Replacing the Seventeenth Amendment

Some Constitutional Amendments #6, Part II: Repeal was the Easy Part

Many writers propose constitutional amendments in order to demonstrate their fantasy vision of the perfect regime. In Some Constitutional Amendments, I propose realistic amendments to the Constitution aimed at improving the structure of the U.S. national government, without addressing substantive issues.

---

Society if we fixed the Senate

Recap!

<Majel Barrett voice> Last time on Some Constitutional Amendments… </voice>

In “A Senate, If You Can Keep It,” we considered the Senate as a “cooling saucer” or “control rod” regulating the untempered power of democracy in the House. We saw how the Founders had built the Senate with a different electorate—the state legislatures rather than the voters—to imbue it with different virtues (and, yes, different vices) from the House. Then we took a whirlwind tour through the collapse of that system. Unlike, say, the electoral college (which never worked), the Founders’ plan for the Senate worked for almost a century. However, America’s burgeoning size and increasingly organized political parties eroded its structural supports, causing electoral deadlocks in state legislatures, an oversized and under-deliberative culture in the Senate, and the diminishment of state legislative elections (which became proxy elections for Senate). When the populist progressivism of the turn-of-the-century arrived, it found a hollowed-out Senate ripe for collapse, and collapse it did. The Seventeenth Amendment took the power to choose senators away from state legislatures and gave it over to the voters.

This had consequences. It (mostly) solved the deadlocks problem¹ and diminished the proxy-elections problem. However, it drastically weakened federalism (by removing states’ primary defense against federal encroachment), it made money and special interests far more powerful in Senate elections than ever before, and senators became less representative, rather than more.

Above all, the character of the Senate broke down, because U.S. Senators elected the same way as Congressmen, by popular vote, had to learn to share all the vices of Congressmen: the demagoguery needed to win on the campaign trail, the lust for fundraising, the ability to oversimplify issues in order to sell them to poorly-informed voters, and more besides. The Senate came to resemble the House, increasingly a vestigial organ rather than an invaluable counterpoint.

The Objective

In this constitutional amendment, our objective is to restore the Senate as a “cooling saucer” to the raw democratic power of the House. To live up to its original purpose, the Senate needs its own, contrasting virtues (and, yes, its own vices).

This is an ambitious objective! I have never had a harder time writing an amendment for Some Constitutional Amendments. Although it is not my longest proposal by word count, it is the most complicated. That is bad. In a constitutional amendment, greater complexity = more ways to go wrong = more places for people to object = less chance of becoming law.

Alas, fixing the Senate is a hard problem! We can’t simply repeal the Seventeenth Amendment, for reasons we discussed last time, but replacing it means venturing out into unexplored territory! We simply have to accept that. My final amendment proposal text will run about the same length and complexity as the 1866 deadlocks bill, which makes sense because it serves a similar purpose, but that makes it longer than the longest existing amendment.² Hard problems often require complex solutions.

A Quick Win

Fortunately, there is an easy way to begin. Call this the low-hanging fruit of senatorial reform.

I have argued that the Senate is too big (nearly as big as the Founding-era House!), that its culture has broken down under the sheer weight of numbers, and that it isn’t able to deliberate properly with so many bodies jostling. So shrink it!

The Constitution guarantees every state equal suffrage in the Senate. This cannot be altered, even by a constitutional amendment.³

However, the “two senators per state” rule was not handed down to Moses at Sinai. Just change the rule to “one senator per state.” Now, instead of a crowded 100-person Senate, you have a 50-person Senate, and you’ve lost… absolutely nothing.⁴

This is low-hanging fruit, and we should do it even if nothing else in this article sticks.⁵

Returning to State Legislatures… Safely

Alas, a smaller Senate of populist demagogues is still a Senate of populist demagogues. To promote the special virtues of the Senate, we really must change who picks them, so that the system stops favoring the House-style demagogues who thrive in mass popular election. Therefore, we will restore the Senate to election by the state legislatures.

However, not the whole state legislature. How about another quick win?

Upper House Only

Some of the problems that arose from the original method for electing senators came from the fact that states with two legislative houses (which is all of them except Nebraska) had to coordinate agreement between both houses. First, the state house would pass a resolution electing such-and-such to U.S. Senate under its rules, then the state senate would have to take up that same resolution and pass it, just like any ordinary bill. You might think passing a resolution through two houses instead of one would be roughly twice as hard. In reality, it is more like ten times as hard.

Because this was so hard, many states blurred the intent of the original Constitution by having both houses of their state legislatures elect senators together, in joint convention, to reduce this friction. The 1866 deadlocks bill then institutionalized this bending of the constitutional rule.⁶ However, joint conventions of state house and senate are still harder to manage than single-house proceedings. There’s a lot more egos to stroke, much more complicated lines of negotiation, and even the floor rules are weirded up. A little chaos is a good thing, insofar as it scrambles party discipline and frees legislators to exercise independent judgment, but not if it also makes it harder for legislators to make deals.

Meanwhile, involving the entire legislature in the U.S. Senate election made the entire legislature subject to possible proxy elections and deadlocks.

I therefore propose that our replacement for the Seventeenth designate one house in each state legislature to elect U.S. senators.

We must then decide whether the state house or state senate is better suited to this task. It seems appropriate for the U.S. senator to be elected by the state senate. A wise state will have already configured their senate to have many of the special virtues sought in the federal senate. It might even have some insulation of its own against direct democracy, reducing the risk of proxy elections and increasing the members’ ability to exercise independent judgment. Such a chamber will therefore be uniquely well-suited to elect senators.

Alas, most states are not so wise. Most state senates were already superfluous even before Reynolds v. Sims came down. State senates today are largely trash. However, even an unwise state will at least have the benefit of fewer people to negotiate between—and fewer ballots to count. Moreover, limiting the U.S. Senate election to the upper house allows the lower house to get on with the state’s most pressing business. Since that is usually the state budget, which, in most state constitutions, must originate in the state house anyway, it is befitting for the state senate to handle this other matter.

The Alaska state senate chamber. There are only 20 senators, split 11-9 Republican vs. Democrat. However, the Democrats have formed a governing coalition with 5 of the Republicans, while 6 “non-caucusing Republicans” form the effective minority party. This is fascinating, but I only included the image to show just how small (and deliberative) an effective senate can be.

Of course, we can’t just say, “the U.S. Senator shall be chosen by the state senate,” because not all states have a senate. Nebraska has only a single legislative house, the so-called “Unicameral.”⁷ Instead, paralleling similar language in Article I, Section 2, Clause 1 and the Seventeenth Amendment, we can say, “the least numerous branch of the state legislature.”

This gives us the first section of our amendment, aka The Easy Bit:

Section 1

The seventeenth article of amendment to the Constitution of the United States is hereby repealed.

The Senate of the United States shall be composed of one Senator from each State, chosen by the least numerous branch of the state legislature thereof.

Conclave Considered Harmful

This isn’t enough. We need to end deadlocks, not contain them to one house.

Nor do we want to (once again) reduce state legislative elections to proxy votes on the U.S. Senate race. State legislators, voting in a federal Senate election, must not become mere agents of their political party, nor mechanical cogs in a machine (like presidential electors who are “bound” to vote for their party’s candidate). We want to see state legislators exercising independent judgment as state legislators. If we repeal the Seventeenth, but they still can’t do that, then we’ve gained very little. Worse, if state legislative elections become proxy elections for U.S. Senate, we will lose quite a bit while gaining nothing at all.⁸

As you have probably guessed, my preference would be to lock state legislators in a conclave and make them stay there until they reach a two-thirds consensus. I think that conclaves are a very good electoral system, which is why I proposed a gubernatorial conclave for presidential elections earlier in this series. Conclaves, with their deliberative structure, secrecy, and supermajority requirements over multiple ballots, are very good at shattering electoral restraints imposed on the participants by partisans and outsiders. Conclaves therefore thwart proxy elections.

For example, if a state party picks a candidate through a primary, the party’s legislators might support him in the first round of voting… but, if no candidate can win without buy-in from the other party, the one person almost guaranteed to be too toxic to elect is the party’s official standard-bearer! This is why, in papal conclaves, the factional leaders of the “liberal” and “conservative” cardinals rarely win, even when one camp holds a clear majority.

Conclaves also tend to punish anyone who shows too great an eagerness to hold office, which is just inherently healthy. (Hence the saying, “He who enters the conclave a pope, leaves a cardinal.”) Conclaves are great.

Unfortunately, we can’t use the conclave method here. Conclaves don’t deadlock often, but it happens, and it happens more often if isolation and secrecy are not properly enforced. I don’t trust individual states to enforce this. With fifty states holding fifty Senate elections, there are a lot of opportunities for something to go wrong, way more than with our gubernatorial conclave, and we know from the Seventeenth Amendment’s history that it doesn’t take many deadlocks to turn the People against the system. We must have a system that is virtually deadlock-proof.

Land of a Thousand Liebermans

Still, just because we can’t run a conclave that compels broad consensus doesn’t mean we can’t incentivize broad consensus.

There are two reasons why we are particularly interested in consensus candidates for Senate. The first is that it makes for a better Senate (even though it often means individually worse senators). The second is that a system that drives toward broad consensus provides a thick layer of republican insulation for individual state senators who are inclined to break with their political parties and exercise the supreme senatorial virtue: independent judgment.

To explain the first: If a senator is to represent an entire state in Washington, it promotes the virtues of the Senate if he is supported by a broad consensus. Today, almost everyone elected to the U.S. Senate is anything but a consensus choice. Most senators are considerably to the left or right of their home state’s median voter. They have to be in order to win their party primary. Then, in the general election, the unhappy moderates who decide the outcome are forced to choose between two extremes. Picking the lesser of two evils, they usually give a small edge to the less extreme of the two unwanted candidates. This is the Median Voter Theorem in a nutshell, and the MVT has proved itself empirically a fair number of times. Scott Alexander explored its practical operation last fall. It is also, of course, a major source of America’s frustration with our two-party system.

