge who sued their landlord after discovering two-way mirrors in their bedroom and bathroom. The wife, who came from Mississippi, kept her Mississippi domicile because she was in Louisiana only as a student and the couple had not decided where to settle. That made her diverse from the Louisiana landlord, and diversity is what let the federal verdict stand.

Remote work has made these questions more interesting. Picture a software engineer who grew up in Ohio, keeps a Columbus apartment, spends winters coding from Lisbon, and votes absentee. Her domicile is likely still Ohio, because domicile changes only with physical presence in a new place plus the intent to stay indefinitely. But if she gives up the apartment and moves abroad for good, she may become stateless, and a suit naming her in diversity fails. Courts weigh voting, taxes, driver's licenses, property, and family. McCann v. Newman Irrevocable Trust, 458 F.3d 281 (3d Cir. 2006), is a good example of how a court weighs those facts, including the presumption favoring an established domicile.

Corporations: Hertz and the problem of the remote-first company

28 U.S.C. § 1332(c)(1) makes a corporation a citizen of every state of incorporation and of the state where it has its principal place of business. Before 2010, circuits used competing tests, some of which looked at where the most business happened. That meant a large retailer might be a California citizen simply because California is big.

Hertz replaced those with a single rule. The principal place of business is the place where high-level officers direct, control, and coordinate the corporation's activities, typically its headquarters, called the nerve center. Justice Breyer's reasoning was explicitly practical: simple jurisdictional tests are a virtue because complex ones waste money litigating where to litigate. He also warned against gaming. If the alleged nerve center "is nothing more than a mail drop box, a bare office with a computer, or the location of an annual executive retreat," courts should look to the place of actual direction and control. And a Form 10-K listing "principal executive offices," without more, is not proof.

The most forward-looking sentence in the opinion is an aside: "in this era of telecommuting, some corporations may divide their command and coordinating functions among officers who work at several different locations, perhaps communicating over the Internet." The Court acknowledged these are "hard cases" and said the test still points toward the center of overall direction, control, and coordination. That sentence was written in 2010, before distributed leadership became routine.

Suppose a venture-backed company called Glasswing Ledger is a Delaware corporation with no office at all. Its CEO lives in Denver, its CFO in Atlanta, its general counsel in Portland, and the board meets over video. Nothing on a website settles where its nerve center is. A careful party seeking federal court would gather evidence of where the CEO and the senior team actually make decisions: where they are physically located when they sign contracts and approve budgets, where the corporate records and treasury function are run, and which address the company uses with its bank and regulators. The company should settle on the answer before it needs one, because a mistaken allegation can unravel a win. It also helps to know that a holding company's nerve center is where its own officers direct the holding company, even if a much bigger subsidiary sits elsewhere. Johnson v. SmithKline Beecham Corp., 724 F.3d 337 (3d Cir. 2013).

National banks follow their own statute, 28 U.S.C. § 1348. They are citizens of the state where their main office is located as designated in the articles of association, not every state with a branch. Wachovia Bank, N.A. v. Schmidt, 546 U.S. 303 (2006).

LLCs, partnerships, and trusts: the rule behind many vacated verdicts

Only corporations get the two-state treatment. Every other artificial entity takes the citizenship of each of its members, and that remains true no matter how the entity looks or what state law calls it. Carden v. Arkoma Associates, 494 U.S. 185 (1990), applied this rule to limited partners, who have no management role, by a 5-4 vote, and it candidly left any change to Congress. Americold Realty Trust v. ConAgra Foods, Inc., 577 U.S. 378 (2016), extended it to a Maryland real estate investment trust, whose members included its shareholders. A traditional trust, in which a trustee holds legal title and sues in his own name, generally takes the trustee's citizenship under Navarro Savings Association v. Lee, 446 U.S. 458 (1980). Americold confines that rule to traditional trusts.

The Supreme Court has not addressed LLCs directly, but every circuit to decide the question applies the same rule. See, for example, Zambelli Fireworks Manufacturing Co. v. Wood, 592 F.3d 412 (3d Cir. 2010), and Johnson v. Columbia Properties Anchorage, LP, 437 F.3d 894 (9th Cir. 2006). The Seventh Circuit's version states the drafting rule: an LLC's jurisdictional statement "must identify the citizenship of each of its members as of the date the complaint or notice of removal was filed, and, if those members have members, the citizenship of those members as well." Thomas v. Guardsmark, LLC, 487 F.3d 531 (7th Cir. 2007).

The best teaching case is Belleville Catering Co. v. Champaign Market Place, L.L.C., 350 F.3d 691 (7th Cir. 2003). The parties tried a lease dispute to a jury, which returned $220,000 on the defendant's counterclaim. On appeal, the Seventh Circuit noticed that everyone had treated the LLC like a corporation. Then it noticed that the plaintiff, pleaded as a Missouri corporation, was incorporated in Illinois; the lawyer had taken that from the lease. Judge Easterbrook wrote that counsel should "use the Internet" to check state corporate databases. One LLC member, itself an LLC, was "asserting confidentiality" over its own members, and the court replied that it is not possible to litigate in diversity with details kept confidential from the judiciary. When the winning party asked the court to affirm anyway, the opinion said this request "leaves us agog." The judgment was vacated. The court added that the costs of "a doomed foray into federal court" should fall on the lawyers, and that counsel should finish the case in state court without additional fees.

