Federalism: The Case That It Changed Little

Lesson 8 gave the structural argument that direct election broke the states' hold on the national government. This lesson gives the evidence that the hold was weaker than the argument assumes, that 1913 changed less than the text suggests, and that the growth of federal power has causes that do not run through the Senate. It ends with a scorecard. The two lessons are meant to be read together; neither side gets the last word alone.

What the counter-case claims

The counter-case does not say the Seventeenth Amendment did nothing. Several of its own studies find that senators behave differently under direct election. It says something narrower and, if true, more damaging to the structural argument: that state legislatures in the decades before 1913 were not acting as guardians of state power when they chose senators, so taking the choice away from them removed a guardian who was not on duty. If the legislators voted their party label, followed their voters' advisory primaries, took money from the candidates, and served one term before going home, then the "agent of the states" was already an agent of the parties and the voters. On that reading the amendment ratified a change that had already happened, and the federal growth of the next century has to be explained by things that actually happened in that century.

How indirect election really worked

Schiller and Stewart's data

The most important work on this question is Wendy J. Schiller and Charles Stewart III, Electing the Senate: Indirect Democracy before the Seventeenth Amendment (Princeton University Press, 2015; published December 2014). They and Benjamin Xiong published the first results in the Journal of Politics in July 2013 (vol. 75, no. 3, pp. 835 to 847), and the two authors wrote a public summary for Brookings in May 2013, The 100th Anniversary of the 17th Amendment: A Promise Unfulfilled? The figures below come from that Brookings paper, which is free and gives the numbers plainly.

Measure, 1871 to 1913	Finding
Elections studied	752 Senate elections by state legislatures, 577,042 recorded roll-call votes, rosters of about 106,000 state legislators
Party control of the outcome	In 98 percent of elections the winner belonged to the party with the legislative majority
Incumbents	461 elections (61 percent) had an incumbent seeking reelection; 291 (39 percent) were open seats; 323 incumbents (70 percent) were reelected
Legislators' tenure	More than two-thirds served a single term, usually two years
Winners' backgrounds	About 65 percent of first-time winners held elective office immediately before; over 80 percent of all winners had prior elective experience
Money	Authors' estimate: an average of $12.64 million (2012 dollars) per election, about $214.88 million a year across an average of 17 elections; the 1897 Oregon contest alone involved $305,000 (about $8.47 million in 2012 dollars)
Party data available	Complete party identification for 28 percent of elections (208), partial for 54 percent (409), none for 18 percent (135)

Read the table against the structural argument's picture of legislators as institutional principals with a stake in state power. The legislators who chose senators mostly served two years and left. They voted their party label almost without exception. The contest that mattered happened inside the majority party's caucus, and the Journal of Politics article finds that conflict was most likely when the caucus nominating process broke down, when the majority coalition was small, and when an incumbent was seeking reelection. Money moved. Candidates for the Senate recruited and financed candidates for the legislature. None of that describes a body deliberating about the proper reach of the national government.

Haynes had seen the same thing from inside the period. In The Election of Senators (1906), after his table of deadlocks, he wrote that statistics could not reveal "the extent to which the party caucus often dominates the whole situation," that the caucus was unknown to the law and met behind closed doors, and that many an election recorded as a first-ballot landslide had been preceded by a long and bitter campaign inside the party.

What Schiller and Stewart found did and did not change after 1913

The Brookings paper's title has a question mark for a reason. Comparing the indirect era with the direct era, the authors report that party control shifted toward the Democrats, that senators stayed longer, and that roll-call voting moved toward the moderate views of state electorates. Those are real changes. On the other side of the ledger: candidates' backgrounds looked much the same (experienced officeholders then and now), money was central in both eras, and partisanship was strong in both. In 2012, 91 percent of incumbents who ran were reelected (against 70 percent in the indirect era) and exit polls in 14 states found 93 percent of Democrats and 87 percent of Republicans voting the party line. Senators from ten sample states averaged 76 bills and 6 committees before the amendment and 35 bills and 4 committees from 1993 on, a gap the authors attribute largely to fewer committees and to the pension and relief bills common in the earlier period. Their verdict is that the amendment "has failed to deliver on its promise" of a more responsive and accountable Senate. That is a judgment about democracy, and it is relevant here because the same evidence shows the old Senate was already a partisan, money-driven, electorally minded body. The structural argument needs it to have been something else.