In the harshly partisan House, this pattern is not necessarily a terrible thing. It might even be a good thing, in a large house with lots of frequent, competitive elections, for a lot of loudmouthed partisans to win. However, in a self-consciously deliberative Senate designed to cool the partisan passions of the House, you actually want to aim for the middle. A state divided 51%-49% between the two parties should generally elect somebody like Susan Collins or Joe Manchin—not Tina Smith or J.D. Vance. (On the other hand, a state tilted 70-30% in one direction probably should elect a Tina Smith or J.D. Vance, since that is the actual moderate middle of opinion in their state.)

Now, I am, by temperament, a hardcore partisan. I often respect principled adversaries on the other side, but moderates tend to strike me as slippery customers who can’t commit. I instinctively disdain them, especially those on my own “team.”⁹ I spent years daydreaming about Susan Collins being run out of my party, ideally on a rail, no matter how tactically stupid it would have been. However, sentiments like mine are precisely what the Senate is supposed to counter. Ugly partisanship has run amok in both houses, unchecked, so that, instead of being a healthy force for political dynamism against moribund moderation, partisanship has become a national poison. It befits the unique virtues of the U.S. Senate for its members to reflect a broader, more moderate, statewide consensus. It is easy to imagine that, in a Senate packed with senators who align with their respective states’ median voter, it might actually be possible—even straightforward—to reach the supermajority necessary to ratify a treaty, revise the rules, or (the holy grail) convict a President in an impeachment trial after a high crime. It would also be very, very difficult for either political party to use a narrow election win as an excuse for sweeping, unpopular change if it faced a more broadly moderate Senate. The filibuster currently serves the purpose of forcing legislators to win broad support for sweeping changes, but, as we discussed last time, its days are numbered. Wouldn’t it be better if our system selected federal senators who restrained the House willingly?

The second reason a consensus-driven election system is desirable is because it would break down party discipline and enable greater independent judgment. In a first-past-the-post or simple majority election system, it is easy for party bosses to point out defectors from the party’s anointed one and say, “There! There’s the witch! Burn her!” I’ve taken up the hue and cry myself plenty of times. (I was a Tea Partier who cheered against Mike Castle!¹⁰) However, in a system that encourages supermajority winners, it’s much easier for the legislator to say, “Why, yes, of course, I wanted nothing more than to vote for Radrage McBurnItDown, the wonderful candidate endorsed by our party’s rank-and-file and beloved by all good people in our state. Unfortunately, that darned minority party used its leverage and forced us into voting for Wishwash Talkypants III instead.” Without party discipline, the relationship between a state senator and the ultimate outcome of the U.S. Senate race becomes much less obvious. Without a clear relationship between a state senator and the federal Senate seat, it’s harder for the state legislative election to become a proxy election for federal Senate. In an era of strong political parties, this doesn’t come close to solving the problem of republican insulation, but rewarding consensus is a good start.

In fact, having spent a lot of time in the past few weeks reading over the records of nineteenth-century Senate races, I think creating incentives for consensus is an absolutely necessary first step. I have come to believe that the Senate election process the Founders framed was always doomed to be subverted and eventually destroyed by political parties. The Founders did not address this, because they did not believe America would have political parties. As we have often seen in Some Constitutional Amendments, this was the Constitution’s most serious structural defect, leading many parts of the original Constitution to fail, albeit some faster than others. To accomplish the goals the Founders had for the Senate, Senate elections need insulation from political bosses and party rank-and-file, and the Founders provided none. Designing a system that favors consensus gets us partway there.¹¹

Incentivizing Consensusizing

We can incentivize legislators to choose a supermajority consensus candidate by giving the majority a very unpalatable alternative. I’ll sketch a broad outline of my proposal, then dig into the details.

I propose that, if two-thirds of the state legislature can privately agree, before they hold their federal Senate election, on a single candidate, terrific. She wins. This candidate can be anyone eligible for the seat: a fellow legislator, an ex-governor, a war hero, or anyone else they might fancy. In many legislatures, this will require bipartisan consensus. This is good, for the reasons given above.¹²

If the state legislature can not agree to a consensus candidate in time, we switch to the unpleasant contingency: we will limit their options to a list of known moderates and force an immediate decision by a single, secret ballot. Because this “contingent election” is always decided on the first ballot, it is immune to deadlock. Because parties will dislike limited options and will loathe being forced to elect a moderate, they will have adequate incentives to find a better compromise before the contingent election kicks in.

That’s the idea at a high level. Hopefully, many, if not most, U.S. senators will be chosen by a two-thirds majority without much strife.

However, given the history of the United States, we can be sure that won’t always happen, and it might not even happen very often. As usual, the details, especially of the contingent election, will make or break the scheme. As a result, the majority of this article will be about exhaustingly fine details of the contingent election, even though I deem it the less important part of the amendment proposal overall.

Secret Ballot

During debate over the 1866 deadlocks bill, Congress considered making legislatures carry out Senate elections by secret ballot. They already saw how partisan bases and bosses compelled moderate legislators to vote against compromise candidates, even if it meant deadlock. The motion was defeated, in part because roll-call voting was simply customary, and in part because of a powerful belief in democratic accountability, even when it led to concededly worse republican outcomes.

The 1866 bill was a failure in many dimensions, and I think this was one of them. People tend to think of democratic accountability as an unalloyed good, because it can help and how can it hurt? In reality, there is a direct tradeoff between democratic accountability and independent legislative judgment. The former nullifies the latter. Moreover, the prime movers in inflicting democratic accountability are rarely the People, who have lives and therefore don’t spend much time watching the live feed of their state senate on cable access. The prime movers of democratic accountability are nearly always well-funded special interests, including partisan media and the outrage machine, who use their power to gin up fury (and therefore engagement, and therefore monetization) based on soundbites. This is a key idea in Todd Zywicki’s paper on the Seventeenth Amendment Beyond the Shell and Husk of History. It is also (more digestibly) the key argument in the fittingly anonymous ACX book review of Secret Government, and (more topically) in Thirty-Thousand.org’s brief but clear argument against direct election of Senators.¹³

Because we are re-designing the U.S. Senate in order to restore senators’ independent legislative judgment, direct democratic accountability is the very last thing we want, either against the federal chamber itself, or against the state senators who vote for the federal chamber. We must have secret ballots, especially in our era of strong party discipline. This will be a crucial layer of insulation for our senate electors.

Consensus Election

As the federal Senate election opens on the floor of the state senate, open nominations shall be in order. Upon a motion to nominate and a second, the body must immediately vote by secret ballot on whether to elect that nominee to the U.S. Senate, yes or no. If two-thirds present vote yes, the nominee is elected. If not, the nominee fails. The nominee cannot be re-nominated and the senators who made and seconded the motion cannot make another nomination.¹⁴

This last part limits dilatory nominations, which could otherwise be extended infinitely by a powerful party boss. We will also put a clock on this section, again to avoid deliberate stalling: two calendar days.¹⁵ For similar reasons, we’ll force them to start the election process within a month of the vacancy, either before or after.

Other details, like when and how they meet, are prescribed by state law, under Congressional supervision, because that’s what Article I, Section 4 of the Constitution says, and we are not superseding it.¹⁶

Hopefully, the senators came to an understanding backstage and are able to nominate and elect a candidate with two-thirds support on the first ballot. Despite all our talk of deadlocks, the truth is that most Senate elections prior to the Seventeenth Amendment were resolved on the first ballot. However, if no nominee receives consensus support at this stage, the body must proceed to the contingent election.

Or, in constitutional language:

Section 2

Within thirty calendar days of an imminent or actual vacancy in the state’s Senate seat, the electoral house shall convene to elect a replacement. Upon a motion and a second nominating an eligible candidate, the body shall immediately vote by secret ballot whether to elect that candidate or not. If two-thirds of members present concur, the nominee is elected. If not, the nominee may not be re-nominated for this vacancy under this section, and the mover and seconder may not move or second another motion for this vacancy under this section. If there are no further nominations, or if no U.S. Senator is elected in this way within two calendar days, the body shall proceed to election under Section 3.

Contingent Election: Moderate Nominees

We need the contingent election to use a short list of nominees who are reasonably close to the state’s median voter. The better our method does at this, the better it ensures that any contingent election yields a candidate with the necessary senatorial virtues, and the better it incentivizes the chamber to avoid the contingent election in the first place.

We cannot accomplish this with a clever voting method alone, at least not without making the process vulnerable to deadlock. (Trust me, I tried for days.) With strong political parties in a small (<500 voters) electorate willing and eager to collude, it is trivial for a party boss with a bare majority to either elect his favored candidate or derail the election… if he is allowed to nominate his favored candidate.¹⁷ We must therefore make it prohibitively difficult for him to nominate his favored candidate.

Here, more than anywhere else, the tectonic pressures of American politics will be brought to bear in order to break the system. I don’t just mean in the specific context of Senate elections. I think this provision will be brought under more political pressure than any other part of the Constitution, period. Parties are very powerful in the United States. They desperately want to nominate and elect extremists, who reflect the party’s consensus but not the state’s, to the U.S. Senate. They will go to any legal lengths to do so, no matter what constitutional or legislative norms they have to violate. They will not play fair.

Here, then, more than anywhere else, we are justified in being blunt, prescriptive, and inflexible, even if it means sacrificing the elegance and economy of phrasing used by most of the rest of the Constitution.

Take that as an apology in advance.

The Many Uses of DW-NOMINATE

DW-NOMINATE (Dynamic Weighted Nominal Three-step Estimation) is an extremely clever algorithm devised by Keith Poole and Howard Rosenthal in 1997. They designed the algorithm to help identify how legislators vote in relation to other legislators. Specifically: which legislators tend to vote with each other, and which legislators tend to vote against each other? As input, it takes in roll call votes. As output, it produces a number for each legislator between -1 and +1.¹⁸ People you tend to vote with will have a score close to yours. People you tend to vote against will have a score far away from yours.

