Introduction

When a federal court hears a diversity case — a suit that is in federal court only because the parties are citizens of different states — the rights it adjudicates were created by state law. Contracts, torts, property, and the defenses that go with them are subjects the states, and ordinarily only the states, have power to govern. Before such a court can decide anything on the merits, it must therefore answer a prior question: whose law supplies the rule of decision? The answer matters because the two candidate bodies of law can differ on points that decide cases outright — whether a claim is time-barred, whether a damages ceiling applies, whether a suit may proceed at all. If federal judges were free to fashion their own answers, the same dispute could come out one way in state court and the opposite way in the federal courthouse across the street, and the party able to choose the forum would, in effect, choose the law.

The body of law that answers this question is the Erie doctrine, and this chapter builds it in the order the Supreme Court did. It begins with the foundational command that a federal court sitting in diversity applies state substantive law while keeping federal procedure, and with the constitutional reasoning behind that command. The hard work then lies in drawing the line between the two categories, and the Court drew it in stages: first a test keyed to whether the choice of rule would change the outcome, then a refinement that weighs countervailing federal interests, and finally a two-track framework that separates cases governed by a written Federal Rule from cases governed only by unwritten federal practice. The chapter closes with a modern application of that framework to a genuinely close conflict, and with the companion rule that tells a diversity court which state's law to borrow when a dispute touches more than one state.

The Erie Command

The starting point is a statute nearly as old as the federal courts themselves. The Rules of Decision Act, first enacted in 1789, directs that the laws of the several states, except where the Constitution, treaties, or Acts of Congress otherwise require or provide, shall be regarded as rules of decision in civil actions in the federal courts¹. For almost a century, however, the Court read the word "laws" in that Act to include state statutes but to exclude state judicial decisions on questions of general law². Under that reading, a federal court facing a common-law question with no state statute on point was free to exercise its own judgment about what the common law required, and the federal courts accumulated a body of federal general common law that often diverged from the decisions of the state courts sitting beside them.

The old regime failed on its own terms and produced a distinctive abuse. Its defenders had hoped that state judges would gradually fall in line behind the federal decisions, yielding a uniform national common law; instead the state courts kept their own views, so a single state contained two bodies of law — one applied between citizens of the same state, and another applied whenever diversity let a litigant reach federal court³. Because only some litigants had that choice, the applicable rule came to turn on citizenship rather than on the merits, and parties learned to manufacture the difference: in Black & White Taxicab & Transfer Co. v. Brown & Yellow Taxicab & Transfer Co., a Kentucky taxicab company dissolved and reincorporated in Tennessee for the sole purpose of creating diversity, then sued its Kentucky competitor in federal court to enforce an exclusive contract that Kentucky's own courts refused to enforce⁴. Forum choice, and even citizenship itself, had become instruments for selecting the governing law.

The Court ended the regime in Erie Railroad Co. v. Tompkins. Harry Tompkins, a Pennsylvania citizen, was walking along a well-worn footpath beside the Erie Railroad's tracks in Pennsylvania when something projecting from a passing freight train struck him. Under Pennsylvania decisions, a person using such a longitudinal path was a trespasser owed no duty beyond the avoidance of wanton or willful injury; under the more forgiving general law the federal courts had developed, he was a licensee owed ordinary care. Tompkins sued in a New York federal court precisely to capture that difference, and he won a verdict of $30,000 that the lower courts sustained on the theory that a railroad's duty was a question of general rather than local law⁵. The Supreme Court reversed and disapproved the entire doctrine: except in matters governed by the Federal Constitution or by Acts of Congress, the law to be applied in any case is the law of the state, whether declared by its legislature in a statute or by its highest court in a decision, and there is no federal general common law⁶.