The technology behind the problem is simple: nobody's records show it. A Delaware LLC's filings on the public registry typically show no members. A cap table lives in a private spreadsheet. A fund investing in that LLC may have hundreds of limited partners, some of which are funds of funds, some foreign, some pension plans, and at the bottom some natural person who moved to your client's state last year. Suppose a Tacoma cider maker named Ninebark Cider Works, a Washington corporation, sues a Delaware LLC whose only visible member is a Delaware fund. If one limited partner of that fund, three layers up, is a retired dentist domiciled in Spokane, there is no diversity, and the discovery of that fact four years later on appeal costs both sides everything they spent. Nobody involved needs to have done anything wrong except fail to ask.

Two developments help. First, Fed. R. Civ. P. 7.1(a)(2), effective December 1, 2022, requires each party in a diversity case under § 1332(a) to file, with its first appearance, a statement naming and giving the citizenship of every individual or entity whose citizenship is attributed to it, when the action is filed or removed, and again when a later event could affect jurisdiction. The advisory committee note accepts that a court may limit disclosure where privacy interests are substantial, and it accepts pleading on information and belief at the outset. Second, Lincoln Benefit Life Co. v. AEI Life, LLC, 800 F.3d 99 (3d Cir. 2015), allows a plaintiff who has made a reasonable inquiry to allege that no member shares its citizenship, on information and belief, and let the LLC come forward with the facts. That solves the information asymmetry at the pleading stage. It does nothing if the facts turn out badly later.

The logic reaches newer entity forms, too. If a decentralized autonomous organization is treated as an unincorporated association, Carden suggests its citizenship would be that of every member, which could mean every token holder. Few doctrines age so awkwardly. A party in that position should expect diversity to be nearly impossible to establish and should look for a federal claim instead.

One small addition for a party thinking of arranging things to create diversity: 28 U.S.C. § 1359 bars jurisdiction where a party was improperly or collusively made or joined to invoke it, and Kramer v. Caribbean Mills, Inc., 394 U.S. 823 (1969), applied it when a Panamanian company assigned its contract claim to a Texas lawyer for a stated $1 and a side deal returning most of any recovery. In the other direction, plaintiffs often assign or join to avoid diversity, and courts are less uniform on those moves. Lincoln Property Co. v. Roche, 546 U.S. 81 (2005), held that a removing defendant need not prove there is no unnamed party whose presence would destroy diversity. The parties in the case, not hypothetical ones, are what counts.

Amount in controversy

More than $75,000, exclusive of interest and costs, is required. A plaintiff's good-faith claim controls unless it appears "to a legal certainty" that the claim is really for less, under St. Paul Mercury Indemnity Co. v. Red Cab Co., 303 U.S. 283 (1938). Later reductions do not oust jurisdiction. In removal, 28 U.S.C. § 1446(c)(2) allows the notice to assert an amount when state practice does not call for a specific sum, and the court decides by a preponderance if challenged. Dart Cherokee Basin Operating Co. v. Owens, 574 U.S. 81 (2014), says a notice needs only a "plausible allegation," with evidence required only if the plaintiff contests or the court questions it. Dart Cherokee was a CAFA case, and the Court expressly left open whether a presumption against removal applies in ordinary diversity cases.

For nonmonetary relief, the value of the object of the litigation is the measure. Hunt v. Washington State Apple Advertising Commission, 432 U.S. 333 (1977). Circuits still divide over whose viewpoint counts.

Subject matters diversity will not reach

Two judge-made exceptions survive, in limited form. Federal courts do not issue divorce, alimony, or child custody decrees, under Ankenbrandt v. Richards, 504 U.S. 689 (1992). And they do not probate wills or administer estates, but they may hear other claims involving those matters, under Marshall v. Marshall, 547 U.S. 293 (2006).

Time of filing and what can be cured

Many of the disasters happen here, so it pays to understand how the pieces fit together. The controlling rule dates to Mollan v. Torrance, 22 U.S. 537 (1824): jurisdiction depends on the state of things when the action is brought. Grupo Dataflux called this "hornbook law (quite literally)." In a removed diversity case, the parties must be diverse at both filing and removal.

The rule has two kinds of exceptions, and the Court has spent a generation drawing the line between them.

Dropping a dispensable nondiverse party. Rule 21 lets a court drop a party "on just terms," and Newman-Green held that even a court of appeals may do so to save a judgment, if no one is prejudiced. This works because the party who caused the defect is gone. The original case minus that party is a case the court could always have heard. 28 U.S.C. § 1653 also allows defective allegations to be amended, even on appeal, but it fixes the words only. As Newman-Green explained, it does not fix defective facts.

Caterpillar: the defect cured before judgment. In Caterpillar Inc. v. Lewis, 519 U.S. 61 (1996), the case was removed when a nondiverse defendant was still in it; the district court wrongly denied remand; and that defendant then settled and was fully dismissed before trial. The Court let the verdict stand. The statutory misstep at removal had been cured before judgment, and at that point "considerations of finality, efficiency, and economy become overwhelming." Note the limit in that sentence. The efficiency interest excused a defect in the removal, once the case was jurisdictionally sound.

Grupo Dataflux: a change in a party's own citizenship does not count. The Mexican partners left Atlas, but Atlas was the same party with a changed citizenship. No party was dropped. The Court refused to extend Caterpillar, explaining that that case "broke no new ground" because the cure there came from dismissing the party who destroyed diversity.

