Federalism: The Case That It Changed Everything

Lessons 1 and 2 described the design and how it ran. Lesson 5 described how it ended. This lesson and the next are about consequences, and they are written as a pair. Here is the case that direct election changed the balance between the states and the national government, set out as its authors make it and with the evidence they rely on. Lesson 9 gives the evidence against and ends with a scorecard. Read both before you decide.

The claim

Madison wrote in Federalist No. 62 that letting state legislatures appoint senators gave the state governments an agency in forming the federal government and was a "convenient link between the two systems." The claim under study is that the Seventeenth Amendment cut the link, and that the cut mattered. Senators who owed their seats to a legislature had reason to protect the legislature's powers, its revenue and its freedom from federal direction. Senators who owe their seats to a statewide electorate have reason to deliver federal benefits to that electorate, whatever the effect on the state government's own powers. Over a century, the argument goes, that change in incentives let Congress grow into fields the states had run, pay for it with the income tax that arrived the same spring, and attach strings to the money it sent back.

Stated that way the claim is about incentives and structure rather than about any one law. Its authors do not say that the amendment caused the New Deal or the Great Society by itself. They say the amendment removed a brake, and that everything that followed rolled more easily without it. Keep that distinction in mind. A good deal of the dispute in the next lesson is about whether the brake was ever working.

Who makes the case, and where

The modern literature began in the 1990s, during the Court's federalism revival and the term limits movement. Every work below exists under the title, venue and year given; the citations were checked this month against the journal repositories, the publisher's record or the author's own publication list.

Author	Work	Venue and year	Core claim
Todd J. Zywicki	Senators and Special Interests: A Public Choice Analysis of the Seventeenth Amendment	Oregon Law Review, vol. 73, p. 1007 (1994)	Legislative election made senators the states' agents and made the two houses answer to different constituencies, which filtered interest group legislation. Direct election made the constituencies alike and gave organized interests a direct route to senators.
Todd J. Zywicki	Beyond the Shell and Husk of History: The History of the Seventeenth Amendment and Its Implications for Current Reform Proposals	Cleveland State Law Review, vol. 45, p. 165 (1997)	The traditional explanations for the amendment (deadlock, corruption, Progressive ideology) do not explain its passage; a public choice account does. Reform proposals of the 1990s should be judged in light of what the amendment removed.
C. H. Hoebeke	The Road to Mass Democracy: Original Intent and the Seventeenth Amendment	Transaction Publishers, 1995 (211 pp.)	The framers built the Senate as a deliberative check modeled in part on the House of Lords; the amendment grew out of a democratic ideology that treated every filter on popular will as illegitimate.
Vikram D. Amar	Indirect Effects of Direct Election: A Structural Examination of the Seventeenth Amendment	Vanderbilt Law Review, vol. 49, p. 1347 (1996)	Complaints about unfunded mandates and federal "conscription" of state agencies connect to the removal of legislatures from Senate selection. The amendment also changed separation of powers, for example by reducing rotation between the Senate and executive offices.
Jay S. Bybee	Ulysses at the Mast: Democracy, Federalism, and the Sirens' Song of the Seventeenth Amendment	Northwestern University Law Review, vol. 91, p. 500 (1997)	The framers, like Ulysses, bound the people to a structure that would resist the call of the moment. Direct election untied the knot and removed a structural protection for federalism that courts cannot fully replace.
Ralph A. Rossum	Federalism, the Supreme Court, and the Seventeenth Amendment: The Irony of Constitutional Democracy	Lexington Books, 2001	The framers protected federalism through structure, above all the mode of electing senators. The people removed that structure in 1913. The irony is a Court that now enforces federalism limits the Constitution's amended structure no longer supports.

Two things about this list. First, these are mostly law review articles and books from constitutional scholars, and they argue from structure, history and incentives. The political science studies that measure Senate behavior before and after 1913 are a different literature, and most of them come in the next lesson. Second, the authors disagree with one another about remedies. Zywicki's 1997 article evaluates term limits and other reform ideas against the lost structure. Bybee is skeptical that repeal is realistic. Rossum's target is the Court, and his conclusion points toward judicial restraint rather than repeal. Treat "the case that it changed everything" as a family of arguments, not a single program.