The Oregon System and the 29 states

The Senate Historical Office's Seventeenth Amendment page reports that by 1912 as many as 29 states chose senators through party primaries or general elections, after which the legislature ratified the voters' choice. Oregon's laws of 1904 and 1908, with the Statement No. 1 pledge that bound legislative candidates to vote for the people's choice, are described in lesson 4. In those 29 states the amendment did not take a decision away from the legislature. The legislature had already given it away, by statute or by party practice. That matters for the structural argument because it shrinks the before-and-after difference to the 19 states that had not adopted a popular choice, many of them in the South and Northeast, by the time the amendment went to the states.

The empirical studies

The political science literature measures what the law reviews infer. Each study below exists under the title, journal, volume and year given; the Crossref records and abstracts were checked this month. For each, the table gives what was measured, the period, and the result in the authors' terms.

Study	What it measured	Period	Result
Sara Brandes Crook and John R. Hibbing, A Not-so-distant Mirror: The 17th Amendment and Congressional Change, American Political Science Review 91(4): 845-853 (December 1997)	Variables on the Senate's membership (who served) and its responsiveness, before and after direct election, with House data as a control for history effects	Decades around 1913	Changing the mode of selection "did indeed lead to alterations in the composition and sensitivity of the Senate." The authors draw a lesson of caution for 1990s electoral reforms: reforms do have effects.
William Bernhard and Brian R. Sala, The Remaking of an American Senate: The 17th Amendment and Ideological Responsiveness, Journal of Politics 68(2): 345-357 (May 2006)	W-NOMINATE roll-call scores of major-party senators; a selection model treating the decision to seek reelection and late-term voting as linked	1877 to 1932	More incumbents sought reelection once legislatures no longer controlled the outcome, and post-amendment senators, especially Republicans, moderated their positions as elections approached. The paper opens by noting earlier scholarship had treated the amendment as having little effect.
Scott R. Meinke, Institutional Change and the Electoral Connection in the Senate: Revisiting the Effects of Direct Election, Political Research Quarterly 61(3): 445-457 (2008, online October 2007)	Bill sponsorship, roll-call participation and party voting, comparing senators originally elected by voters with those originally chosen by legislatures while both served side by side	Decades around ratification	"Increased sponsorship and participation among originally elected senators," more constituency bill sponsorship just before reelection, and a stronger link between state partisanship and senators' party-support voting after the amendment.
Sean Gailmard and Jeffery A. Jenkins, Agency Problems, the 17th Amendment, and Representation in the Senate, American Journal of Political Science 53(2): 324-342 (2009)	A principal-agent model of indirect election by an expert intermediary versus direct election by an inexpert mass electorate, tested statistically with the amendment as a quasi-experiment and the House as a control	Before and after 1914	Senators became "more responsive to the mass electorate under direct election" and also gained more discretion to pursue their own ends; both predictions were supported. No similar change appeared in the House.
Wendy J. Schiller, Charles Stewart III and Benjamin Xiong, U.S. Senate Elections before the 17th Amendment: Political Party Cohesion and Conflict 1871-1913, Journal of Politics 75(3): 835-847 (2013)	Recorded ballots in state legislative Senate elections	1871 to 1913	Elections were often contentious; a legislative majority did not guarantee an easy result; conflict rose when caucus nominations broke down, when majorities were small, and when an incumbent ran.
Daniel Wirls, Regionalism, Rotten Boroughs, Race, and Realignment: The Seventeenth Amendment and the Politics of Representation, Studies in American Political Development 13(1): 1-30 (1999)	The politics of the amendment's passage: regional and partisan alignments in Congress and the states	1890s to 1913	Treats the amendment as a product of regional realignment and representation politics rather than of a national verdict on federalism. (The abstract was not retrievable this month; the citation is confirmed and the summary here is kept general.)