Hain Celestial: a mistaken ruling does not count either. Sarah and Grant Palmquist fed their son baby food made by Hain and bought from Whole Foods. After a 2021 House staff report on heavy metals in baby food, they sued both in Texas state court. They and Whole Foods were Texas citizens. Hain removed anyway, arguing that Whole Foods was improperly joined, meaning there was no plausible claim against it. The district court agreed, dismissed Whole Foods, denied remand, tried the case, and granted Hain judgment as a matter of law at the close of the plaintiffs' evidence. The Fifth Circuit reversed on improper joinder, and vacated the judgment, in Palmquist v. Hain Celestial Group, Inc., 103 F.4th 294 (5th Cir. 2024).

Justice Sotomayor's unanimous opinion affirmed. The dismissal was interlocutory, so it merged into the final judgment and was reviewable. Once reversed, Whole Foods was restored to the case, meaning the defect "lingered through judgment." Hain argued the parties were in fact diverse at judgment. The Court's reply: "This Court has never held that a district court can create jurisdiction through its own mistakes." Hain's efficiency argument failed too, because Caterpillar's efficiency reasoning excuses a statutory error once jurisdiction exists and cannot substitute for jurisdiction. The opinion resolved a split in which the Eleventh Circuit, in Henderson v. Washington National Insurance Co., 454 F.3d 1278 (11th Cir. 2006), had gone the same way and the Eighth, in Junk v. Terminix International Co., 628 F.3d 439 (8th Cir. 2010), the other.

Hain also tried Rule 21, and here the Court drew a line that will matter. Newman-Green involved a plaintiff asking to drop a party. A defendant asking to drop a properly joined nondiverse codefendant over the plaintiff's objection is trying to override the plaintiff's forum choice, and the rule does not permit that. The Court left open Rule 21's use for nondiverse parties added after a proper removal and in cases originally filed in federal court.

A footnote offers the practical escape. If a removed case depends on a contested improper-joinder ruling, the parties and the court should consider early appeal, through a partial final judgment under Fed. R. Civ. P. 54(b) or certification under 28 U.S.C. § 1292(b). Those tools, the Court noted, could help resolve uncertainty before trial. Justice Thomas concurred to question the improper-joinder doctrine itself: he reads the older cases as limited to actual fraud, and he doubts that federal courts can dismiss nondiverse defendants based on the merits of claims over which they lack jurisdiction. No other justice joined. The doctrine lives for now, but defendants should expect that concurrence to appear in remand briefs.

Royal Canin: amendments move the line. A plaintiff's operative pleading controls. Anastasia Wullschleger bought Royal Canin's prescription-only dog food, thinking it contained medication not found in off-the-shelf products, and sued in Missouri state court when she learned it did not, pleading Missouri claims that also alleged violations of the federal Food, Drug, and Cosmetic Act. Royal Canin removed. She amended to delete every reference to federal law and asked for remand. A unanimous Court, per Justice Kagan, held that the court lost supplemental jurisdiction and that the case had to go back. Under the statute's text, jurisdiction follows the amended complaint. An amendment "can wipe the jurisdictional slate clean." The Court rejected a contrary footnote in Rockwell International Corp. v. United States, 549 U.S. 457 (2007), as dictum, and it distinguished a court's dismissal of federal claims (which an appellate court can reverse) from a plaintiff's deletion (where the claims are "gone for good"). Notably, Royal Canin preserves the old rule about amount in controversy: a later amendment reducing the amount usually does not destroy diversity, because that rule concerns the value of the suit, not the plaintiff's choice of claims.

Put together, the state of play as of this writing looks like this.

Supplemental jurisdiction: the extra claims that come along

If a federal court had power over only the federal claim, plaintiffs would need to split a single story between two courts, and most would stay in state court. United Mine Workers v. Gibbs, 383 U.S. 715 (1966), fixed that by reading Article III's "case" broadly. State and federal claims that derive from a common nucleus of operative fact form one constitutional case. Congress codified the idea at 28 U.S.C. § 1367.

The statute has three working parts. Subsection (a) grants jurisdiction over all claims in the same case or controversy, including claims involving additional parties. Subsection (b) limits that grant in diversity cases: plaintiffs may not use it to bring claims against persons joined under Rules 14, 19, 20, or 24, or to add plaintiffs under Rules 19 or 24, if doing so would be inconsistent with § 1332. That codifies Owen Equipment & Erection Co. v. Kroger, 437 U.S. 365 (1978), where a plaintiff tried to sue a nondiverse third-party defendant after impleader. Subsection (c) lets a court decline jurisdiction where the state claim is novel, predominates, or is all that remains, or for exceptional reasons.

Exxon Mobil Corp. v. Allapattah Services, Inc., 545 U.S. 546 (2005), read the statute more broadly than many had expected. If one plaintiff in a diversity case meets the amount, co-plaintiffs with smaller claims come along, even when joined under Rule 20 or as class members under Rule 23, as long as complete diversity is intact. The Court distinguished the two requirements: a shortfall in amount does not infect the rest of the action, but a lack of complete diversity does.