When Poole and Rosenthal looked at their outputs, they realized that they had perfectly replicated the left-right spectrum. Senators with high grades from the American Conservatives Union had scores well over zero. Congressmen beloved by the AFL-CIO had scores well under zero. Politicians universally scorned as “moderates” had scores around zero. This happened even though Poole and Rosenthal never told the computer anything about the issues. The computer had no concept of “conservative” or “progressive,” but it could see how people voted, and derived the relationship from that.

If you’re curious about the details, a lovely 2006/2007 paper by Phil Everson, Rick Valelly, and Jim Wiseman (NOMINATE and American Political History: A Primer) explains the basics for laymen at some length. For the purposes of this article, you may consider DW-NOMINATE a mathematical miracle. You can try it yourself: think of the most progressive U.S. Senator you can, then the most moderate, then the most conservative. Once you’ve done that, look up their current NOMINATE scores on VoteView.

DW-NOMINATE is extremely powerful, so much so that I have made it the key to identifying moderate nominees in contingent Senate elections.

I propose that, if the state senate is unable to arrive at a consensus candidate, then they must score everyone in the chamber by DW-NOMINATE. The median legislator, plus the legislator just to his right and just to his left, are nominated. In a large state senate, nominate another legislator just to their right and just to their left. Continue until 10% of the state senate has been nominated.

The median legislator in the state senate is likely to be close to the state’s median voter. Those closest to him on either side present a narrow but meaningful range of choices for the rest of the body to vote on.

Notably, the incumbent U.S. Senator is not included in this list. An incumbent who wants to be re-elected must win at the consensus stage, so stay broadly appealing, federal senators!

That’s all very abstract. I’ll try illustrating what these nominations might look like in practice by turning to a body of legislators with which we are all most familiar: the United States Senate itself.

For Example

This isn’t an ideal example, because the U.S. Senate is so big, and it is not apportioned the way state senates are. Moreover, it is weird to explain elections to the U.S. Senate by using the U.S. Senate itself as the example electors, as if the Senate were electing itself.

However, the Senate is also the legislative body with the most recognizable names. I could use the Minnesota Senate, and it would be a lot more “accurate,” but most of you have never heard of any of them, so it would not be very illustrative. I must therefore trust in the considerable intelligence of De Civ readers to follow along, despite these flaws.

Enough throat-clearing. Who actually gets nominated in the contingent election under this fancy DW-NOMINATE system I’m proposing?

In the last U.S. Senate,¹⁹ the Democrats held a closely divided 51-49 majority.²⁰ Sure enough, the “median legislator”, according to DW-NOMINATE, was the midpoint between Sen. Joe Manchin and Sen. Angus King, the two most moderate Democrats. If the U.S. Senate had to carry out a contingent election for U.S. Senate (which, again, is silly) under the rules I am proposing, they would nominate Manchin, King, and their nearest neighbors on either side until 10% of the chamber had been nominated. Namely:

• Sen. Angus King (I-ME) (median legislator)

• Sen. Joe Manchin (I-WV) (median legislator)

• Sen. Tom Carper (D-DE) (first neighbor to the left)

• Sen. Susan Collins (R-ME) (first neighbor to the right)

• Sen. Krysten Sinema (I-AZ) (second neighbor to the left)

• Sen. Lisa Murkowski (R-AK) (second neighbor to the right, etc.)

• Sen. John Fetterman (D-PA)

• Sen. Shelley Moore Capito (R-WV)

• Sen. Mark Warner (D-VA)

• Sen. Mitt Romney (R-UT)

It is worth recalling that, since state senates are much smaller than this, they will have fewer nominees. Most will end up with three to five nominees, not ten. However, the U.S. Senate is absurdly large, so they get ten nominees.

Now, if you’re remotely as partisan as I am, whether for the Right or for the Left, you probably threw up in your mouth a little at that list. Ugh, stick me in a sack and ship me to Abu Dhabi before you make me pick between Mitt Romney and Lisa Murkowski! However, the Senate isn’t for us partisans. The House is for us. We can run foaming-at-the-mouth ideologues to our heart’s content in the House. The fact that we partisans all find this list slightly queasy is precisely what (we hope) will motivate party bosses to strike a deal and elect somebody else with broad consensus support. If they don’t, well… I guess I’d lie back and think of England, for Mitt. Again. I don’t have to like Mitt Romney to recognize that, if both sides are forced toward more moderate candidates than they’d otherwise get away with, the Senate as a whole will be better off.

Of course, if the state senate is tilted in one direction, that will affect the nominees. In an evenly-divided chamber, like the most recent U.S. Senate, you will get a truly moderate candidate list. In a chamber tilted red or blue, you’ll get a more red or blue candidate list—because the median voter will have been more red or blue.

For example, in the current Congress, the Senate tilts Republican, 53-47. The median vote falls between Sen. Shelley Moore Capito and Sen. Mike Rounds. Consulting their DW-NOMINATE scores, if this Senate had to carry out a contingent election under the rules I’ve proposed, the nominees would be (from Left to Right):

• Sen. John Fetterman (D-PA)

• Sen. Angus King (I-ME)

• Sen. Susan Collins (R-ME)

• Sen. Lisa Murkowski (R-AK)

• Sen. Shelley Moore Capito (R-WV)

• Sen. Mike Rounds (R-SD)

• Sen. Chuck Grassley (R-IA)

• Sen. Lindsey Graham (R-SC)

• Sen. Roger Wicker (R-MS)

• Sen. Thom Tillis (R-NC)

Democrats have little to plump for here. Their one official candidate on the list is widely considered a party traitor. On the other hand, hardcore Republicans have little to be excited about here, either. The median Republican today is James Risch of Idaho. If left to their own devices, the Republican Conference would want someone ideologically similar to Risch to represent them. We know this because the heads of the Senate Republican Conference are Tom Cotton (R-AR) and James Lankford (R-OK), both of whom are a little to Risch’s right—and very far to Lisa Murkowski’s right. Even with comfortable control of the chamber, Republicans under this system are still forced to either make a bipartisan deal in the consensus election, or settle for somebody more hostile to the base, like Roger Wicker,²¹ in the contingent election.

The same effect works in reverse when Democrats hold the majority: Republicans end up left choosing between Democrats, but Democrats are stuck choosing between moderate Democrats.

I guess I’d better print that list, just to reassure Democrats reading this, so here’s what the nominations list would have looked like at the end of the 114th Congress (2009-2010), when Democrats had just lost their filibuster-proof majority but still held 59 seats:²²

• Sen. Max Baucus (D-MT)

• Sen. Mark Warner (D-VA)

• Sen. Joe Lieberman (ID-CT)

• Sen. Kay Hagan (D-NC)

• Sen. Mark Pryor (D-AR)

• Sen. Clare Nelson (D-MO)

• Sen. Tom Carper (D-DE)

• Sen. James Webb (D-VA)

• Sen. Blanche Lincoln (D-NE)

Not a single Republican on this list. You can see how, the less balanced the chamber, the more ideological the nominees become. Still, we’re mostly looking at red-state Democrats, a few purple-state Democrats, and Joe Lieberman, who so annoyed the Democrats he left the party and identified as an “Independent Democrat.”

That’s enough for now. Later, we’ll see how the voting system will put further pressure on the majority party to take the minority party’s preferences into account.

Gaming the DW-NOMINATE System

It is possible to imagine someone attempting to game this system. Indeed, Campbell’s Law teaches us to expect it: as soon as an objective indicator like DW-NOMINATE gets used to affect real-world power, power will attempt to subvert, suborn, and corrupt it.

Suppose a party with a narrow state senate majority wants to elect an extremist to the U.S. Senate, perhaps the aforementioned RadRage McBurnItDown, or Candace Owens. They could run McBurnItDown for state senate. As a new senator, he would have little or no roll-call record and would therefore score as a moderate. The party bosses could then ensure that nobody cleared the two-thirds threshold for election-by-consensus, force the chamber to the contingent election, and instruct their members to vote for McBurnItDown. No matter how you structure your voting method, this stands a chance at working—perhaps a good chance. That’s exactly why we went to such lengths to prevent extremists from getting nominated in the first place.

To protect the system from manipulation, we will further restrict candidates to those who have served for at least two years and who have cast at least a hundred roll call votes in their legislative career. For long-serving legislators, we will take the average of their DW-NOMINATE score across each legislature in which they’ve ever served.²³

Under these rules, the only way to have an extremist sneak onto the nomination list would be to designate him the party’s desired senator several years in advance of the election and have him maintain a relatively moderate voting record that closely track’s the chamber’s median member over those years. Indeed, because the system only nominates a set percentage of members, our stalking-horse would have to out-moderate many of the real moderates! He must do this without losing the base’s support due to said moderate voting record or the mere changing of political fashion, and must not find himself suddenly left far from the median by a wave election. He must outcompete members of other factions within his party, and potential stalking-horses from other parties, who might be attempting the same thing. This strikes me as largely impractical.

What this method might do is persuade some state senators to vote more like their state’s median voter, in hopes of eventually ending up on the senate shortlist, becoming actually more moderate in the process. However, that would be great! Their selfish tactical maneuvering ends up benefitting the system as a whole. That’s already the Constitution’s animating principle!

Here, then, is our constitutional text:

Section 3 (Part 1)

Each member qualified for the U.S. Senate who has served in the body for at least two full years (even if non-contiguous) and who has cast at least one hundred roll-call votes shall be eligible for nomination.

Each eligible member shall be scored by linear, eight-iteration Dynamic Weighted Nominal Three-Step Estimation (DW-NOMINATE) in the first dimension, taking the member’s average score across every legislative term in which the member has served.

The median eligible legislator or legislators shall be nominated. If they do not amount to ten percent of the eligible members of the body, then the eligible legislator or legislators who have the closest score above the median, plus the legislator or legislators who have the closest score below the median, shall be nominated. If all these combined do not amount to ten percent of the eligible members, this procedure shall be repeated until ten percent of eligible members have been nominated.