Notice what these results have in common. Every study that looks for a behavioral change finds one, and the change is in the direction the amendment's supporters promised: senators paying more attention to voters. Crook and Hibbing say the Senate's composition and sensitivity changed. Bernhard and Sala find moderation toward the state electorate. Meinke finds more sponsorship and participation. Gailmard and Jenkins find more responsiveness to the mass public. These findings cut against the claim, found in older political science and in Schiller and Stewart's own framing, that the amendment changed little about the Senate.

Now notice what none of them measures. No study in the table tests whether senators voted to protect state governments' powers, revenue or autonomy before 1913 and stopped afterward. The dependent variables are ideology, participation, sponsorship, reelection and party loyalty. Gailmard and Jenkins come closest by modeling the legislature as an "expert" principal, but the expertise in their model is about judging the senator, not about defending federalism. So the empirical literature establishes that the amendment changed senators' relationship with voters. It leaves the federalism claim where lesson 8 left it: an inference from structure. That is the fair reading, and it is why the scorecard below gives neither side a clean win on the central question.

What else happened: the alternative explanations for federal growth

The structural argument's table of outlays and grants (lesson 8) shows a government that grew. The same OMB tables show when it grew, and the dates do not cluster around 1913.

Fiscal year	Outlays, millions of current dollars (Table 1.1)	Outlays, percent of GDP (Table 1.2)	What was happening
1913	715	not in series	Seventeenth Amendment certified May 31
1916	713	not in series	Second year of all-direct Senate elections; spending unchanged
1917	1,954	not in series	United States enters the First World War
1919	18,493	not in series	War spending peaks
1925	2,924	not in series	Postwar retrenchment
1929	3,127	not in series	Last pre-Depression year
1930	3,320	3.4	GDP series begins
1933	4,598	7.9	Depression; New Deal begins
1936	8,228	10.3	New Deal spending
1937			NLRB v. Jones & Laughlin (April 12) widens the commerce power
1939	9,141	10.1	Eve of war
1942			Wickard v. Filburn (November 9) adopts the aggregation principle, 9 to 0
1943	78,555	42.6	Second World War
1944	91,304	42.7	Wartime peak
1945	92,712	41.0	War ends
1948		11.4	Postwar low, still above any prewar year
1965		16.7	Great Society programs begin
1983		22.9	Recession high
2009		24.3	Financial crisis
2020	6,516,817	30.5	Pandemic
2025	7,011,105	23.1	Latest actual

The pattern is familiar to anyone who studies public finance: a ratchet driven by crises. Spending leaps in a war or a depression and never returns all the way to where it started. The 1920s got back to about $3 billion a year, roughly four times the 1913 level in a larger economy. After 1945 the floor was over 11 percent of GDP, where before 1933 it had never reached 8. Grants to the states, which were 0.9 percent of GDP in 1940 and 0.8 percent in 1950, did not reach 2 percent until about 1970, after the Great Society programs of the mid-1960s, and reached 4 percent only in 2010. Every step on that ladder has a proximate cause with a name and a date, and none of those causes is the mode of Senate election.

The Court's role in this growth is also independent of the Senate. NLRB v. Jones & Laughlin Steel (1937), 5 to 4, let Congress reach labor relations at a steel plant because a strike would disrupt interstate commerce. Wickard v. Filburn (1942), unanimous, let it reach wheat grown and eaten on one farm because of the aggregate effect of all such wheat. Those two decisions, more than any statute, opened the regulatory field the states had occupied. The justices who wrote them were confirmed by a directly elected Senate, which is a point for the structural side, but nothing in either opinion depends on how senators were chosen.

What the Senate still does for the states

The structural argument sometimes reads as if the Senate stopped being a states' body in 1913. Four features survive.