The title of Rossum's book names the irony that gives this whole debate its edge. If the people deliberately removed the structural protection for the states, a Court that invents judicial substitutes is overriding a constitutional choice. If the people did not understand what they were removing, the Court may be the only guardian left. Every author in the table has to pick a side of that fork.

Zywicki's public choice account in more detail

Zywicki's 1994 article is the one most often cited, so its logic deserves a paragraph. Interest groups seek legislation that benefits them at general expense. A legislature with two houses chosen by different constituencies is harder to capture, because a bill must satisfy both. Before 1913 a group that wanted federal legislation had to persuade the House, chosen by voters in districts, and the Senate, chosen by state legislators who had their own institutional interest in keeping power and revenue at the state level. After 1913 both houses answered to the same statewide and district electorates, and the same campaign donors could reach both. Zywicki's review of Hoebeke in The Independent Review (Winter 1996/97) restates the argument compactly: senators had been seen as, and acted as, the states' ambassadors to the federal government; the two-house system filtered special-interest legislation; direct election made the chambers' constituencies more alike and gave organized interests a direct route to senators, which he argues fueled federal expansion in the twentieth century.

His 1997 Cleveland State article asks a prior question: why did state legislatures give away a power? He finds the standard answers incomplete. Deadlocks and bribery were real but limited (lesson 3 has the counts), and Progressive ideology does not explain why only the Senate was changed. His alternative, in the article's own words from its abstract, is "rooted in a public choice analysis of constitutional change": the legislators who voted to ratify had reasons of their own, including freeing their sessions from Senate fights and shifting blame for unpopular federal choices to a statewide vote. Whatever one thinks of that account, it has the merit of taking the ratifiers seriously as actors with interests.

The mechanisms

An argument about structure has to say how the structure did its work. The literature names six channels. Each is stated here as its proponents state it, with the law and the numbers that go with it.

1. The end of the agent

Under the original design the legislature could instruct a senator, and senators sometimes resigned when instructed (lesson 2 gives verified examples). The legislature could also decline to reelect. A senator therefore had a principal whose interests were institutional: the state's taxing power, its police power, its control over local government. After 1913 the senator's principal became the state's voters, whose interests are personal and immediate. A voter who gets a federal highway, a hospital grant or a farm payment is pleased regardless of what the grant does to the state budget office's freedom. The amendment, on this view, replaced a principal who cared about federalism with one who does not, and then gave the senator six years and a large state to campaign in, which made money and national party organizations more important than any legislator's goodwill.

2. Spending

The federal government in fiscal 1913 spent $715 million, according to OMB Historical Table 1.1. OMB's series for outlays as a share of GDP begins in 1930, when the figure was 3.4 percent. The table below gives the decades since; every number comes from Table 1.2 and Table 12.1 of the FY 2027 Historical Tables, downloaded from OMB this month.

Fiscal year	Federal outlays, percent of GDP (Table 1.2)	Grants to state and local governments, millions (Table 12.1)	Grants as percent of federal outlays	Grants as percent of GDP
1930	3.4	not reported (series begins 1940)
1940	9.6	872	9.2	0.9
1950	15.3	2,253	5.3	0.8
1960	17.3	7,019	7.6	1.3
1970	18.7	24,065	12.3	2.3
1980	21.2	91,385	15.5	3.3
1990	21.2	135,325	10.8	2.3
2000	17.7	285,874	16.0	2.8
2010	23.2	608,390	17.6	4.1
2020	30.5	829,093	12.7	3.9
2025	23.1	1,214,777	17.3	4.0

Proponents read this table as the shape of the problem. A federal government that spent under a tenth of the economy before the Second World War has spent between a sixth and a quarter of it since 1950, and a rising share of that spending is money sent to the states with instructions. Grants were under one percent of GDP through 1950 and have been near or above four percent since 2010. The next lesson will point out what else happened in those decades. For now, note what the table does and does not show: it shows growth, and it shows the growth of grants as a tool, and it says nothing by itself about why.

3. Conditions on the money

Money sent to the states is the lever the argument cares about most, because a condition on a grant lets Congress reach a subject it could not regulate directly. The Court approved the lever in South Dakota v. Dole, 483 U.S. 203 (June 23, 1987), 7 to 2. Congress had told states to adopt a drinking age of 21 or lose about 5 percent of their federal highway funds. The Court said Congress could not set a drinking age directly. It could attach the condition all the same, provided the spending served the general welfare, the condition was unambiguous, it related to the federal interest in the program, and no other constitutional provision barred it. The opinion allowed that a financial inducement could become so large that "pressure turns into compulsion," and found 5 percent well short of that.