Aside: The Ugliness of a Beautiful Algorithm

I must confess: I don’t like the idea of putting a specific, named algorithm in the Constitution.

The Constitution has a couple of algorithms in it, like the apportionment process and the presidential election process, but these algorithms are simple enough to be defined in the Constitution’s text itself. DW-NOMINATE is complicated and is not suitable for full description in an amendment. Just dropping “score the blokes by DW-NOMINATE” smack into the Constitution, without a definition, feels as foolish as putting “protect free speech and the right to bear arms” into the Constitution without any definition.

Oh, I guess we did do that.

Nevertheless, it was inelegant. The vagueness of “the freedom of speech” and similar expansive phrases led to centuries of court cases, and ongoing disputes about legal interpretation.

I believe that DW-NOMINATE has a sufficiently clear and distinct definition today to avoid these vagueness problems. Nevertheless, it still feels inelegant. I don’t like it.

Alas, hard problems require hard solutions, and using DW-NOMINATE is the least unlikable idea I had. It’s kind of ugly, but it seems to work. It keeps the nominations out of the hands of party bosses and still gives us at least a reasonable set of nominees. Most other approaches do not.

With our nominees in hand, we can finally proceed to the voting.

Contingent Election: The Actual Voting

As we have seen, we absolutely must not allow deadlocks. Therefore, the contingent election must be guaranteed to finish on the first ballot.

We could do this by simply declaring that the candidate with the most votes wins. However, with multiple candidates running, it’s easy to imagine the most popular candidates splitting the vote, thereby allowing the least popular candidate to win. (This happens frequently.) That would be bad for state representation, and it would probably be bad for the composition of the U.S. Senate.

Fortunately, psephologists²⁴ have developed a wide variety of well-vetted, mathematically sound ranked-choice voting methods that can guarantee a result, supported by a majority of the body, in a single round of voting. We’ve got Borda counting, Instant Runoff Voting, Bucklin voting, Approval ballots,²⁵ and more besides. Instant Runoff has become so (undeservedly) popular in the United States that it is now almost synonymous with “ranked-choice voting,” but it’s only one ranked-choice method among many—and, frankly, kind of a bad one.

None of these methods are as good at building consensus as a simple supermajority requirement across multiple rounds of balloting (aka a conclave). However, most of these methods are better at consensus-finding than typical plurality voting. Crucially, they get the job done fast, in one ballot with no chance of deadlock.

In normal, mass, popular elections, ranked voting methods are disfavored, in part because they are complicated. Their inner logic is sometimes opaque, which reduces voter trust in the outcome. Try explaining Schulze Pairs to your grandma! (Or even to your college-educated spouse who majored in CompSci and actually took Discrete Math!) Schulze Pairs are weird. So are several other ranked voting methods, especially Instant Runoff, whose implementation is simple but whose outcomes can be very counterintuitive.

Moreover, ranked-choice ballots are often lengthy and confusing. (This is partly because they need to be machine-countable.) As a result, it is very easy for normal voters to accidentally spoil their ballots. In fact, in general elections, it is often impossible, as a practical matter, to actually fully implement a ranked-choice system! For example, in most ranked-choice methods, every voter should rank every choice to ensure the best outcome. In reality, most jurisdictions only let you rank your top three or top five, because more than that is just too unwieldy to count. Moreover, it takes much longer to count a ranked-choice election, especially if a hand count is needed. An election that might be otherwise tabulated in six or seven hours might stretch to six or seven days if it uses a ranked ballot.

Happily, however, none of this is a concern in a state senate. There are no more than sixty-seven legislators voting. The ballots don’t have to be machine-readable, and there are so few that they can be counted quickly, even by hand. All of the voters here are reasonably smart people who live and breathe politics. They can be trusted to follow ballot instructions, and they can understand how ranked-choice methods work. Indeed, their understanding is sophisticated enough that they will be well-versed in all possible tactical exploits… and eager to use them.

No Bullet Ballots

Many ranked-choice voting systems are defeated by a tactic called the “bullet ballot,” where the voter only ranks their top choice, leaving the rest of the ballot blank. This encourages all sorts of tactical mischief that prevents the consensus candidate from being identified. Major-party candidates love when their supporters cast bullet ballots, because they are confident that they won’t be eliminated in early rounds, and bullet ballots lower the majority threshold needed to win in later rounds, magnifying the power of an early lead.

Fortunately, the remedy is simple: make bullet ballots invalid.

You cannot, practically speaking, do this in large general elections, because voters simply don’t follow directions and you’d end up with tons of invalid ballots. However, it’s easy to implement in this election, with fewer than a hundred voters, all well-informed. In fact, many state and local political parties already have rules against bullet ballots in internal elections, so legislators know what a bullet ballot is and are used to not casting them.

Condorcet Winner

Suppose there are three candidates in an election: Robbie Right-Wing, Millie Moderate, and Larry Lefty. The voters are partisan and polarized, so Robbie and Larry each get 40% of the first-choice votes, while Millie gets only 20% of first-choice votes. However, all of Larry’s supporters would rather see Millie win than Robbie, and Robbie’s supporters feel the same way. If forced to choose between Millie and Robbie, then, the voters go 60%-40% for Millie. If forced to choose between Millie and Larry, the voters again go 60%-40% in favor of Millie. Millie beats all other candidates head-to-head. Millie is the clear consensus candidate of the voters. Millie should win this election, right?

Surprisingly, in most election systems (including the plurality voting system Americans typically use), there is no guarantee that the most broadly acceptable compromise will win. In fact, many of our voting systems militate against this happening, including both our two-party plurality voting system (which favors more extreme candidates backed by a party base) and the Instant Runoff ranked-choice system in use in a few places like Maine and Alaska (which always eliminates the candidate with the fewest first-choice votes—in this case, like many cases, Millie Moderate!).

There are good reasons why you might choose an election system where the consensus candidate doesn’t win. (You can read electowiki on voting system criteria to learn some of the tradeoffs. There are always tradeoffs.) However, in the context of state senators voting for U.S. Senate, after failing to reach a broad consensus, forcing a consensus candidate upon them anyway is kind of the whole point of our proposal.

If, when we count up the ranked ballots, we find one candidate who beats every other candidate head-to-head, that person (known in election science as the “Condorcet winner”) wins the election. Parties will do everything in their power to ensure that the Condorcet winner is their favorite candidate, but they must choose their horse from a limited set of nominees, they can’t use bullet ballots to manipulate the outcome, and the secret ballot creates a lot of ways for legislators to defect, virtually undetectably, especially on second or third choices. It’s not impossible—nothing’s impossible if you believe hard enough (and have enough blackmail)—but it won’t be easy for party bosses to force their choice through without something approaching a sincere majority consensus.

Gaming the Condorcet Winner

This does not mean that the majority party is without power. Even a Condorcet voting system is weak to a well-coordinated majority.

Suppose a fictional, left-leaning state senate with 27 seats has 15 Democrats and 12 Republicans. This chamber holds a U.S. Senate election, fails to reach a consensus, and is forced to the contingent election, where the algorithm nominates the following candidates:

• Roberta Janeti, the pro-life Democrat. She is the median legislator—the 14th most progressive and the 14th most conservative member of the chamber. She is mostly a comfortable Democrat, but her insistence on the rights of the unborn have made her a pariah to many in the party.

• Bob Bigdeals, the centrist. He’s nominated because he falls just on Janeti’s right. Known as a “maverick” who routinely waters down or outright blocks progressive legislation, the Democratic base considers him a “Democrat in name only” who serves corporate interests more than the party’s, and he’s certainly the most conservative candidate on this ballot. Still, he’s from a conservative district where Democrats likely can’t do better.

• Leslie Knope, former small-town city councilor. She falls just to Janeti’s left. Knope, a mainstream Democrat through-and-through, still adores Barack Obama and Joe Biden, but she also has relationships with folks on the other side (esp. Sens. Wyatt and Gergich) which have led her to occasionally vote centrist positions. She is, nevertheless, unambiguously progressive.

Democratic leadership will, of course, want Knope. She’s not their preferred choice by far, but she’s basically a left-wing vote. Republican leadership will likely want Bigdeals, who is even less their cup of tea, but they lost the fall elections and elections have consequences. If Republican and Democratic leadership can get all senators to vote their preferred way, the ballots will come out like this:

• 15 ballots: 1st-choice Knope, 2nd-choice Janeti, 3rd-choice Bigdeals

• 12 ballots: 1st-choice Bigdeals, 2nd-choice Janeti, 3rd-choice Knope

Head-to-head, Knope beats Janeti 15-12 and Knope beats Bigdeals 15-12:

Knope beats all candidates, so Knope wins.

However, it might not be so easy to rig this, especially with a secret ballot. Suppose that Janeti and Bigdeals both want to go to the U.S. Senate, and they both like one another better than either of them likes Knope, so (secretly) they vote for themselves first and one another second. Then the result actually looks like this:

• 13 ballots: 1st-choice Knope, 2nd-choice Janeti, 3rd-choice Bigdeals

• 13 ballots: 1st-choice Bigdeals, 2nd-choice Janeti, 3rd-choice Knope

• 1 ballot: 1st-choice Janeti, 2nd-choice Bigdeals, 3rd-choice Knope

Now Knope loses head-to-head to Janeti (because 14 ballots ranked Janeti ahead of Knope). Janeti also beats Bigdeals 14-13.

Janeti wins!

This is a trivial example of how small defections in the majority, combined with the weight of minority party opinion, can combine to identify and elect consensus candidates, despite attempts by majority-party bosses to control the outcome. (The fact that this is not guaranteed is precisely why it was so important to control the nominees so tightly.)