The tolling provision deserves more attention than it gets. Under § 1367(d), the limitations period on a state claim dismissed from federal court is tolled while it is pending and for 30 days after. Artis v. District of Columbia, 583 U.S. 71 (2018), held 5-4 that tolled means stopped: the clock pauses and restarts, so a plaintiff has whatever time remained plus 30 days. Do not assume § 1367(d) covers every kind of dismissal or that it applies when there was never federal jurisdiction over anything in the action; check your forum's state savings statute too, and see Statutes of Limitations, Accrual, and Tolling: The Clock That Ends Most Claims.

Discretionary remand of state claims after federal claims drop out by the court's own ruling remains governed by Carnegie-Mellon University v. Cohill, 484 U.S. 343 (1988), and that kind of remand order is reviewable on appeal, because it does not rest on a lack of jurisdiction. Carlsbad Technology, Inc. v. HIF Bio, Inc., 556 U.S. 635 (2009). Kokkonen is the other surprise in this neighborhood. A federal court that dismisses a case after settlement does not automatically keep jurisdiction to enforce the deal; the order has to retain it or incorporate the terms. And the Federal Circuit held that a district court abused its discretion in taking supplemental jurisdiction over foreign patent claims in Voda v. Cordis Corp., 476 F.3d 887 (Fed. Cir. 2007).

Arbitration shows how a jurisdictional question can be dependent on how a case began. Vaden v. Discover Bank, 556 U.S. 49 (2009), lets a court "look through" a petition to compel arbitration to the underlying dispute. Badgerow v. Walters, 596 U.S. 1 (2022), refused to extend that to free-standing petitions to confirm or vacate awards. Then Jules v. Andre Balazs Properties, No. 25-83 (U.S. May 14, 2026), held that a federal court that stayed a pending federal-question suit under § 3 of the Federal Arbitration Act keeps jurisdiction to confirm or vacate the resulting award. So the same award may belong in federal court if the lawyer filed the original claims there and in state court if she filed a stand-alone petition. See Confirming and Vacating Arbitration Awards: The Narrow Door of FAA Review.

Removal: how a defendant changes courts

Removal is entirely statutory, and it is the mirror image of the plaintiff's choice. 28 U.S.C. § 1441(a) allows removal of any civil action over which the district courts would have had original jurisdiction, to the district and division embracing the state court, unless Congress says otherwise. The Court reads these statutes narrowly: Shamrock Oil & Gas Corp. v. Sheets, 313 U.S. 100 (1941), says federal courts must scrupulously confine themselves to the limits Congress set. Some cases cannot be removed at all under 28 U.S.C. § 1445, including suits under the Federal Employers' Liability Act and actions arising under a state's workers' compensation laws.

A complete walk-through of the procedure is available in Removal and Remand: Getting a Case to Federal Court and Keeping It There and in the Getting Into or Staying Out of Federal Court: A Practical Guide to Jurisdiction and Removal. This section concentrates on the doctrine.

Who is "the defendant"

Only the defendant to the original complaint can remove under § 1441(a). A plaintiff who faces a counterclaim cannot, under Shamrock Oil. Home Depot U.S.A., Inc. v. Jackson, 587 U.S. 435 (2019), extended that 5-4 to a third-party counterclaim defendant. Citibank sued Jackson on a credit card debt; Jackson counterclaimed and brought class claims against Home Depot and another company; Home Depot removed. The Court held that neither § 1441(a) nor CAFA's removal provision allowed it. The Court reasoned that "civil action" refers to the action as framed by the plaintiff's complaint, and a counterclaim is irrelevant to original jurisdiction. Section 1454 is the deliberate exception for IP claims.

The clocks

The ordinary deadline is 30 days from formal service. In Murphy Brothers, Inc. v. Michetti Pipe Stringing, Inc., 526 U.S. 344 (1999), a defendant received a faxed courtesy copy of the complaint, and formal service followed two weeks later. It removed 30 days after service. The Court held the clock runs from formal service, because service is what subjects a defendant to a court's authority. Early receipt alone does not trigger the deadline.

If a case is not removable as first pleaded, § 1446(b)(3) provides a new 30 days from receipt of "an amended pleading, motion, order or other paper from which it may first be ascertained" that the case is removable. Circuits have tried to make this mechanical to avoid forcing defendants to play detective. Harris v. Bankers Life & Casualty Co., 425 F.3d 689 (9th Cir. 2005), looks to the four corners of the papers rather than the defendant's subjective knowledge. Moltner v. Starbucks Coffee Co., 624 F.3d 34 (2d Cir. 2010), says the clock starts only on a paper explicitly specifying the amount. Walker v. Trailer Transit, Inc., 727 F.3d 819 (7th Cir. 2013), requires a paper that affirmatively and unambiguously reveals removability. The point of these bright lines is that a defendant is not penalized for failing to infer. They also tend to separate two questions, as Cutrone v. Mortgage Electronic Registration Systems, Inc., 749 F.3d 137 (2d Cir. 2014), explains: when the clock runs and whether the case is removable are not the same thing. A defendant who finds out on its own that a case is removable can generally remove earlier if the facts support it.