Twenty-five years later the Court found the compulsion point. In NFIB v. Sebelius, 567 U.S. 519 (June 28, 2012), the Affordable Care Act told states to expand Medicaid or lose all of their existing Medicaid funding. The Chief Justice wrote that Medicaid funds were "over 10 percent of a State's overall budget," and that threatening the whole of it was coercion. Justices Breyer and Kagan joined him on that point and the four dissenting justices agreed the condition was unconstitutional, so seven justices found coercion. The remedy left the expansion in place as a choice. For the argument in this lesson, Dole and NFIB together mark how far conditional spending had come: from 5 percent of highway money to the whole of the largest grant program, with the Court drawing a line only at the extreme.

4. Unfunded mandates

A mandate orders state or local governments to act at their own expense. Complaints about them peaked in the early 1990s, which is the period in which this scholarship was written, and Congress responded with the Unfunded Mandates Reform Act of 1995, Public Law 104-4, signed March 22, 1995. Its stated purposes include ending the imposition, without full consideration by Congress, of "Federal mandates on State, local, and tribal governments without adequate Federal funding." The Act requires the Congressional Budget Office to estimate the cost of any intergovernmental mandate in a reported bill that would reach $50 million a year (adjusted for inflation) and creates a point of order against bills that lack the estimate. It does not forbid mandates. Amar's 1996 article ties the mandate complaint directly to the amendment: a Senate chosen by legislatures would not have voted to spend the legislatures' money for them. Note that UMRA came from a Congress elected under direct election, two years after the complaint peaked, which is a fact both sides use.

5. Preemption

Under the Supremacy Clause a valid federal law displaces conflicting state law. The Congressional Research Service's Federal Preemption: A Legal Primer (updated May 18, 2023) sorts the doctrine into express preemption, where the statute says so, and implied preemption, which covers field preemption (federal law occupies a subject, as the Court has held for some areas of grain warehousing and immigration) and conflict preemption (compliance with both is impossible, or state law obstructs the federal purpose). The structural argument is simple: a Senate of the legislatures' agents would have written fewer preemption clauses and narrower ones. The evidence for the claim is mostly the growth in federal regulatory statutes since the 1960s. Counting preemptive statutes precisely is hard, and the counts in older reports could not be verified for this lesson, so the claim rests here on the doctrine's reach rather than on a number.

6. Commandeering

The last channel is the federal government ordering state governments to do its work. Two cases define the limit the Court eventually drew. New York v. United States, 505 U.S. 144 (June 19, 1992), struck down the "take title" provision of the Low-Level Radioactive Waste Policy Amendments Act, which told states to either regulate waste as Congress directed or take ownership of it. Justice O'Connor's opinion, 6 to 3 on that point, held that Congress may regulate individuals and may encourage states with conditional spending or a choice between federal and state regulation, but may not "commandeer the legislative processes of the States." Printz v. United States, 521 U.S. 898 (June 27, 1997), extended the rule to state executive officers. The Brady Act's interim scheme required local chief law enforcement officers to run background checks on handgun buyers. Justice Scalia's opinion for five justices held that the federal government "may not compel the States to enact or administer a federal regulatory program." Amar's word for this practice, "conscription," appears in his 1996 abstract, a year before Printz.

The Court's turn, and the turn back

The structural argument gained force from an unlikely source: the Supreme Court's own account of how federalism is protected. In 1954 Herbert Wechsler of Columbia published The Political Safeguards of Federalism in the Columbia Law Review (vol. 54, p. 543). His thesis was that the states are protected by their role in composing the national government: the Senate's equal representation, House districts drawn by states, the Electoral College and the parties. Courts therefore need not patrol the boundary of federal power; the political process does it. Wechsler could write, as Justice Powell's dissent in Garcia later quoted him, that national action was still regarded as "an intrusion to be justified by some necessity." He wrote forty-one years after the Seventeenth Amendment and treated the Senate's state basis as intact.