The holding rests on two grounds that do different work. The constitutional ground concerns allocated power: Congress has no authority to declare substantive rules of common law applicable within a state, and no clause of the Constitution confers such a power on the federal courts, so a federal judge who fashioned a general rule of tort or contract for a diversity case was exercising a lawmaking power the federal government simply does not possess⁷. The equality ground concerns administration: when the governing rule depends on which courthouse hears the case, an out-of-state litigant able to invoke diversity gets a body of law unavailable to a local litigant on identical facts, and the system rewards forum manipulation rather than the merits. Erie removes that incentive at its source by making the substantive rule identical in the state and federal courthouses of the same state. The command carries a matching limit that follows from its logic: where the Constitution or a federal statute supplies the rule of decision — a claim arising under federal law — Erie has nothing to say, because there the federal government does possess the lawmaking power, and federal law governs the merits⁸.

The Outcome-Determinative Test

Erie settled who makes substantive law, but it left the boundary unmarked. Everyone agreed that a federal court keeps its own procedure, so the doctrine's reach turned on where substance and procedure divide. The Court confronted that boundary in Guaranty Trust Co. v. York, a diversity class action brought on the equity side of a New York federal court. York, who had received corporate notes as a gift, sued the indenture trustee for breach of trust in connection with an exchange offer for the notes; New York's statute of limitations would have barred the suit in state court, while traditional federal equity practice measured timeliness by the more flexible doctrine of laches, under which her claim might have survived⁹.

The Court held that the state limitations period controlled, and in doing so it replaced labels with a function. Categories like substance and procedure are drawn for many different purposes, Justice Frankfurter explained, and a statute of limitations can honestly be called procedural in some contexts and substantive in others; letting the label decide would let a federal court disregard a state rule that completely extinguishes the claim¹⁰. The question that matters for Erie is instead whether disregarding the state rule would significantly affect the result of the litigation — whether the outcome in federal court would be substantially the same as it would be in the state court a block away¹¹. This outcome-determination test follows directly from Erie's purpose, because diversity jurisdiction gives an out-of-state litigant a different tribunal, and York insisted that it does not give them a different body of law. A limitations bar is the paradigm case: with the state rule the claim is dead, and without it the claim proceeds, so applying federal practice would revive in federal court a claim that state law had extinguished, and every plaintiff holding a stale claim would head straight for the federal courthouse¹².

The test's strength is also its weakness. Read literally, almost any procedural difference can affect the result of some case: a shorter answer deadline, a different method of delivering papers, even the hour the clerk's office closes could each prove decisive in the case where a party misses the mark, and a literal reading would therefore require federal courts to abandon their own procedure piece by piece until nothing independent remained. The test cannot mean that much, and the Court has never applied it that way: purely housekeeping differences do not become state-law questions merely because some conceivable case could turn on them, and the inquiry asks whether the difference significantly affects the result, viewed from the perspective of a litigant deciding where to sue¹³. Stating that limit, however, is easier than administering it, and the next two decisions supply the machinery — one by weighing federal interests against the outcome effect, the other by rebuilding the test around Erie's actual purposes.

Balancing Federal Interests

The first correction came in Byrd v. Blue Ridge Rural Electric Cooperative, Inc., which holds that effect on outcome is one factor rather than the whole analysis¹⁴. Byrd, a North Carolina lineman employed by an independent contractor, was injured while connecting power lines for a South Carolina electric cooperative and sued the cooperative in tort. The cooperative's defense was that Byrd was its statutory employee under the state workers' compensation act, which would make compensation benefits his exclusive remedy and immunize the cooperative from suit. South Carolina practice assigned that factual question to the judge; federal practice, shaped by the Seventh Amendment's policy favoring juries, sends disputed questions of fact to the jury. Whether judge or jury decided could plausibly change who won, so under a literal reading of York the federal court would simply have copied the state's allocation¹⁵.