There are many more complicated scenarios where other members (of either party) secretly defect for various reasons, potentially changing the outcome. Maybe, instead of Janeti voting for herself, one of the far-left reps, who likes Knope personally and ideologically, has recently learned about corruption allegations involving one of Knope’s longtime associates (a certain Mr. Haverford). This worries him, so he quietly votes Janeti > Bigdeals > Knope. This leads to the same result as Janeti defecting in her own self-interest… and nobody except the defector will ever know who the defector is, providing defectors with a measure of protection from vengeful party bosses.

You can play around with Condorcet winners yourself on the VotingMethods.net Condorcet simulator.

Condorcet Tiebreak: Benham’s Method

Unfortunately, elections do not always have Condorcet winners. This is rare, especially in ideological elections, but it is possible for voters to have a rock-paper-scissors cycle between multiple candidates such that no candidate beats all others. Essentially, we’re talking about a tie. All Condorcet methods must account for this possibility, even if it is rare.

For a tiebreak, I prefer Benham’s Method, which was invented by (naturally) Douglas Woodall. In Benham’s Method, you first identify the smallest group of candidates who beat every other candidate head-to-head. This is generally the three rock-paper-scissors candidates who are locked in a cycle with each other. (To voting nerds, this group is called the “Smith set.”)

Then, you do Instant Runoff Voting on everyone outside the Smith set. That is, you eliminate the candidate with the fewest first-choice votes who is not in the Smith set, transfer all his votes to his voters’ second choices, and re-check. Now do you have a Condorcet winner? If so, you’re done. If not, do it again. I’ll put an example in this footnote:²⁶

After all my Instant Runoff Voting bashing earlier, my use of IRV here might come as a surprise, but the biggest problem with IRV is its tendency to eliminate the Condorcet winner! Since we are applying IRV specifically to candidates who could not be the Condorcet winner, this problem is muted.

Of course, if we run out of candidates outside the Smith Set, and still don’t have a winner, we will have to use IRV against candidates within the Smith set. That leads us to the other problem with IRV: it’s just kind of random. No intuition will help you predict what exactly will happen in an IRV elimination round. Its rules lead to all sorts of weird outcomes. However, we are resolving what is, in essence, a complicatedly tied election. All election systems resolve ties with some sort of randomness. IRV’s randomness is, at least, a little more directed than just drawing lots.

Again, cycles should rarely happen, so Benham’s Method should rarely come up at all. I’ve thus just spent five paragraphs on something that might come up in one Senate seat in one state twice in a decade. But, if the proposal doesn’t cover edge cases, the party bosses will exploit them, so here I am, typing it all out anyway.

We are finally ready for the rest of Section 3:

Section 3 (Part 2)

The body shall immediately vote by secret, ranked ballot. Any ballot that does not rank all nominees shall be invalid. If one nominee defeats all others head-to-head, that nominee shall be elected. Otherwise, the smallest set of nominees shall be identified, such that each nominee in the set defeats every nominee outside the set. The nominee with the fewest highest-ranked votes who is not in that set shall be eliminated. If there is now a single candidate who defeats all others head-to-head, that candidate shall be elected. Otherwise, candidates shall be eliminated in this way until a candidate is elected, eliminating nominees in the set only if all other nominees have already been eliminated. Exact ties shall be broken by lot.

E-Z Amendment

As we have seen, Section Three of this proposal is complicated and technical. Worse: as I’ve said, every partisan hack in America will be looking for ways to exploit it. Even without that, it is very possible that, like the Framers themselves overlooking the existence of political parties, we might simply have gotten some of this wrong.

Because the Constitution is exceedingly difficult to amend, any mistakes we make might lock the country into a dysfunctional Senate election system with no plausible hope of escape.

…again.

The entire amendments process is too hard and needs to be made easier. This is one of the rare constitutional questions where Justices Antonin Scalia and Ruth Bader-Ginsburg agreed. We’ll come back to the general process for constitutional amendments in a future installment.

However, even if the amendment process is eventually made easier, I fear it still will not be easy enough to make sure this process, especially the finicky contingent election, keeps working for centuries into the future. On the other hand, I don’t think it would be a good idea for this amendment to be amendable by ordinary legislation. If it were, the first party to win a bare trifecta after this amendment was ratified would go to town rigging it for their own side!

I therefore think we should include a provision that makes it possible for a broad bipartisan coalition in Congress to amend this section of the amendment, without needing to submit the amendment to the states for ratification. Since the Senate itself cannot be impartial in judging its own case, we should leave this duty to the House.

Maybe this sounds weird. However, the Constitution already has entrenched clauses, which have to follow a special, harder process to be amended! This is just an un-entrenched clause, a section that can follow a special, easier amendment process. Thus, as our backstop against major failures in Section 3:

Section 4

Congress may, by the concurrence of three-quarters of the members duly chosen and sworn in the House of Representatives, amend Section 3 of this Article.

No Pledges

Our amendment should have a section expressly forbidding any attempt, formal or informal, to “bind” state senators to vote for a particular candidate. The secret ballot does most of the heavy lifting with that, but we should still make it legally clear that no one can be legitimately bound, not even in honor, to support a specific candidate. This will discourage political parties from holding primaries and “binding” nominating caucuses. It will also give sitting state senators some legal and moral defenses if the parties hold them anyway and try to impose their decisions on the state senates. This helps ensure our state senators will be able to exercise independent legislative judgment. Perhaps this is unnecessary, given the other protections we have already included. However, given the sheer power and determination of the political parties, I’d rather have one more layer of insulation.

Similar concerns motivated a very similar passage in Some Constitutional Amendments’ gubernatorial electoral college, so I’ve simply copy-pasted that passage into this amendment, close to verbatim:

Section 5

Any pledge, vow, oath, or any other commitment by a Senate elector regarding his vote for U.S. Senator shall be null, void, and utterly without force from the moment it is made (excepting her oath to this Constitution and her state’s Constitution). Any instruction, advice, or requirement laid upon an elector, outside the provisions of this Constitution, shall be likewise null and void.

Boilerplate & Housekeeping

Many amendments need a section authorizing Congress to enforce its provisions by appropriate legislation. That is unnecessary here, since the Constitution already supplies the necessary powers under the “Times, Places, and Manner” clause in Article I. We could stick it in anyway, as boilerplate, but I am concerned that it might be construed as giving Congress additional powers, beyond those granted by Article I.

However, we should make the provisions of this amendment explicitly “justiciable.” That means courts can enforce the text of this amendment against states (or even Congresses) that try to subvert it, rather than being forced to dismiss it as a political question or otherwise stick their fingers in their ears. I don’t especially trust the courts, of course, but this isn’t granting them freewheeling power to rewrite Senate elections. It only permits them enforce this pretty strict procedure against other bodies I don’t trust: Congress and the states. I’ll add that as Section 7.

Next, we need to account for the transition between the current Senate, with one hundred senators, and the new Senate, with only fifty. Since all states must have equal suffrage in the Senate at all times, this is a delicate procedure, but not an especially exciting one. (I discussed it in Footnote 5.) That is covered in Section 6.

Finally, we must account for states that have just joined the Union. Although this has not happened in a very long time, it could happen. If consensus election fails, only state senators who have served in the body for at least two years are eligible in the contingent election. If a state has just joined up, though, their senate won’t have existed for two years! We can insert a sentence at the end of Section 3 to make an exception for them, without giving other states a loophole they can exploit if they can somehow eliminate all their eligible candidates.

That’s it! We’ve got an amendment!

---

AMENDMENT XXXIII

Section One

The seventeenth article of amendment to the Constitution of the United States is hereby repealed.

The Senate of the United States shall be composed of one Senator from each State, chosen by the least numerous branch of the state legislature thereof.

Section Two

Within thirty calendar days of an imminent or actual vacancy in the state’s Senate seat, the electoral house shall convene to elect a replacement. Upon a motion and a second nominating an eligible candidate, the body shall immediately vote by secret ballot whether to elect that candidate or not. If two-thirds of members present concur, the nominee is elected. If not, the nominee may not be re-nominated for this vacancy under this section, and the mover and seconder may not move or second another motion for this vacancy under this section. If there are no further nominations, or if no U.S. Senator is elected in this way within two calendar days, the body shall proceed to election under Section Three.

Section Three

Each member qualified for the U.S. Senate who has served in the body for at least two full years (even if non-contiguous) and who has cast at least one hundred roll-call votes shall be eligible for nomination.

Each eligible member shall be scored by linear, eight-iteration Dynamic Weighted Nominal Three-Step Estimation (DW-NOMINATE) in the first dimension, taking the member’s average score across every legislative term in which the member has served.

The median eligible legislator or legislators shall be nominated. If they do not amount to ten percent of the eligible members of the body, then the eligible legislator or legislators who have the closest score above the median, plus the legislator or legislators who have the closest score below the median, shall be nominated. If all these combined do not amount to ten percent of the eligible members, this procedure shall be repeated until ten percent of eligible members have been nominated.

The body shall immediately vote by secret, ranked ballot. Any ballot that does not rank all nominees shall be invalid. If one nominee defeats all others head-to-head, that nominee shall be elected. Otherwise, the smallest set of nominees shall be identified, such that each nominee in the set defeats every nominee outside the set. The nominee with the fewest highest-ranked votes who is not in that set shall be eliminated. If there is now a single candidate who defeats all others head-to-head, that candidate shall be elected. Otherwise, candidates shall be eliminated in this way until a candidate is elected, eliminating nominees in the set only if all other nominees have already been eliminated. Exact ties shall be broken by lot.

In a state admitted to the Union less than two years ago, members who have served in the body continuously since its establishment shall be deemed to meet the tenure and roll call qualifications. Otherwise, if there are no eligible candidates, the election shall be suspended until such time as there are.

Section Four

Congress may, by the concurrence of three-quarters of the members duly chosen and sworn in the House of Representatives, amend Section 3 of this article.