1. Equal suffrage. Each state still has two senators regardless of population, and Article V provides that no state, without its consent, "shall be deprived of its equal Suffrage in the Senate." The Seventeenth Amendment left that untouched, and the equal-suffrage clause is the one part of the Constitution that even the ordinary amendment process cannot reach over a state's objection. The smaller states' weight in the Senate is the same as it was in 1912.
2. Statewide constituencies. Senators are elected by the people of a state, not by a national electorate or by districts. Their political interest in bringing federal money, projects and favorable rules to their state is as strong as it ever was; what changed is that the state's voters, rather than its legislators, judge the result.
3. Daily advocacy for state governments. Governors, legislatures, attorneys general and state agencies lobby their senators constantly over grant formulas, waivers, mandates and preemption clauses. The Unfunded Mandates Reform Act of 1995 came from exactly that pressure, two years after the mandate complaint peaked.
4. The parties. Larry Kramer's Putting the Politics Back into the Political Safeguards of Federalism (Columbia Law Review 100: 215, 2000) argues that Wechsler was right about the conclusion and wrong about the mechanism: the real safeguard was never the mode of Senate selection or the Electoral College but the decentralized party system, which made national officials' careers depend on state and local party organizations. On Kramer's account the Seventeenth Amendment removed a formal link while the informal one carried on.

David Schleicher's 2014 Hastings Law Journal article (65 Hastings L.J. 1043) accepts part of Kramer and disputes the rest. He argues that the rise of ideologically coherent national parties in the late nineteenth century made state legislative elections turn on national issues, since voters used them to choose senators; supporters of direct election wanted to separate state from national politics and give state elections back to state issues. He also argues that national parties have not in fact acted as political safeguards, and that state elections still turn largely on the president's popularity. His conclusion on repeal is blunt: it would turn state legislatures into "electoral colleges for U.S. Senators" and make state democracy worse, which is the opposite of what the structural argument promises. The remedy he favors is election law reform to make state elections about state government.

Garcia, read from this side

The majority in Garcia (1985) knew about the Seventeenth Amendment. It wrote that the substitution of popular election of senators was among the structural changes since 1789 that "may work to alter the influence of the States in the federal political process," and it still concluded that the states' position in the federal system had served to minimize the burdens they bear under the Commerce Clause. From this lesson's side of the debate, that is the right reading of the evidence: the amendment altered the process without ending the states' influence, and the record of federal legislation shows states winning exemptions, delays, waivers and money in Congress every year. The dissent's reliance on the ACIR's 1984 report describes a Congress "particularly insensitive to state and local values"; the next decade's Congress passed the Unfunded Mandates Reform Act, and the Court's anti-commandeering and coercion rulings were handed down against statutes that had already been softened in the legislative process. Whether that record shows a working political safeguard or a failing one is the question the scorecard tries to answer honestly.

What the two sides agree on

Set the two lessons side by side and the common ground is larger than the argument suggests.

• The framers meant legislative election to tie the Senate to the state governments. Madison's Federalist 62 says so, and no one on either side disputes it.
• The federal government grew enormously after 1913, mostly in bursts tied to the two world wars, the Depression and the 1960s, and grants with conditions became its main tool for dealing with the states. The OMB tables are not in dispute.
• The old system had real defects: deadlocks that left seats empty (Haynes's 1906 table has 46 deadlock entries and 14 failures to elect from 1891 to 1905, the same figures the Senate Historical Office gives), bribery cases, and a voters' revolt that produced the Oregon System and 29 states with popular choice by 1912.
• Direct election changed senators' behavior toward voters. The empirical studies agree on the direction, and the structural side does not deny it.
• The Supreme Court's political-safeguards doctrine of Garcia and its judicial-limits cases of 1992 to 2012 are both responses to the question this course asks, and the Court has never settled on one answer.
• Repeal is very unlikely under Article V (lesson 10), so the live question is what, if anything, should substitute for the lost structure: courts, parties, election law, or nothing.

Scorecard

Each row takes one effect claimed in lesson 8, gives the best evidence for it and against it, and offers a verdict in plain words. The verdicts are this course's judgment, offered so you can disagree with something specific.