The Court instead structured the inquiry in three steps. The first asks whether the state rule is bound up with the definition of the state-created rights and obligations themselves — whether it is part of what the parties are owed. Rules of that kind bind the federal court absolutely, because they are the substantive law the Constitution reserves to the states, and no federal interest can outweigh them¹⁶. The second step applies when the state rule is one of mere form and mode of enforcing the right: there the federal court ordinarily conforms to state practice where the difference would bear substantially on the outcome, but it must weigh that concern against affirmative countervailing federal considerations. The federal courts are an independent system for administering justice, and some features of that system — most prominently the distribution of trial functions between judge and jury — are essential to its character, so a state rule of form and mode cannot commandeer them¹⁷. The third step tempers the outcome inquiry itself: what matters is the degree of certainty that a different result would actually follow, and the identity of the fact-finder affects outcomes far less predictably than a rule that bars the claim outright, particularly since a federal judge retains the power to comment on the evidence and to grant a new trial against the weight of the verdict¹⁸.

Applying those steps, the Court held that a jury, following federal practice, would decide Byrd's employment status. South Carolina's judge-decides rule had grown out of the practical habit of reviewing workers' compensation commission decisions without juries; nothing suggested the rule was integral to the immunity the statute created, so it was form and mode, the federal interest in the judge-jury relationship was strong, and the effect on outcome was speculative¹⁹. The contrast with York shows which facts carry the analysis: a limitations bar certainly determines who wins and implicates no comparable federal interest, so state law governs there, while a rule allocating a factual question between judge and jury only might change the result and collides with a structural feature of the federal system, so federal practice prevails²⁰. The balancing also protects something Erie never meant to surrender, because if any conceivable effect on outcome compelled conformity, the states could restructure the internal workings of the federal courts one procedural rule at a time.

The Hanna Framework

Byrd restored federal interests to the analysis, but it left courts balancing without a scale, and it never addressed the situation in which the federal practice is written down in a Federal Rule of Civil Procedure. Hanna v. Plumer supplied the modern organization²¹. An Ohio plaintiff sued the executor of a Massachusetts estate in federal court over an automobile accident, and she served process by leaving the summons and complaint with the executor's wife at his residence, exactly as Rule 4 then permitted; Massachusetts law demanded in-hand delivery to the executor, and by the time the objection was decided the limitations period had run, so the choice of service rule would decide the case²². From those facts the Court built the two-track sequence courts still run: first ask whether a federal statute or Federal Rule directly covers the disputed point; if one does, the analysis proceeds under the Rules Enabling Act, and only if none does must the court make the relatively unguided Erie choice, now governed by the refined test described below. The tracks are separate because the source of federal authority differs — a Federal Rule is federal law promulgated under a congressional delegation resting on Congress's power to regulate practice in the courts it created, so Erie's constitutional holding about substantive lawmaking simply does not speak to it²³.

A Federal Rule on Point

When a Federal Rule of Civil Procedure directly covers the disputed point, the Rule governs, state law notwithstanding, so long as the Rule is valid — meaning it really regulates procedure, the manner and means by which rights are enforced, and does not abridge, enlarge, or modify any substantive right²⁴. In Hanna itself the collision was unavoidable, because Rule 4 said with unmistakable clarity that in-hand service was unnecessary while Massachusetts said it was essential. Prescribing how a defendant learns that suit has been filed is procedure in the plainest sense, so the Rule was valid, and service that satisfied it sufficed even though a Massachusetts court would have dismissed the identical suit²⁵.

Each half of the validity standard polices the same boundary from a different side. The requirement that a Rule really regulate procedure keeps the rulemaking process inside the congressional delegation: Congress handed over authority for practice and procedure and kept for itself, and for the states, authority over the rights that litigation enforces, so a rule that altered the rules of decision by which courts adjudicate rights would exceed the grant. The proviso against abridging substantive rights guards the same line from the other direction, protecting state-created entitlements from erosion dressed up as court administration²⁶. The presumption of validity, meanwhile, is nearly conclusive for an institutional reason: every Rule passes through Advisory Committee drafting, Supreme Court approval, and a congressional waiting period before it takes effect, so a court asked to void one would be declaring that all of those actors erred in their considered judgment, and no Federal Rule of Civil Procedure has ever been invalidated on this ground²⁷.