The Court tried both approaches within a decade. In National League of Cities v. Usery, 426 U.S. 833 (June 24, 1976), 5 to 4, it held that Congress could not apply the Fair Labor Standards Act's wage and hour rules to state and local employees performing traditional governmental functions such as police, fire and sanitation. Nine years later, in Garcia v. San Antonio Metropolitan Transit Authority, 469 U.S. 528 (February 19, 1985), also 5 to 4, it overruled that decision. Justice Blackmun's majority adopted Wechsler's view: the states' sovereign interests are protected by procedural safeguards inherent in the structure of the federal system rather than by judicially created limits on federal power.

What makes Garcia central to this lesson is that both the majority and the dissent talked about the Seventeenth Amendment. The majority conceded that structural changes had occurred since 1789, "not the least of which has been the substitution of popular election of Senators," and that these changes might alter the states' influence in the federal political process, then held that the process still protected the states well enough. Justice Powell's dissent answered that Wechsler's assumptions no longer matched reality. He quoted the Advisory Commission on Intergovernmental Relations' 1984 report Regulatory Federalism for the proposition that structural and political changes had made Congress "particularly insensitive to state and local values," and, citing the same report, he named the adoption of the Seventeenth Amendment, the weakening of political parties on the local level and the rise of national media as the changes that had made Congress "increasingly less representative of state and local interests." The dissent in Garcia is the argument of this lesson in judicial form, written before any of the law review articles in the table above.

The dissenters' view then won a series of cases without Garcia being overruled. New York (1992) drew the anti-commandeering line. United States v. Lopez, 514 U.S. 549 (April 26, 1995), struck down the Gun-Free School Zones Act as beyond the commerce power, the first such holding since the 1930s; Justice Kennedy's concurrence spoke of the states' role as "laboratories for experimentation." Printz (1997) extended anti-commandeering to executive officers. NFIB (2012) found a spending condition coercive for the first time. In none of these did the Court say the Seventeenth Amendment required the result, but the premise of each is the one Powell stated in 1985: the political process alone does not protect the states, so the Court must.

Case	Date	Vote	Holding	Role in the argument
NLRB v. Jones & Laughlin Steel, 301 U.S. 1	April 12, 1937	5 to 4	Congress may protect organizing at a steel plant because a strike there would disrupt interstate commerce	The commerce power widens; the Senate of 1937 was the first fully directly elected one
Wickard v. Filburn, 317 U.S. 111	November 9, 1942	9 to 0	Congress may regulate wheat grown for home consumption because of its aggregate effect on the market	The high-water mark of the aggregation theory
National League of Cities v. Usery, 426 U.S. 833	June 24, 1976	5 to 4	FLSA cannot be applied to states' traditional governmental functions	Judicial protection of the states
Garcia v. SAMTA, 469 U.S. 528	February 19, 1985	5 to 4	Overrules National League of Cities; the political process protects the states	Adopts Wechsler; majority and dissent both discuss the Seventeenth Amendment
South Dakota v. Dole, 483 U.S. 203	June 23, 1987	7 to 2	Congress may condition 5 percent of highway funds on a 21 drinking age	Conditional spending approved with limits
New York v. United States, 505 U.S. 144	June 19, 1992	6 to 3	Congress may not commandeer state legislatures	Judicial limits return
United States v. Lopez, 514 U.S. 549	April 26, 1995	5 to 4	Gun-Free School Zones Act exceeds the commerce power	First commerce-power limit since the 1930s
Printz v. United States, 521 U.S. 898	June 27, 1997	5 to 4	Congress may not command state officers to administer a federal program	Anti-commandeering extended
NFIB v. Sebelius, 567 U.S. 519	June 28, 2012	7 to 2 on coercion	Threatening all existing Medicaid funds is coercive; the expansion becomes optional	First spending condition held coercive

Rossum's book takes the position that this revival is the wrong answer to a real problem. If the people chose in 1913 to protect the states through elections rather than through structure, the Court is not entitled to restore by doctrine what the amendment removed. Larry Kramer's Putting the Politics Back into the Political Safeguards of Federalism (Columbia Law Review, vol. 100, p. 215, 2000) reaches a similar conclusion about the Court from the other direction: the real safeguard was never the mode of Senate selection but the political parties, which still tie national officials to state and local organizations. David Schleicher's 2014 Hastings Law Journal article (vol. 65, p. 1043) argues that national parties have not in fact acted as safeguards, and that the fix lies in election law rather than in constitutional structure. Those are answers to this lesson's case, and they get their full hearing in lesson 9.