Section Five

Any pledge, vow, oath, or any other commitment by a Senate elector regarding his vote for U.S. Senator shall be null, void, and utterly without force from the moment it is made (excepting her oath to this Constitution and her state’s Constitution). Any instruction, advice, or requirement laid upon an elector, outside the provisions of this Constitution, shall be likewise null and void.

Section Six

In the first twelve months after this article becomes part of this Constitution, each state shall, under such rules as Congress shall prescribe, be simultaneously deprived of one of its senate seats, as randomly as possible, while preserving the equality of the three classes of the Senate. To fill the state’s remaining seat for the remainder of the current term, the state shall hold a senate election as prescribed by this article, except that the only eligible nominees shall be the current U.S. Senators from the state. Until this is completed, the first four sections of this article are suspended.

Section Seven

The provisions of this article are justiciable.

---

What I Expect the Senate to Look Like

If this amendment were adopted, the Senate would change.

The most obvious change would be the increase in moderates. Today, perhaps 3-5% of senators can be considered reasonably likely to buck their party. If this amendment were adopted, I expect that number would jump to 25-40% within six years. Of course, that leaves plenty of room for true-red and true-blue senators from much of the country, and that’s as it should be: the fifty states cover the ideological spectrum, and the Senate should reflect that spectrum. It should not be all-moderate any more than it should be all-partisan!

However, even the reddest (or bluest) state is less red (or blue) than its average partisan primary voter, and I think this system would give those less extreme elements a stronger voice in who gets to be their state’s federal senator. Having a chamber where moderates of various stripes are meaningfully represented would, by itself, transform American politics, although it is not even our primary goal.

You may have noticed another side effect: if this system is adopted, we are likely to see fewer incumbents re-elected. This amendment does not impose term limits, but it requires incumbents to hold the confidence (or at least the acquiescence) of two-thirds of their state senate in order to win a second term, or a third, much less a fourth. This is a high bar.

Voters have, historically, strongly preferred incumbents. Even as the country has polarized, a party that runs an incumbent in a mass election is still much more likely to hold the seat. As a result, parties run the same candidates term after term after term, and trying to unseat an incumbent from the same party is considered a form of internal treason. This has had a dramatic effect on Senate term lengths. For U.S. Senators who served entirely before the Seventeenth Amendment, the average length of service was 6 years, 2 months—just over one term. U.S. Senators whose service began after the Seventeenth Amendment, however, have served an average of 11 years, 7 months, nearly double the pre-Seventeenth figure!²⁷ All of the twenty-five longest-serving senators are children of the Seventeenth Amendment, except Francis Warren (R-WY) whose long service merely overlapped it.

Life expectancy alone doesn’t seem likely to explain this. Instead, state legislatures were more likely to rotate their senators as circumstances changed. Of particular note: the voters’ recent habit of returning obviously senile senators to office was unheard of before the Seventeenth. If this amendment is adopted, moreover, it will make re-election more difficult than it was prior to the Seventeenth. I therefore expect a substantial reduction in average Senate tenure. This provides most of the benefits of term limits without the annoying costs.

However, the most important changes in the Senate would not show up on any blunt poli-sci metrics like moderation and tenure. The primary goal of this amendment was to foster not just a different Senate composition but a renewed Senate culture. By electing an entirely different sort of man to the Senate, with a different constituency, I anticipate a Senate that is more deliberative, less combative. I predict committee hearings that are a genuine search for truth, not a truffle hunt for partisan soundbites. I expect floor debates that see real attempts to persuade other senators, not just posture for the audience back home—and I even, optimistically, believe that those debates might actually succeed at changing minds, and votes, and the future of our country.

I hope, too, that senators elected by their state senates will more fiercely police the boundaries between state and federal power. The current arrangement reduces states to mere provinces of the federal Behemoth, to the cheers of the unthinking mob. States must be empowered to govern their people as their people see fit—not as the people in other states see fit.

The new Senate will have new vices in addition to its virtues, of course. I’m sure they’ll be insular and sometimes insensitive to the needs of the American People… but the House will be there, too, more democratic than ever, to counteract its vices and complement its virtues.

Conclusion

This proposed amendment was a heck of a lot of work, wasn’t it? It took me years to figure out, weeks to write down, and it probably took you days just to read. (This piece is more than half as long as The Old Man and the Sea! I’m taking the next week or two off!)

Still, at its heart, I’ve presented a straightforward solution to a well-known problem: the Senate was designed to protect states from the encroachment of the federal government, and to provide a counterpoint / containment system for the populist demagoguery that was sure to dominate the House. Because we have made the Senate elected by the People, it clearly can no longer serve either purpose effectively. However, even before that change, the Senate was already in serious trouble, because the Founders had failed to anticipate the rise of political parties and their many attendant difficulties. Their error led to deadlocks and proxy elections.

This proposal seeks to return power over the Senate from the People to the state legislatures, so that the Senate can once again fulfill its worthy purposes. (This will also have the happy effect of reducing the importance of money and special interests in Senate elections.) However, in so doing, this proposal does not naively roll back the clock to the already-broken conditions that prevailed in 1912. Instead, it accounts for the problems that had arisen by then, especially the rise of strong political parties, and attempts to account for them. The resulting scheme honors the Founding Fathers’ wise design, but—hopefully—addresses the oversights that have become visible in the two-and-a-half centuries since they did their work.

At an Article V convention, the details of the scheme would be open to negotiation. I’d be happy to quibble over all of them. I’m not married to DW-NOMINATE; I’m very open to a better answer to the same problems, or even a slightly worse answer that’s more palatable to the Article V delegates.

Just remember: any proposal for Senate election that doesn’t fully account for the Founders’ oversights is doomed to fail. If that means a wordy, complicated amendment instead of a crisp, clear one, so be it.

P.S. A belated happy Fourth of July. Two hundred forty-nine years ain’t quite Roman Republic numbers, but it’s nothing to sneeze at. I am very pessimistic about my country’s future, but I love her so much my heart swells up with the ache of it.

Walker Percy’s prophetic novel, Love in the Ruins, opens on the Fourth of July with these striking words:

Now in these dread latter days of the old violent beloved U.S.A. and of the Christ-forgetting Christ-haunted death-dealing Western world I came to myself in a grove of young pines and the question came to me: has it happened at last?…

Is it that God has at last removed his blessing from the U.S.A. and what we feel now is just the clank of old machinery, the sudden jerking ahead of the roller-coaster cars as the chain catches hold and carries us back into history with its ordinary catastrophes, carries us out and up toward the brink from that felicitous and privileged siding where even unbelievers admitted that if it was not God who loved the U.S.A., then at least some great good luck had befallen us, and that now the blessing or the luck is over, the machinery clanks, the chain catches hold, and the cars jerk forward?

I hope not. I must really mean that, too, because a man doesn’t write this much about the Seventeenth Amendment if he’s truly given up.

So, I hope you saw some fireworks, ate some brats, read the Declaration, and talked about loving America where your kids—or someone who has lost faith— could hear you. Enjoy a belated American Tune:

We come on the ship they call the Mayflower
We come on the ship that sailed the moon
We come in the age's most uncertain hours
And sing an American tune

1. “Mostly” because, while we no longer have deadlocks, we have recount lawsuits instead. A goodly number of these last just as long as an average pre-1913 deadlock. Still, the situation seems improved.

2. The Fourteenth, if you’re curious.

3. It requires a super-amendment: unanimous consent of all fifty states. There are various clever plans floated by various clever anti-Senate law professors to work around this restriction. Some of them, even, are plausible readings of the plain text, but all of these schemes violate the clear intention and structure of the clause so obviously and so utterly that I would take a very dim view of anyone who seriously tried one of those tricks outside the just-for-fun atmosphere of a law review article.

4. I must credit Robert George, Michael McConnell, Colleen Sheehan, and Ilan Wurman for this idea. They suggested it in “The Conservative Constitution,” part of a project at the Constitution Center that also yielded “The Progressive Constitution,” “The Libertarian Constitution,” and a set of constitutional amendments that all three teams agreed with. I almost don’t want you to read any of it, because Some Constitutional Amendments runs close enough to it in enough places that I’m afraid it will constitute spoilers, but it really is a neat project.

However, George, McConnell, and Sheehan were working under the assumption of an all-new constitution, so they did not deal with the transition process as I will in the next footnote. I therefore hope to have at least added to their good idea.

5. The transition from 100 senators to 50 can be equally straightforward: each state would immediately lose one senate seat (maintaining equal suffrage throughout). The state’s senatorial electors would vote, using the procedure described later on, on which of their two current senators to retain in that seat until the end of its term. The other senator’s term would expire immediately. The state’s next senate election, several years hence, would be held entirely under the new procedure.

The eliminated seat itself (e.g. whether it’s a Class 2 or Class 3 seat) would be chosen by lot, one after another, in such a way as to keep the size of each senate class as close to equal as possible. This removal process is roughly the reverse of how senators are added when new states join the Union. The result would be two classes with 17 seats and one class with 16 seats, adding up to 50.

Unfortunately, we can’t just hand-wave this and say that currently sitting senators may serve out their terms, because that would compromise equal suffrage in the Senate. No state may have fewer seats than any other, not even for a day, without that state’s consent. We must therefore work around it, with something like what I described above. For example:

In the first twelve months after this article becomes part of this Constitution, each state shall, under such rules as Congress shall prescribe, be simultaneously deprived of one of its senate seats, as randomly as possible, while preserving the equality of the three classes of the Senate. To fill the state’s remaining seat for the remainder of the current term, the state shall then hold a senate election as described elsewhere in this article, except that the only eligible nominees shall be the current U.S. Senators from the state.

6. Specifically, the deadlocks bill required each chamber to take a vote separately. If a single candidate received a majority in both houses, that candidate won. If not, the houses met daily in joint convention and took a vote until somebody won. This was the worst way to marry the two methods and encouraged more deadlocks, by forcing everyone to commit to a public position just when they were least likely to successfully elect (then scrambling the electorate). The 1866 deadlocks bill did that a lot. Remember, they were busy with Reconstruction when they wrote it, and I think it shows.