Enbridge ended the debate on stretching these deadlines. Michigan's attorney general sued in June 2019 to shut down the Line 5 pipeline, and Enbridge was served on July 12, 2019. It litigated in state court. After winning a federal jurisdictional ruling in a separate suit brought by the Governor, it removed the attorney general's case on December 15, 2021, which the Court calculated as 887 days after receipt of the complaint. A unanimous Court held equitable tolling unavailable. Even assuming the presumption favoring tolling applies, Congress wrote explicit exceptions: § 1446(b)(3) works like a discovery rule, § 1446(c)(1) has a bad-faith exception, and §§ 1441(d) and 1454(b)(2) allow enlargement for cause. A general equitable exception would make those redundant. The Court added that plaintiffs "usually can be confident" that once the deadline has passed, the forum question is settled. The lesson is about a defendant who sees a federal theory and waits to see how things go: the wait may be fatal, even if the theory is good.

The one-year limit and the voluntary-involuntary rule

A diversity case that becomes removable later cannot be removed under subsection (b)(3) more than one year after commencement, unless the plaintiff acted in bad faith to prevent removal. 28 U.S.C. § 1446(c)(1). Read the text closely, because it is narrower than many summaries suggest: it bars later-arising diversity removals under (b)(3), not removals based on federal questions and not an initial-pleading removal by a defendant served late. Congress added the bad-faith exception in 2011, after the Fifth Circuit had recognized an equitable exception in Tedford v. Warner-Lambert Co., 327 F.3d 423 (5th Cir. 2003). A plaintiff's deliberate nondisclosure of the true amount is deemed bad faith, § 1446(c)(3)(B). Keeping a defendant in the case while knowing the evidence would not support the claim can also qualify: Hoyt v. Lane Construction Corp., 927 F.3d 287 (5th Cir. 2019), found no clear error in a bad-faith finding of that kind.

The voluntary-involuntary rule is judge-made and old. A case not removable at the start becomes removable only through a voluntary act of the plaintiff. If a court dismisses the nondiverse defendant over the plaintiff's objection, the case stays where it is. The Supreme Court established the principle in Powers v. Chesapeake & Ohio Railway Co., 169 U.S. 92 (1898), Whitcomb v. Smithson, 175 U.S. 635 (1900), and Great Northern Railway Co. v. Alexander, 246 U.S. 276 (1918). Self v. General Motors Corp., 588 F.2d 655 (9th Cir. 1978), and Poulos v. Naas Foods, Inc., 959 F.2d 69 (7th Cir. 1992), carry it forward. Judge Easterbrook's opinion in Poulos explains the practical reason: an involuntary dismissal might be reversed by a state appellate court, which would restore the nondiverse party and force a remand, and the court was "anxious to avoid this sort of yo-yo effect." That logic should look familiar. It is close to the reason Hain Celestial gave for treating an interlocutory dismissal as not "gone for good." Improper joinder is the recognized exception, as Poulos and Hoyt both show: a defendant who proves it does not need a voluntary act by the plaintiff.

The forum-defendant rule and snap removal

Diversity exists to protect out-of-state defendants from local prejudice, so a local defendant has no such need. 28 U.S.C. § 1441(b)(2) therefore bars removal of a case based solely on diversity "if any of the parties in interest properly joined and served as defendants is a citizen of the State in which such action is brought." Most circuits treat this as a nonjurisdictional defect, waived if not raised within 30 days, as in Lively v. Wild Oats Markets, Inc., 456 F.3d 933 (9th Cir. 2006). The Eighth Circuit, an outlier for years, came around in Holbein v. TAW Enterprises, Inc., 983 F.3d 1049 (8th Cir. 2020) (en banc).

The words "and served" produced a technology story. They appear aimed at plaintiffs who name a local defendant they never planned to pursue. But once courts moved to electronic filing, and news and docket services began pushing new-case alerts to subscribers, a defendant could learn about a new suit within hours and remove before anyone could possibly complete service. The result is called snap removal.

Three circuits have approved it on the plain text. Encompass Insurance Co. v. Stone Mansion Restaurant Inc., 902 F.3d 147 (3d Cir. 2018), involved a defense lawyer who had agreed to accept service and then wrote to opposing counsel that service would have to wait until after the notice of removal was filed. The court called the result one that "may be peculiar," but "not so outlandish as to constitute an absurd or bizarre result." Gibbons v. Bristol-Myers Squibb Co., 919 F.3d 699 (2d Cir. 2019), and Texas Brine Co. v. American Arbitration Association, Inc., 955 F.3d 482 (5th Cir. 2020), agree. Texas Brine extends the idea to a non-forum defendant who removes before the local codefendants are served.

The rest of the map is less settled. The Eleventh Circuit had harsh words for a pre-service removal in Goodwin v. Reynolds, 757 F.3d 1216 (11th Cir. 2014), which affirmed a plaintiff's voluntary dismissal without prejudice after such a removal, and refused to "tie the district court's hands in the face of such gamesmanship." The Eighth Circuit has noted the debate without resolving it and held in M & B Oil, Inc. v. Federated Mutual Insurance Co., 66 F.4th 1106 (8th Cir. 2023), that snap removal "cannot cure a lack of complete diversity." The unserved nondiverse defendant still counts for jurisdiction. That limit is easy to miss and important: snap removal avoids the forum-defendant rule and nothing else. District courts in circuits without a controlling decision divide, and the Supreme Court has not taken the question. Congress held a hearing in 2019, documented in the Congressional Research Service's legal sidebar on snap removal, and scholars including Arthur Hellman have proposed amending the statute, but § 1441(b)(2) reads the same today.