Claimed effect	Evidence for	Evidence against	Where the evidence points
The states lost their direct agent in Congress	Article I, Section 3 and Federalist 62 made the legislature the senator's principal; instruction and resignation on instruction occurred (lesson 2); Amar 1996 and Zywicki 1994 trace the incentive change	Schiller and Stewart: 98 percent party-label outcomes, two-thirds of legislators one-term, money and caucus politics ran the process; 29 states had popular choice by 1912; Haynes on the caucus	True in law, weak in practice by the 1890s. The agent existed; the principal had mostly stopped giving orders.
Senators became less protective of state governments' powers	Structural inference; Powell's Garcia dissent and the ACIR 1984 report; UMRA passed only in 1995	No study measures this directly; studies measure responsiveness to voters, which rose. State governments still win exemptions, waivers and UMRA	Unproven either way. The central claim has not been tested with the tools used for the other claims.
Senators became more responsive to voters	Crook and Hibbing 1997; Bernhard and Sala 2006; Meinke 2008; Gailmard and Jenkins 2009, with the House as control	Schiller and Stewart: parties and money dominate both eras; incumbency rose to 91 percent	Supported. This is the best-measured effect, and it is what the amendment's authors promised.
Federal spending grew because the brake was gone	Outlays from 3.4 percent of GDP (1930) to 23.1 percent (2025); the Sixteenth Amendment the same spring	Flat spending 1913 to 1916; jumps in 1917, 1933 and 1941 each with its own cause; postwar ratchet is common to all industrial democracies	Not shown. Timing and mechanism both favor the crises. The structural view survives only as a claim that the ratchet did not reverse because no state-based body pushed back.
Conditional grants became the main lever over the states	Grants from 0.9 percent of GDP (1940) to 4.0 percent (2025); Dole 1987; NFIB 2012	States lobby for the grants and most conditions; the Court drew a coercion line in 2012; grants did not pass 2 percent of GDP until 1968 (Table 12.1)	True as description, unproven as consequence of the amendment. Grants grew with the programs that funded them.
Unfunded mandates and preemption spread	Amar 1996; the UMRA preamble; CRS on field and conflict preemption	A directly elected Congress passed UMRA; preemption counts could not be verified; the Court polices commandeering since 1992	Real phenomena, contested cause. The 1995 statute is evidence both of the problem and of the political process answering it.
The Court had to substitute judicial limits for the lost structure	National League of Cities, New York, Lopez, Printz, NFIB; Powell's dissent	Garcia majority acknowledged the amendment and still relied on the process; Rossum says the substitution is illegitimate; Kramer says parties are the safeguard	The Court has acted as if the structural view were partly right, without ever saying so. Whether it should is a question of constitutional theory, not of evidence.
Repeal would restore the check	Zywicki 2010 and the state resolutions in lesson 10	Schleicher 2014: national parties would turn legislatures into electoral colleges; Schiller and Stewart on money and party in the old system	Doubtful. The system repeal would restore is the 1890s system the data describe, not the 1790s one the argument imagines.

Questions and answers

If the behavioral studies found changes, how is this the "changed little" lesson?

Because the changes they found are in senators' responsiveness to voters, which is what the amendment set out to produce, and not in senators' protection of state governments, which is what the structural argument says was lost. The studies by Crook and Hibbing, Bernhard and Sala, Meinke and Gailmard and Jenkins are evidence that the amendment worked as advertised. They are silent on federalism.

How many deadlocks were there, really?

Haynes's 1906 table lists 46 deadlock entries from 1891 through 1905 (two-seat contests appear once per seat), of which 14 ended with no election; the Senate Historical Office gives the same 46 and 14 for the period. Against 752 elections over 1871 to 1913, deadlocks were a minority, but the ones that happened cost Delaware all representation in the Senate for the 57th Congress and left seats empty in Montana, Washington, Wyoming, Oregon, California, Pennsylvania and Utah for most of a Congress each. Lesson 10 tabulates them by state and year.

Did the Oregon System actually bind legislators?

Statement No. 1 was a pledge, enforced by voters rather than courts, and lesson 4 covers how it worked and how many states copied it. The Senate Historical Office's figure of as many as 29 states with popular primaries or elections by 1912 shows the scale. In those states the legislature's formal vote was a ratification of a result already announced.

Why do Schiller and Stewart say the amendment failed to deliver?

Their Brookings paper compares the two eras on the things reformers promised to fix: money in Senate elections, party domination, incumbency and insider candidates. They find all four persisting or growing, with 91 percent of incumbents reelected in 2012 against 70 percent in the indirect era. Their point is about democratic accountability, not federalism, and it is a judgment that the behavioral studies above partly complicate.

Does Garcia still govern?