The track opens only on a genuine collision, and that limit does real work. If the Rule can fairly be read not to reach the disputed question, there is no direct conflict, the Enabling Act analysis never begins, and the court proceeds down the unguided track instead, where the state rule gets its full weight²⁸. Courts have often taken that interpretive exit, reading Federal Rules narrowly to avoid displacing state rules that carry important state policies²⁹. In Walker v. Armco Steel Corp., for example, the Court read Rule 3 — which fixes when an action commences for purposes of the Federal Rules' own timing requirements — as saying nothing about when a state limitations period stops running, so the state rule requiring service within the limitations period governed and the late-served claim was barred. Whether a Rule is on point is therefore where the real fights occur, a battleground the final case in this chapter puts on full display.

The Twin Aims Test

Where no federal statute or Rule covers the point, the court is back in Erie territory, and Hanna refit York's test for that setting. Outcome-determination was never meant to be applied mechanically; it must be read in light of the twin aims of the Erie rule, which are the discouragement of forum shopping and the avoidance of inequitable administration of the laws³⁰. The operative question is whether disregarding the state rule would lead plaintiffs to choose the federal courthouse or would give diversity litigants a substantive advantage denied to in-state litigants on identical facts. If it would, state law applies; if the difference is too trivial to influence anyone's choice of forum, the federal court may follow its own practice³¹.

Each aim names a distinct mechanism. Forum-shopping deterrence is about incentives: if the federal courthouse offers a materially better rule, every plaintiff eligible for diversity will systematically select it, the applicable law will come to depend on the parties' citizenship rather than on the merits, and the system will have recreated the vice that discredited the general-common-law era³². Equitable administration is about fairness between litigants: an in-state plaintiff locked into state court and a diverse plaintiff holding a federal option should not face different substantive prospects on the same facts, because that difference is discrimination worked through the accident of citizenship³³.

Both aims are measured from the ex ante vantage of a litigant choosing a forum at the moment of filing, and that vantage point is what keeps the test from collapsing back into York's literal breadth. Judged at filing, only differences large enough to influence the forum decision count, and the countless procedural variations that matter only after litigation has gone wrong drop out of the analysis. The two anchor cases sort themselves accordingly. The limitations difference in York flunks the test, because a plaintiff whose claim is dead in state court but alive in federal court will obviously file federal, so the state rule binds³⁴. The service difference in Hanna passes it, because no rational plaintiff selects a forum over whether the summons may be left with a resident adult rather than pressed into the defendant's hand; that difference became outcome-determinative only in hindsight, after the plaintiff had failed to comply with the state method, and hindsight is the wrong perspective³⁵.

Hard Cases Under Hanna

The Enabling Act track works smoothly when the displaced state rule is as thoroughly procedural as the Federal Rule — no substantive policy rode on Massachusetts's in-hand service requirement. The framework is tested when a nominally procedural state rule carries genuine substantive policy, and Shady Grove Orthopedic Associates, P.A. v. Allstate Insurance Co. is the modern demonstration³⁶. New York required insurers to pay claims within thirty days and imposed statutory interest of two percent per month on overdue benefits, but a separate New York statute barred class actions to recover statutory penalties — a limit designed to keep aggregated penalty awards from crushing defendants. Shady Grove, a medical provider holding an assigned insurance claim worth roughly five hundred dollars in unpaid interest, sued Allstate in federal court and sought to represent a class of every provider Allstate had paid late, converting a five-hundred-dollar claim into an aggregate exposure of about five million dollars. Rule 23 permits certification whenever its criteria are met; New York's own courts would have refused the class outright³⁷.

A divided Court held that Rule 23 answers the question in dispute — whether a suit may be maintained as a class action — and therefore controls, allowing a class action in federal court that New York's courts would have rejected³⁸. The Justices split three ways, and the split maps the framework's pressure points. Justice Scalia's plurality asked only whether Rule 23, consulted on its face, really regulates procedure; because a class action merely enables a court to adjudicate many claims at once while leaving the parties' legal rights and the rules of decision unchanged, the Rule was valid in all applications, whatever its incidental effect on state-created rights³⁹. Justice Stevens, who supplied the fifth vote, agreed that Rule 23 applied here but rejected the face-of-the-Rule approach: in his view a Federal Rule cannot displace a state rule so intertwined with the state's substantive rights and remedies that it functions as substantive, and New York's bar escaped that description only because it was a general procedural limit that applied to penalty claims arising under any source of law⁴⁰. Justice Ginsburg's dissent would have read Rule 23 narrowly to avoid the collision altogether, treating New York's bar as a remedial limit on a defendant's maximum liability — the functional equivalent of a damages cap, which Erie would plainly require the federal court to respect⁴¹.