There is a floor. In Casola v. Dexcom, Inc., 98 F.4th 947 (9th Cir. 2024), a California company learned about complaints through a Courthouse News Service report on the day they were electronically submitted, the day before Thanksgiving, and removed before the clerk had processed them. The Ninth Circuit held that a complaint is not filed in California until the clerk has processed and accepted it. There was no pending civil action to remove, the notices were legal nullities, and the 30-day remand clock never started. As the court put it, "we close the book on California defendants' attempts at ever-snappier snap removals." Because the remand rested on § 1447(c), the appeal was dismissed. The Ninth Circuit left the broader snap-removal question open. Anyone running a monitoring workflow should remember that an electronic submission is often only a request to a clerk, and that the docket event the software reports may precede legal filing.

Unanimity, and improper joinder

When removal rests solely on § 1441(a), every defendant properly joined and served must join or consent, under § 1446(b)(2)(A). Each defendant has its own 30 days, and an earlier-served defendant may consent to a later-served defendant's removal. Consent may be shown by a co-defendant's representation in most circuits. Mayo v. Board of Education of Prince George's County, 713 F.3d 735 (4th Cir. 2013), holds that a notice signed by one lawyer stating unambiguously that the others consent will do, because Rule 11 stands behind the statement. Compare the Fifth Circuit's demand in Getty Oil Corp. v. Insurance Co. of North America, 841 F.2d 1254 (5th Cir. 1988), for some timely written indication from each served defendant. Failure can be unforgiving: Taylor v. Medtronic, Inc., 15 F.4th 148 (2d Cir. 2021), involved a defendant who wrongly thought it had not been served and consented too late.

Improper joinder, also called fraudulent joinder, lets a court disregard a nondiverse or forum defendant who has no real claim against it. The standard varies. The Fourth Circuit requires no possibility of a claim, and it enforced that standard in Skidmore v. Schinke, No. 25-1436 (4th Cir. Mar. 30, 2026), vacating a dismissal and a denial of remand where a Virginia supervisor might be liable under state law. The Fifth, following Smallwood v. Illinois Central Railroad Co., 385 F.3d 568 (5th Cir. 2004) (en banc), uses a Rule 12(b)(6)-type analysis. The Seventh Circuit declined to decide whether the doctrine reaches diverse forum defendants in Morris v. Nuzzo, 718 F.3d 660 (7th Cir. 2013), calling it a very close question. After Hain Celestial, a denial of remand on this ground is a live risk throughout the case and may wipe out a defense win.

Federal officer removal under 28 U.S.C. § 1442 follows different rules. It needs no unanimity and has a broader scope, reinforced by Chevron USA Inc. v. Plaquemines Parish, No. 24-813 (U.S. Apr. 17, 2026), which held that a coastal-environmental suit over wartime crude-oil production plausibly related to the company's federal contract to refine aviation fuel.

Remand: the asymmetry that shapes strategy

Section 1447(c) creates two tracks. A motion based on a defect other than lack of subject-matter jurisdiction must be filed within 30 days of the notice of removal. A lack of jurisdiction requires remand at any time before final judgment. Fees and costs may be awarded, and the standard from Martin v. Franklin Capital Corp., 546 U.S. 132 (2005), is whether the removing party lacked "an objectively reasonable basis" for removal.

Section 1447(d) says a remand order "is not reviewable on appeal or otherwise," except in § 1442 and § 1443 cases. The bar applies to remands based on grounds in § 1447(c). Thermtron Products, Inc. v. Hermansdorfer, 423 U.S. 336 (1976), allowed review of a remand based on a crowded docket, a ground the statute does not recognize. Powerex Corp. v. Reliant Energy Services, Inc., 551 U.S. 224 (2007), limits inquiry to whether the district court's characterization of its ground as jurisdictional was colorable. Kircher v. Putnam Funds Trust, 547 U.S. 633 (2006), accepts that this sacrifices some accuracy. When an exception does apply, BP p.l.c. v. Mayor & City Council of Baltimore, 593 U.S. 230 (2021), allows review of the whole order.

The asymmetry is the heart of strategy. A wrong remand generally ends the matter with no review. A wrong denial of remand survives through final judgment and can undo everything. That means a defendant who wins a close remand fight has not actually won anything yet. Thanks to Hain Celestial, the risk runs for the life of the case.

The standing boomerang

Here is a trap few defendants see coming. A defendant removes a state-law privacy or statutory claim to federal court and then moves to dismiss, arguing the plaintiff has no concrete injury under TransUnion LLC v. Ramirez, 594 U.S. 413 (2021). If that argument succeeds, the court does not dismiss with prejudice. Section 1447(c) says that a court without subject-matter jurisdiction "shall" remand. In Collier v. SP Plus Corp., 889 F.3d 894 (7th Cir. 2018), a defendant removed a federal credit-card-receipt claim and then argued the plaintiffs had no injury. The Seventh Circuit vacated a dismissal with prejudice and ordered the case returned to state court, rejecting the idea that removal gets a defendant's foot in the door and then wipes the slate clean. In Polo v. Innoventions International, LLC, 833 F.3d 1193 (9th Cir. 2016), the Ninth Circuit reversed a dismissal on the same reasoning, doubting that any futility exception survives and finding remand would not be futile anyway. And in Thornley v. Clearview AI, Inc., 984 F.3d 1241 (7th Cir. 2021), plaintiffs kept a biometric privacy class action out of federal court by pleading a bare statutory violation. A party that wants a federal forum has to be ready to prove the plaintiff's standing, which is an awkward thing for a defendant to do. For the mechanics of this kind of motion, see Pleading and Attacking Standing in Federal Court: A Practical Guide.