It has never been overruled. Garcia still means Congress may apply generally applicable laws such as the Fair Labor Standards Act to state governments. The later cases carve out commandeering (New York, Printz), coercive spending conditions (NFIB) and non-economic activity under the commerce power (Lopez) without touching Garcia's core. Both views of federalism therefore coexist in current law.

Isn't the Sixteenth Amendment the real culprit?

It is a better candidate than the Seventeenth for explaining how growth was financed, since an income tax free of apportionment made war finance and later a large peacetime budget possible. The Constitution Annotated dates its certification to February 25, 1913. But revenue capacity explains the possibility of growth, and the OMB figures show the demand came from wars and the Depression. Money and motive are different questions.

What does Kramer mean by putting the politics back?

Wechsler's 1954 article listed formal structures (Senate, House districts, Electoral College) as the safeguards. Kramer argues the formal structures mattered less than the party system, which made federal officials depend on state and local party organizations for nomination and campaigns. On his account the Seventeenth Amendment, which changed a formal structure, mattered less than the later nationalization of parties and campaign finance, which changed the informal one.

What is Schleicher's alternative to repeal?

Election law reform that helps voters judge state governments on state performance: measures that reduce the degree to which state legislative elections are decided by the president's popularity. His article argues this would do more for federalism than returning Senate selection to legislatures, which would re-nationalize state elections and make legislatures the Senate's electoral colleges.

Hard questions

You admit no study measures protection of state governments. Then the empirical literature proves nothing against the structural argument, and this lesson's title is wrong.

The literature proves two things against it. First, that the legislatures of 1871 to 1913 were not behaving as the guardians the argument needs them to have been: 98 percent party-label outcomes, one-term legislators, caucus control and money, in Schiller and Stewart's data. Second, that the amendment's measurable effects run toward voter responsiveness, which is what it was designed to do. What the literature does not do is directly refute the inference that incentives toward federalism weakened. The scorecard says so. The title claims the amendment changed little about the Senate's relationship to state governments as institutions, and the evidence that there was little to change is strong.

The 1890s legislatures were partisan and bought because the system had decayed. The structural argument is about the design, and the design worked for a century before that.

That is a fair reframing, and it moves the argument's target from 1913 back to the rise of mass parties in the 1830s and 1840s and the money politics of the Gilded Age. If so, the thing that broke the states' check was the party system, which is Kramer's point, and the Seventeenth Amendment was a late formality. It also raises the question why anyone expects repeal to restore the 1790s rather than the 1890s. Lesson 10 takes that up.

Wars and depressions happen to every country. The United States was the one with a Senate designed to resist centralization, and it stopped resisting. Isn't that the point?

The comparison is the right one to make, and it cuts both ways. Federal systems with strong state-based upper houses, and unitary states with none, both built large central governments in the twentieth century; the ratchet is general. The claim that the pre-1913 Senate would have held the line has to explain why it did not hold it in 1917, when a third of the chamber still consisted of members chosen by legislatures and most of the rest were former legislative choices who had won popular reelection, or in the late nineteenth century, when it passed the first federal regulatory and antitrust statutes. The honest answer is that nothing in the record shows the old Senate resisting centralization when the country wanted it.

Doesn't the Court's own behavior since 1992 prove the political safeguards failed? Judges do not strike down laws when the process is working.

It proves the Court concluded the process was insufficient in a handful of cases: a take-title provision, a background check mandate, a school-zone gun law, a Medicaid condition. Those are narrow holdings against a background in which thousands of federal statutes affecting the states were negotiated, amended and passed with the states' participation. One can read the cases as the safeguards failing, or as a backstop for the rare failure. Rossum reads them as the Court overstepping. The evidence does not pick among those readings; constitutional theory does.

If the amendment changed little, why did the states fight about it for twenty years, and why did Utah reject it?

Because it changed something for the people in the fight: senators who would lose their legislative patrons, legislators who would lose a bargaining chip, machines that would lose a lever, and reformers who wanted the symbol. Schleicher's account adds that legislators wanted their own elections back. The Senate's refusal from 1893 to 1911 (lesson 4) shows the incumbents understood the stakes for themselves. None of that requires the amendment to have changed the balance between Washington and the states, which is the only claim this lesson disputes.