The practical lessons are structural. Once a valid Federal Rule is found to govern the question, the state policy loses no matter how weighty it is, and the twin aims never enter the analysis — even where, as in Shady Grove, the stakes make the outcome look intensely substantive and the decision hands class counsel an obvious reason to file in federal court to unlock a remedy the state courthouse denies⁴². The real battleground in a Rules-track case is therefore the interpretive first step: whether the Rule actually answers the disputed question and how broadly it should be read. Lower courts frequently treat Justice Stevens's intertwinement inquiry as the controlling opinion, so a litigant defending a state rule should be prepared to show that the rule defines the scope of a state-created right or remedy rather than merely regulating courtroom mechanics⁴³. The converse also remains true: where the Federal Rule can fairly be read not to address the question, no collision exists, and the conflict is analyzed under the twin aims⁴⁴.

Which State's Law Applies

One question remains after the vertical framework is complete. Erie directs the federal court to state law, but a diversity case by definition involves citizens of different states, and it may involve events in yet another, so some rule must select which state's substantive law governs. Every state maintains choice-of-law rules for exactly that purpose — doctrines that pick the governing law in disputes with multistate contacts — and the question for a federal court is whether it uses the forum state's conflicts rules or is free to devise its own. Klaxon Co. v. Stentor Electric Manufacturing Co. answered that a federal court sitting in diversity must apply the choice-of-law rules of the state in which it sits, because conflicts rules are themselves substantive for Erie purposes⁴⁵. In Klaxon, a New York corporation had won a $100,000 contract verdict in a Delaware federal court and moved to add prejudgment interest under a New York statute; the Third Circuit applied the New York statute because it considered that result the better view of the law of damages, without asking what Delaware would do. The Supreme Court reversed: the Delaware federal court had to ask how a Delaware state court, applying Delaware's own conflicts rules, would answer the question, and it could not engraft its preferred exceptions onto those rules⁴⁶.

The rule follows from Erie's own logic through two mechanisms. First, a choice-of-law rule is outcome-determinative in the strongest sense: it determines whose substantive law supplies the parties' rights, so changing the conflicts rule changes the governing law and often the result. If federal courts applied independent conflicts principles, the accident of diversity would again make the applicable law turn on the courthouse chosen — a plaintiff would pick federal court for a friendlier pointer to some other state's tort or contract rule — and the equal administration Erie promised within each state would collapse⁴⁷. Second, the rule respects state authority over the reach of state law. Within constitutional limits, each state decides whether a given matter is governed by its own law or by another state's, and a federal court that overrode the forum's conflicts rules would be contradicting the state on a question the state is entitled to answer⁴⁸.

The consequence students find counterintuitive is that Klaxon buys uniformity within each state at the price of uniformity across the federal system. The state and federal courthouses in the same city will now reach the same answer, which is the vertical uniformity Erie demands; but because different states use different conflicts rules, federal courts in different states may apply different substantive law to identical facts. Interstate forum shopping among federal districts therefore survives even though intrastate forum shopping is eliminated, and the Court has accepted that lack of horizontal uniformity as the deliberate price of a federal system that leaves each state free to pursue local policies diverging from its neighbors'⁴⁹. What the federal court may never do is treat itself as a neutral umpire empowered to select the law it considers best for a multistate dispute. Imagine a district judge who concludes that the jurisdiction named by the forum state's conflicts rule has no genuine interest in the case; that judge must follow the rule wherever it points anyway, because the freedom to improve on state conflicts law is precisely what Klaxon forecloses⁵⁰.