The end of "never": when a jurisdictional defect finally goes quiet

The title of this article is a slight exaggeration, and it is good to know where. A jurisdictional objection can be raised at any point through direct review. After the appeal is over, the rules change.

Chicot County Drainage District v. Baxter State Bank, 308 U.S. 371 (1940), held that a federal court that determines its jurisdiction, even wrongly, produces a judgment that binds the parties and cannot be attacked later. Travelers Indemnity Co. v. Bailey, 557 U.S. 137 (2009), reaffirmed the point in the context of bankruptcy orders roughly 20 years old. United Student Aid Funds, Inc. v. Espinosa, 559 U.S. 260 (2010), explained that relief from a void judgment under Fed. R. Civ. P. 60(b)(4) is reserved for the exceptional case where the court lacked even an arguable basis for jurisdiction, and it is not a substitute for a timely appeal. Finality eventually wins, though only after every direct route has run out.

The lesson for the party holding a judgment is that a defect is most dangerous while the appeal is pending. For the party trying to attack a judgment years later, the odds are poor. Res Judicata, Collateral Estoppel, and the Preclusive Effect of Judgments discusses the broader doctrine. Limitations periods also continue to matter if a case is dismissed for want of jurisdiction, which is one more reason to confirm jurisdiction at the start. Also, 28 U.S.C. § 1631 authorizes transfer to cure a want of jurisdiction, mostly useful for cases in the wrong federal court, and 28 U.S.C. § 1919 lets a court order just costs when dismissing for want of jurisdiction.

What this means in practice

Two hypotheticals show how the pieces interact.

A worked example: the verdict that disappears

Suppose Ninebark Cider Works, the Tacoma cider maker, wins a $2.1 million jury verdict against a Delaware LLC called Pale Harbor Capital over a failed licensing deal. Both sides pleaded diversity in the usual way, and Ninebark's lawyer took Pale Harbor's own statement that it was a "Delaware limited liability company with its principal place of business in New York" and did not ask further. On appeal, Pale Harbor's new counsel reviews the fund documents and discovers that one investor in an upstream fund, at the time the complaint was filed, was domiciled in Washington.

The court must vacate. It does not matter that Pale Harbor chose not to raise this at trial or that its own disclosure was incomplete. Ninebark can try to save something: if the upstream investor's fund had been a separate named party, Rule 21 might help, but here it is the defendant's own citizenship, and the entity cannot be dropped without ending the case. Grupo Dataflux forecloses reliance on the investor's later exit. Ninebark's remaining route is state court, and § 1367(d) is no help for a diversity claim dismissed with no original jurisdiction at all, so Ninebark needs a state savings statute that fits. Many states have one, each with its own conditions, and if Washington's does not reach this dismissal the claim may be gone. Had Ninebark obtained a full Rule 7.1(a)(2) disclosure at the outset, or served a single interrogatory on the membership chain, it would have filed in Washington state court and tried the same case to a jury whose verdict would stand. Rule 11 has force here too. Fed. R. Civ. P. 11(b) requires an inquiry reasonable under the circumstances, and as Belleville Catering shows, a court may send the bill to counsel.

A worked example: the amendment that sends everyone home

Suppose Jonquil Analytics, an Ohio data company, is sued in Ohio state court by a former contractor, Odette Marsh, also of Ohio. The complaint alleges one claim under the federal Computer Fraud and Abuse Act and five under state law. Jonquil removes within two weeks, then spends eight months and a large budget on a Rule 12 motion and discovery in federal court. The judge dismisses two state claims. Marsh's lawyer, seeing the judge's rulings, then amends to drop the federal count.

Under the pre-2025 practice in several circuits, the judge would have had discretion to keep the case. After Royal Canin, there is no such discretion. The federal claim was the only source of original jurisdiction, and the leftover claims are out, so the case goes back to state court with whatever record it has built. Marsh may need leave to amend under Fed. R. Civ. P. 15 if her time to amend as of right has passed, and a court has some control through that gatekeeping function; see Amending Pleadings Under Rule 15: Leave to Amend, Relation Back, and the Scheduling Order Trap. But once the amendment is allowed, the jurisdictional consequence is automatic. If diversity had existed, this would not have happened. So a defendant deciding whether to remove a mixed case should ask whether the federal claim is one the plaintiff can afford to abandon. If it is, the removal may buy only delay, and interlocutory rulings won in the meantime will not bind the state court the way a final judgment would.

How a careful lawyer should use this

A practical reading of these cases suggests a few habits:

Do the citizenship work before filing, and document it. The moment to ask who owns the LLC is before the complaint. A short engagement-letter clause or intake question can get it; the Federal Jurisdiction Toolkit: Notices of Removal, Remand Motions, and Jurisdictional Allegations has a worksheet. The Federal Jurisdiction and Removal Checklist: A Practical Checklist has the same items in a form to keep at hand.