7. It is also possible for states to have more than two legislative houses, although, as far as I know, this has never been tried.

8. As Christopher M. Russo of Econ Intel Briefs pointed out in a recent comment, it isn’t necessarily a problem (at least not for our Senate reform) if there’s a high correlation between a state’s two-party vote for state legislature and, say, U.S. House:

Political parties are brands, and brands communicate information. A candidate's party affiliation tells me (with moderate confidence) their general views on taxes, healthcare, climate, immigration, gun control, etc. Since everyone has limited time and attention, and it is extraordinarily unlikely that any individual vote sways the election, the strategy of “pull the lever for the candidates of the major party you agree with” makes sense. This strategy will give correlated results across state and federal elections. One's political preferences at the state level generally mirror one's political preferences at the federal level. Hence, we see straight-ticket voting.

Straight-ticket voting is therefore not the outcome we are trying to avoid here. If it turns out that red states elect red state senators who then incidentally end up voting for red federal Senators, and blue states the opposite, that’s just polarization, not a failure of the Senate election system.

What we are trying to avoid is the federal Senate race becoming a central issue in state senate races. State senate races should remain about state issues, at least to the (very questionable) extent that they are about state issues today.

9. Alas, I am a partisan for a dead political party, the Republican Party of 2014, and even that party had big problems. I have all the partisan spirit, but no actual political party to apply it toward.

10. Yes, I knew Christine O’Donnell was a lunatic! I just wanted Castle made an example of, and was willing to lose Delaware for the opportunity.

11. Here we may briefly and brutally dispense with the most common “solution” proposed to the deadlocks problem: if both houses of the state legislature (these proposals always keep both houses involved) cannot agree to a candidate within a certain number of days, the proposal goes, then let the governor appoint the U.S. Senator directly. The concept appears in this proposal by ALEC, and this similar one by Mark Levin, among others.

It’s simple, it’s obvious, it has many precedents… and I am sure you see by now why it would be a disaster. The problems with it seem so evident to me, in light of what I’ve already written, that I almost didn’t include this footnote at all. I trust you readers to recognize the implications. However, the idea has become common enough that I must at least briefly explain why I’ve passed over it.

Governors are not neutral third-party arbiters. They are, themselves, powerful party bosses—usually the most powerful in the entire state. They universally crave more power, and they nearly always have breathtaking leverage over their state legislators. This immense power base, which allows them to resist pressures from their own party, is precisely why I proposed giving them the power to collectively elect the president. (That proposal then used conclave mechanisms, including secrecy, to force the governors to build a consensus.)

However, if you give a single governor the power to appoint the state’s U.S. Senator, even as a contingency, then you effectively make him the dictator of all U.S. Senate appointments in his state so long as his party (or faction of the party) has enough power to engineer an artificial deadlock in at least one legislative house, by hook or by crook. (Imagine the minority party filibustering a Senate election in order to run out the clock and let the governor, a member of their party, fill the seat.) The mere threat, even the mere implication, that the governor’s numerous legislative supporters will cause a deadlock will usually be enough to get the governor’s favored candidate elected. If not, a governor can do a great deal to force an election to fail in order to seize the appointment power himself. Only in rare cases, where the governor belongs to a political party that controls neither legislative house (and there are no procedural tricks the minority can use) will the legislature truly elect their senator—and the threat of an opposite-party governor breaking any deadlock will discourage them from having any real deliberation about it after the initial caucus vote.

Since gubernatorial appointment is inherently public (thus cannot be made a secret ballot), and since the governor is, in our system, always a powerful party boss, the Senate appointment will effectively always be decided by the governor to pay dues to key factions or allies in his political coalition, not to benefit the state at large. We already see this today in temporary gubernatorial appointments to the Senate.

Repealing the Seventeenth Amendment, but giving governors the power to break deadlocks, is (mostly) just a roundabout way of saying, “U.S. Senators are appointed by the governor of the state.” This idea isn’t meritless! However, I don’t think it promotes the virtues of the Senate very well, it works against the whole consensus ideal I’ve been pushing, it runs a high risk of turning the state gubernatorial election into a proxy Senate election (with none of the safeguards from my proposed gubernatorial conclave), and, worst of all, the people behind this proposal don’t even seem to realize that these would be the obvious consequences. They’re sleepwalking into gubernatorial control of the Senate.

I hate to draw a hard line against so many potential allies in our quest to repeal and replace the Seventeenth Amendment, but this is a terrible idea. It would immediately recreate many of the problems repealing the Seventeenth is supposed to solve, then create several more.

12. On the other hand, due to America’s highly polarized politics, quite a few state senates are under super-majority control of a single party. You’d be surprised by how many! Or maybe you wouldn’t be. Since I live in a closely divided state, with a tied house and a one-seat DFL advantage in the senate, I am really taken aback how many others are “one-party states”:

• Alabama: 75%R

• Arkansas: 83%R

• California: 75%D

• Connecticut: 70%D

• Delaware: 71%D

• Florida: 68%R

• Hawaii: 88%D

• Idaho: 83%R

• Illinois: 68%D

• Indiana: 80%R

• Iowa: 68%R

• Kansas: 78%R

• Kentucky: 82%R

• Louisiana: 72%R

• Maryland: 72%D

• Massachusetts: 88%D

• Mississippi: 69%R

• Missouri: 71%R

• Nebraska: 67%R

• New Hampshire: 67%R

• North Dakota: 89%R

• Ohio: 73%R

• Oklahoma: 81%R

• Rhode Island: 87%D

• South Carolina: 74%R

• South Dakota: 89%R

• Tennessee: 82%R

• Utah: 76%R

• West Virginia: 94%R

• Wyoming: 94%R

It seems that, even without the prompting of the Senate election, the People are still voting for state legislators based on the letter next to their name! The Seventeenth Amendment’s supporters had hoped to end that. They thought the Seventeenth Amendment would usher in a new era of voting for the person, not the party, in state legislative races. However (as we saw a few footnotes ago), that was always rather pie-in-the-sky. As long as we live under high polarization, especially with a mostly-national news environment, people are going to vote for the party, not the person, in state senate elections whether U.S. Senate elections are tied to the outcome or not.

Single-party dominance is not ideal for our consensus-driven Senate election model, but it is acceptable. If a party dominates a state this completely, its supporters have democratically earned the right to control the state’s U.S. Senate seat. Forcing a moderate on such a state would deprive the state’s median voter of genuine representation.

Fortunately, in a one-party state, legislative races don’t become partisan proxies for the U.S. Senate race because, hey, there’s only one party anyway. This alleviates the strongest concern about returning Senate elections to the state legislatures. Moreover, even in a one-party state, the need for a two-thirds majority and the secret ballot may force moderate and extreme factions of the party to compromise, which is still a modest improvement on partisan primaries, where extremists tend to dominate.

Unfortunately, partisan primaries might well still happen in these one-party states, and they may even place pressure on legislators to conform to the popular will. Party discipline will be stronger when there’s no powerful minority to scapegoat for defections and compromises. We will attempt to insulate these legislators with a secret ballot and an anti-pledges clause, but that’s probably all we can do. Highly partisan states are going to be more likely to elect highly partisan U.S. senators; news at 11.

For what it is worth, the other states, where no single party can muster two-thirds of the senate by itself, are: Alaska, Arizona, Colorado (barely), Georgia, Maine, Michigan, Minnesota, Montana, Nevada, New Jersey, New Mexico, New York (barely), North Carolina, Oregon, Pennsylvania, Texas (barely), Vermont, Virginia, Washington, and Wisconsin. (In other words, basically every state that ever shows up on an electoral map as remotely competitive, plus Vermont and New York somehow.)

13. Thanks to reader

Epikouros

for pointing the Thirty-Thousand.org article out in a comment as this article went to press.

14. This procedure could be condensed into a single ballot by putting forward all possible nominees at once and then voting by approval ballot. However, my gut tells me that this would unduly discourage nominations and compromises.

At the start of voting, there is often a clear frontrunner. It is difficult for other candidates to be taken seriously until the frontrunner has been tested, because nobody wants to go on record opposing the frontrunner, just in case he wins and decides to punish his opponents. However, once the frontrunner falls, other candidates can step forward without fear of retribution. (Something like this happened to Cardinal Parolin in the most recent papal conclave.) We really do want the consensus election to succeed, so we’ll give them a chance to make it work even after an early setback.

15. We must specify “calendar days,” because legislatures otherwise have been known to play games with time.

16. In fact, we are restoring it: the clause about states having final authority to decide “the Places of chusing Senators” was made obsolete by the Seventeenth Amendment, but its repeal makes it relevant again.

17. Really, I tried for days.

It is trivial to convert approval voting into a party-line vote where each member of the chamber is forced to vote approval for the party’s pre-selected nominee and disapprove all others, even if that isn’t their honest preference. All the Democrats would approve their extremist primary winner and disapprove all others; all the Republicans would support their extremist primary winner and disapprove all others. The winner would simply come down to whether Democrats or Republicans had more seats. The secret ballot may bring about some defections, but, overall, the system is quite likely replicate today’s electoral dynamics, where the parties present two unpalatable candidates to the electorate and force them to choose. This would encourage proxy elections and defeat our whole search for consensus.

Worse, this turns out to be true for all so-called cardinal voting systems. These systems are fantastic if you have sincere voters, but a chamber full of colluding partisans is nearly the opposite of that. Our system must assume rampant tactical voting and still generate good outcomes from that.

Instant Runoff Voting (which, ironically, is the most popular alternative voting method in the United States) has a deeply degenerate elimination mechanism that tends to randomly elect extremist candidates who would lose in head-to-head matchups with centrist candidates. This is called the IRV “center-squeeze” problem, and it makes IRV so problematic it’s barely worth adopting (and rarely lasts long when it is adopted). As we’ll see later on, the main use of IRV is as a tiebreak mechanism in superior voting systems. Note to American electoral reformers: drop IRV like a hot potato. It certainly wouldn’t stop party bosses from nominating and electing extremists; it’s far more likely to help!