The Sixteenth Amendment and the spring of 1913

Writers on both sides notice the calendar. The Constitution Annotated records that Congress proposed the Sixteenth Amendment on July 12, 1909, that ratification was completed on February 3, 1913, and that Secretary of State Knox certified it on February 25, 1913 (37 Stat. 1785). The Seventeenth was proposed May 13, 1912, completed April 8, 1913, and certified by Secretary Bryan on May 31, 1913 (38 Stat. 2049). Within fourteen weeks the national government gained a revenue source not tied to apportionment among the states and lost the states' agents in the Senate. For the structural argument, that pairing is the whole story in miniature: the means to spend and the removal of the body most likely to object.

Two cautions, which the next lesson develops. The amendments were separate proposals with separate ratifying coalitions, three years apart in Congress; the Sixteenth drew support from states that never ratified the Seventeenth. And the income tax produced little revenue until the First World War. Table 1.1 shows outlays of $715 million in 1913 and $713 million in 1916; they reached $18,493 million in 1919. The spring of 1913 made the growth possible. The war made it happen.

The case at its strongest

Put together, the argument runs like this. The Constitution protected the states with structure, not with words: Article I, Section 3 gave the state legislatures a standing agent in the national legislature, and Madison said so in Federalist 62. Interest groups and national majorities that wanted federal action had to get past a chamber whose members answered to the governments they proposed to displace. In 1913 that chamber's electors changed. Within a generation the commerce power was read to reach a farmer's own wheat (Wickard, 1942), within two the federal government was spending a fifth of GDP, and by the 1980s it was sending a sixth of its outlays to the states with conditions attached, mandating state spending without paying for it, preempting state law across whole fields, and ordering state officers to run federal programs. The Supreme Court first announced that the political process would protect the states (Garcia), then, finding that it did not, built judicial substitutes case by case (New York, Lopez, Printz, NFIB). On this view the Court is doing badly and inconsistently what the Senate once did by its composition.

The honest limit of the case is that it is a structural inference supported by a chronology. Its authors show that the brake was removed and that the car sped up. Showing that the brake was what had been holding the car is harder, and that is the job the political scientists in the next lesson took on.

Questions and answers

Did the Supreme Court ever say the Seventeenth Amendment weakened federalism?

Not as a holding. The closest it came is Garcia (1985), where the majority acknowledged that the substitution of popular election of senators might alter the states' influence in the federal political process, and Justice Powell's dissent listed the amendment among the changes that had made Congress less representative of state interests. The anti-commandeering and spending cases that followed rest on the Tenth Amendment and the structure of the Constitution generally, not on the Seventeenth.

Is Zywicki's 1994 article in the Northwestern University Law Review?

No. It is in the Oregon Law Review, volume 73, page 1007 (1994), as his own 1996/97 review of Hoebeke cites it. Bybee's article is the Northwestern one: volume 91, page 500 (1997), as his UNLV publication list confirms. The two are often paired because Zywicki draws on Bybee's structural account.

How much did the federal government spend in 1913 compared with today?

OMB Table 1.1 gives outlays of $715 million in fiscal 1913 and $7,011,105 million in fiscal 2025. Raw dollars are a poor comparison across 112 years of inflation and growth; use Table 1.2's percent of GDP, which runs from 3.4 percent in 1930 to 23.1 percent in 2025. There is no official GDP share for 1913 in the OMB series.

What exactly is a "condition" on a grant, and why can Congress attach one?

A condition is a requirement a state accepts in exchange for federal money, such as the 21 drinking age in Dole. Congress may attach conditions under the Spending Clause even on subjects it could not regulate directly, provided the four Dole tests are met and the inducement does not cross into compulsion, the line NFIB found when all existing Medicaid funding was at stake.

Does the Unfunded Mandates Reform Act stop unfunded mandates?

No. Public Law 104-4 requires CBO cost statements for intergovernmental mandates at or above $50 million a year and creates procedural hurdles. Congress can still pass a mandate by voting to do so. Its passage in 1995, by a directly elected Senate, is cited by one side as proof the political process works and by the other as proof the problem was real enough to need a statute.