Treat improper joinder as a loan and not a purchase. If removal depends on it, consider Rule 54(b) or § 1292(b) certification, as the Court suggested in Hain Celestial. A defense verdict resting on an uncertified improper-joinder ruling is a win with an asterisk.

If you have a jurisdictional doubt, say so. Courts have repeatedly placed that duty on counsel. The Seventh Circuit's approach in Belleville Catering makes clear that remaining silent, or pretending the court can overlook it, is a losing strategy and a risky ethical posture.

Know when federal court is the wrong destination. Federal court is not always better. For choosing among forums and defending the choice, see Forum Selection, Transfer, and Parallel Proceedings: Choosing and Keeping the Court You Want. A forum-selection clause can also close the federal door: FindWhere Holdings, Inc. v. Systems Environment Optimization, LLC, 626 F.3d 752 (4th Cir. 2010), read a clause choosing the courts of a named state, framed in terms of sovereignty, to exclude federal courts located there, and Atlantic Marine Construction Co. v. United States District Court, 571 U.S. 49 (2013), governs how those clauses are enforced in federal court. For the appeal side of these decisions, see Federal Appellate Practice: Preserving Error, Standards of Review, and Briefs That Win.

What is unsettled

• Snap removal remains undecided in most circuits, and the Supreme Court has not taken it up.
• Improper joinder's future. Justice Thomas has invited a challenge to the doctrine, and the standard remains uneven across circuits.
• Rule 21 in other postures. Hain Celestial reserved cases where a nondiverse party was added after proper removal, as in Gorfinkle v. U.S. Airways, Inc., 431 F.3d 19 (1st Cir. 2005), and cases originally filed in federal court.
• Amendments in diversity cases. Royal Canin sets a broad principle that jurisdiction follows the operative pleading, while preserving the amount-in-controversy rule. How far that reasoning goes is being worked out. The Ninth Circuit has applied it to a joint stipulation dismissing federal claims in Walker v. Arizona, No. 24-1806 (9th Cir. Oct. 22, 2025), and to deletion of class allegations under CAFA in Faulk v. Jeld-Wen, Inc., No. 24-4078 (9th Cir. Nov. 14, 2025). Other circuits may not go as far, so check before relying on either.
• Rule text. Proposed amendments to Rule 7.1 (expanding the parent-company disclosure to any publicly held business organization that directly or indirectly owns 10% or more of the party) and Rule 81(c)(3) (requiring a jury demand after removal when none was made before) advanced through the Judicial Conference's rules process in 2026. They have not taken effect as of this writing, and an amendment approved on the usual timetable would not take effect before December 2027.
• The remote nerve center. Hertz flagged the problem and left it for later. Lower courts are still applying the test to companies without a physical headquarters.

For deeper reading, the standard treatment is Wright, Miller, Cooper, and Steinman, Federal Practice and Procedure, particularly § 3562 on arising-under jurisdiction, §§ 3721 and 3723 on removal and diversity, and § 3733 on the notice. Hart and Wechsler's The Federal Courts and the Federal System remains the casebook that explains why these rules exist. On snap removal, see Arthur D. Hellman et al., Neutralizing the Stratagem of "Snap Removal": A Proposed Amendment to the Judicial Code, 9 Fed. Cts. L. Rev. 103 (2016). Primary sources are easy to check on UpLaw.

This article is general information and not legal advice for a specific situation; if a jurisdictional question is live in your case, a short conversation with litigation counsel early on is much cheaper than discovering the answer on appeal.

Related Documents

Related Articles

• CAFA Jurisdiction and Class Action Removal: Amount in Controversy, Exceptions, and Mass Actions: the separate minimal-diversity regime for class cases.
• Article III Standing After TransUnion: Concrete Injury, Traceability, and the End of No-Injury Claims: the constitutional limit behind the standing boomerang.
• Declaratory Judgment Actions: Jurisdiction, Discretion, and the Race to the Courthouse: how a declaratory posture interacts with arising-under doctrine.
• Personal Jurisdiction Over Online and Foreign Defendants: From Zippo to Ford Motor: the other kind of jurisdiction, and the one that can be waived.
• Forum Selection, Transfer, and Parallel Proceedings: Choosing and Keeping the Court You Want: the broader strategic picture.
• Hiring From a Competitor: Inevitable Disclosure, Employee Raiding, and the New Employer's Trade Secret Exposure: a dispute type where the DTSA supplies the federal anchor.
• The Computer Fraud and Abuse Act After Van Buren: Gates, Insiders, Scrapers, and the Loss Requirement: another claim that is often a case's only federal hook.

Related Guides

• Getting Into or Staying Out of Federal Court: A Practical Guide to Jurisdiction and Removal: the companion step-by-step guide for plaintiffs and defendants.
• Removal and Remand: Getting a Case to Federal Court and Keeping It There: a procedural walk-through, including CAFA and fraudulent joinder standards.
• Pleading and Attacking Standing in Federal Court: A Practical Guide: how to plead and challenge Article III standing.

Related Checklists

• Federal Jurisdiction and Removal Checklist: A Practical Checklist: the companion checklist for complaint, removal, remand, and late-discovered defects.

Related Toolkits

• Federal Jurisdiction Toolkit: Notices of Removal, Remand Motions, and Jurisdictional Allegations: the companion templates and worksheets.
• Civil Procedure Toolkit: Pleadings, Jurisdiction, Preclusion, and Deadlines: a broader civil procedure working set.