Bucklin voting inverts IRV. First, you count all first-ranked choices. If someone has hit the victory threshold, he wins. Otherwise, you add everyone’s second-ranked choices to the tally (unlike IRV, which eliminates votes and transfers them). You keep adding votes from deeper down the ballot until somebody wins. This is very cool in thought experiments, it avoids the center-squeeze problem, and it could almost be adapted to use a pro-consensus supermajority requirement. (The rule would be something like “first candidate to reach two-thirds support wins.”) However, Bucklin exhibits a weird property where your side is more likely to win if you have more candidates on the ballot. That means canny party leaders will simply nominate every member of their party they can to ensure that one of them wins, and then the majority party’s preferred extremist candidate wins anyway. Even if the multiplication of candidates didn’t ruin the system, a ballot with, say, forty candidates to rank is simply not manageable as a practical matter, even in a small legislative body.

And so on down the line.

I spent many, many hours trying to find (or create) a voting system that would elect consensus candidates based on open nominations while resisting the tactical voting behavior of modern, well-disciplined political parties. When there are so few voters, and they are each able to collude with every other voter, this is extremely hard. To be fair, some systems are more resistant than others. However, if there is a perfectly resistant system, it is not known to modern psephology.

To truly protect the independent judgment of legislators from the political parties (and their own primary voters), we cannot depend on our voting method. We must put restrictions on the nominees.

18. As a matter of fact, DW-NOMINATE can analyze legislative voting behavior in multiple dimensions. For example, if there’s two legislators who never ever vote with one another on one set of issues—say, economic issues—but they always vote together on a different set of issues—say, drug issues—DW-NOMINATE can detect that by analyzing votes in two dimensions. Such legislators might be very far apart on the X-axis, but occupy exactly the same point on the Y-axis.

…or vice versa. DW-NOMINATE doesn’t actually know what the issues are. All it can see is the voting behavior. The first dimension DW-NOMINATE calculates, the X-axis, is the dimension that influences voting behavior more strongly. In a Congress focused on economic issues (which is pretty typical), economic issues generally turn out to be the X-axis (or at least a big part of it). Then, if there’s a “drug issues” dimension to voting behavior, it will be on the Y-axis (or the Z-axis, or another even higher-dimensional axis, depending on how important it is). However, if there’s a Congress that’s totally obsessed with drug issues to the exclusion of all else, drugs will be the X-axis in that Congress, and economic issues will probably by the Y-axis. Examining all these axes is mentally and computationally bewildering.

However, good news for political science: the higher dimensions don’t really matter. Across all Congresses ever held, the x-axis (the first dimension) explains 83% of voting behavior. What the x-axis is from Congress to Congress has varied as the issues facing the nation have changed, but the first dimension tells you almost everything you need to know about the relative ideological positioning of Congresscritters. What is that first dimension? It’s not anything as simple as “economic issues,” although some try to simplify it that way. I think the first dimension is best understood as “the set of all issues that currently divide the Left from the Right.”

Adding a second dimension to the math does help: over the history of all Congresses, it adds an additional 2% “accuracy,” making the two-dimensional DW-NOMINATE score 85% predictive. That second dimension is often a secondary fault line that cuts across party lines, like racial politics during the New Deal era, or “morals” issues like abortion and homosexuality in the 1970s before the parties successfully “sorted” on those issues.

According to Poole and Rosenthal’s research, however, adding additional dimensions beyond the second provides no additional useful information or accuracy. You can do it, and you can certainly find groupings of Congresspeople who vote together on weird, not-well-identified issues, but it’s extraneous information. Throughout American history, U.S. politics have been almost entirely one-dimensional. They are occasionally a little bit two-dimensional. The only time they were ever meaningfully three-dimensional or more was in the antebellum political chaos leading up to the Civil War. (In those years, multi-dimensional analysis can be useful.) American political change has largely been about managing that dimensionality: gently transitioning some issues into the Left versus Right spectrum while transitioning other issues out of that spectrum.

This is all extremely interesting, but I mention it here only to state that it is completely irrelevant for this article. We are using DW-NOMINATE to identify where each legislator sits on a straightforward Left-to-Right spectrum, so, for the purposes of this amendment, we are exclusively concerned with its first dimension.

P.S. because I haven’t thought of another place to put this: DW-NOMINATE can run in static, linear, quadratic, or cubic modes. Poole & Rosenthal report that linear mode works best, so that’s what we will use. Official documentation recommends 4 iterations. We will use 8 iterations, because compute is cheap and getting Senate elections right is really important.

19. We must exclude Sens. Feinstein, Menendez, Helmy, Butler, and Manchin’s “initial” run as a Democrat before he turned Independent, since none of these were present at the end of the last Senate.

20. …counting “independents” who caucused with the Democrats, like Bernie Sanders…

21. Thom Tillis is quantitatively a little more right-wing than Wicker, but the GOP wouldn’t support Thom Tillis because this would vacate Tillis’s competitive seat, forcing an election on the GOP that they might well lose. This is another incentive for the GOP to make a deal with the Democrats.

…oh wait! While I’ve been writing this, Thom Tillis announced he is not running for re-election. Why? He voted against Pres. Trump’s “One Big Beautiful Bill Act,” drawing the ire of North Carolina Republican primary voters, making the competitive seat untenable for him. This is a good example of the kind of partisan-base pressure on senators that our amendment should help moderate.

22. Excluded from this list are Sens. Ted Kennedy, Mel Martinez, John Ensign, Ken Salazar, Roland Burris, Robert Byrd, Edward Kaufman, Paul Kirk, Carte Goodwin, and the Republican version of Arlen Specter, since they had all stopped being senators by the end of the session.

23. This will prevent wannabe senate candidates suddenly running to the middle, after long careers on the extremes, in order to get themselves on the contingent election candidates list.

Of course, by “each legislature” we merely mean, for example, to score state senate roll call votes cast in the 94th Minnesota Legislature separately from those cast in the the 93rd Minnesota Legislature.

One interesting, albeit subtle, side effect of this rule is that it gives a little extra weight to whichever party won the election before the most recent election. This is good, because it further corrodes the linkage between the fall state legislative elections and the federal Senate election, reducing the risk of turning them into proxy elections. Actually, my original idea for this amendment was to require legislatures to elect U.S. Senators as their last act before dissolving (after fall elections) rather than (as is typical) their first act after forming. My thinking was that this would put enough distance between an election and the U.S. Senate race to forestall the proxy election effect. Obviously, I abandoned this idea for a number of reasons, but I’m not unhappy to see a faint echo of that idea show up as an unintended side effect of the nomination rules.

24. My spell-checker doesn’t recognize this word, which means it’s obscure enough I ought to provide a definition: psephology is the study of voting and elections. Mostly it’s polling analysis and so forth, your basic Nate Silver stuff, but, obviously, election nerds are hugely interested in actual voting systems as well. However, credit where it’s due: a tremendous amount of the real leg work in this sub-field has been contributed by mathematicians and economists, from the Marquis of Condorcet to Nicolaus Tideman to Douglas Woodall and Chris J Benham, who has made some meaningful contributions to the field despite apparently just being (as far as I can tell) an Australian USENET poaster?

25. I know, nerds: strictly speaking, approval voting is a “graded” method, not a “ranked” method, as are all cardinal ballot methods (as opposed to ordinal ballot methods). Indeed, one can say that approval is simply degenerate score voting, with only two possible scores. However, for lay cromulence, I’m gonna refer to everything that isn’t plurality/first-past-the-post (with or without a primary) as a “ranked voting method” in this section. I apologize for the grating sensation this may create.

26. Take our toy election from the previous section between Bigdeals, Janeti, and Knope.

Now suppose there’s a fracture in the GOP. The pro-life Republican caucus doesn’t care much for Bigdeals, whom they also see as a money-grubbing “fiscally conservative, socially liberal” type. They’d rather have the pro-lifer. Five of the GOP’s eleven make Janeti their first choice, Bigdeals their second, and Knope their third. On the other hand, the pro-choice Republicans really do like Bigdeals, who’s on their wavelength, and dislike both Knope and Janeti. However, because Knope is pro-choice, they dislike her less, so they vote Bigdeals > Knope > Janeti. Our ballots are:

• 13: Knope > Janeti > Bigdeals

• 6: Bigdeals > Janeti > Knope

• 6: Janeti > Bigdeals > Knope

• 2: Bigdeals > Knope > Janeti

Janeti beats Bigdeals head-to-head, 19-8, but loses to Knope 15-12. Knope, however, loses to Bigdeals 14-13. There is no Condorcet winner!

Uh-oh!

Now let’s apply Benham’s method. All three candidates have exactly one victory, so all candidates are in the Smith set.

(I think three-way Condorcet cycles with fully-ranked ballots always result in all three candidates being in the Smith set. To have no Condorcet winner, but still have candidates outside the Smith set, you need to either have bullet ballots or more than three candidates or, possibly, both.)

Because all candidates are in the Smith set, we go immediately to IRV’s semi-random elimination. For all her downballot strength, Janeti was the first choice of only 6 voters, so she is eliminated from all ballots (see that center squeeze problem?) and we rescore them accordingly:

• 13: Knope > Bigdeals

• 6: Bigdeals > Knope

• 6: Bigdeals > Knope

• 2: Bigdeals > Knope

These voters said they’d rather have Bigdeals than Knope, 14-13. Bigdeals wins!

27. Excluding currently-serving senators from this figure reduces the tenure to 11 years, 1 month.

I did original research for this, so, as always, I’ve posted my data online. You can get my Senate Service data file at jamesjheaney.com/jjh_uploads/senateservice.xlsx. Filter by “Service” to get averages for each of the categories (or combinations thereof). The file is based on current and historical data from a consortium of nerds.