Why does the argument care about the Senate's constituency being "like the House's"?

Because the bicameral filter depends on difference. If the two houses answer to the same voters and donors, a measure that pleases those voters and donors passes both. Zywicki's 1994 article argues that legislative election gave the Senate a constituency with institutional interests opposed to federal growth, and that direct election removed the difference. The counter, in lesson 9, is that state legislators were already voting their party and their voters' national preferences by the 1890s.

Where does Rossum part company with Zywicki and Bybee?

On what to do. Zywicki's 1997 article weighs reform proposals against the lost structure, and Bybee's article treats the amendment as a mistake whose effects courts can only partly offset. Rossum's book argues that because the people removed the structural safeguard, the Court should stop substituting judicial doctrines for it; the "irony" of his subtitle is a Court defending a federalism the amended Constitution no longer builds in.

What did the ACIR report that Justice Powell quoted say?

The Advisory Commission on Intergovernmental Relations, a federal advisory body, published Regulatory Federalism: Policy, Process, Impact and Reform in 1984. Powell's dissent in Garcia quoted it for the view that structural and political changes had made Congress "particularly insensitive to state and local values," and he named the Seventeenth Amendment, weaker local parties and national media as those changes. The report itself was not retrieved for this lesson; the quotation is taken from the opinion.

Hard questions

Isn't this just a story told backward? The federal government grew, the amendment came first, so the amendment must have caused it.

That is the strongest objection, and the authors know it. Their answer is that they are not relying on timing alone but on a mechanism: a change in who senators answer to, and therefore in what they vote for. The mechanism is plausible and the chronology fits. What the case lacks on its own terms is a measurement showing senators protected state governments before 1913 and stopped afterward. The studies that try to measure exactly that are in lesson 9, and their results cut in more than one direction.

If the Senate was such a brake, why did the states themselves ratify the amendment in eleven months?

Zywicki's 1997 article treats this as the central puzzle and answers it with public choice: legislators gained freedom from Senate fights, shed responsibility for unpopular choices, and faced voters who had already been given advisory primaries in 29 states by 1912 (lesson 4). Schleicher's 2014 article offers a different answer: national parties had made state legislative elections turn on national issues, and legislators wanted their own elections back. Either way, the states' consent is a fact the structural argument must absorb, and Rossum's book treats it as the reason courts should not undo the people's choice by doctrine.

The Court protects the states now. Doesn't that make the loss of the Senate brake moot?

Only if you think the Court is a good substitute. The record since 1976 is a 5 to 4 decision overruled by a 5 to 4 decision nine years later, followed by a line of 5 to 4 and 6 to 3 rulings that draw lines (commandeering, coercion, non-economic activity) without touching the bulk of federal spending and preemption. Bybee's point is that structural protection works every day in every vote, while judicial protection works at the margin in the cases that reach the Court. Rossum's point is that the substitution is also illegitimate. Both see the Court as a poor replacement, for different reasons.

Weren't the senators chosen by legislatures the ones who blocked direct election for twenty years and sat in what the press called the Millionaires' Club? Why trust that body to protect anything but itself?

The argument does not require the pre-1913 Senate to have been admirable. It requires only that its members' incentives ran toward protecting state governments' powers, and a wealthy senator who owed his seat to a legislature had that incentive whatever else he was. Lesson 3 records the bribery cases and the Senate's own vote of July 13, 1912, 55 to 28, declaring Lorimer's election invalid (he was not expelled, which would have needed two-thirds; the Senate used its Article I, Section 5 power by majority). The fair statement is that the old system produced both a structural check and real corruption, and that the amendment traded away the first to get rid of the second. Whether that was a good trade is the debate.

The spending table starts in 1930. Isn't it convenient that the federalism argument's own data skip the first seventeen years after the amendment?

The gap is OMB's, not the argument's: the official GDP-share series begins in 1930. Table 1.1 covers the earlier years in dollars, and it shows outlays flat from 1913 ($715 million) to 1916 ($713 million), a war spike to $18.5 billion in 1919, and a return to about $3 billion a year in the 1920s. Those figures help the other side more than this one, since they show that the first direct-election Congresses did not spend more until the war forced them to. A careful version of the structural argument says the amendment removed a constraint whose absence mattered once the demand for federal action arrived, in 1917 and again in 1933.