Correcting Douthat on the filibuster

[Updated with a response from Keith Krehbiel]

New York Times columnist Ross Douthat has mounted a defense of the filibuster on his blog as producing centrist outcomes:

If you were so inclined, then, you could cite the final vote on both Bush-era bills [the 2001 tax cut and Medicare coverage of prescription drugs] as exhibits in a bipartisan case against the filibuster, using them to argue that supermajority requirements tend to block legislation no matter who’s in power. But then you’d have to claim that there would have been no Bush tax cut in 2001 without Congressional Republicans’ willingness to use reconciliation, and no Medicare Part D in 2003 if the Democrats had been willing to filibuster the final bill.

That’s possible, obviously. But it seems much more likely that a supermajority requirement would have produced a slightly smaller tax cut — closer to the one that earned those 65 Senate votes, probably — and a slightly different pattern of House-Senate negotiation on prescription drugs. In both counterfactuals, I suspect the bills would have still passed — and what’s more, that their final shape would have vindicated Jay Cost’s suggestion that the filibuster tends to pull legislative efforts toward the center, rather than killing them off entirely.

While Douthat may be right that the 2001 tax cuts and Medicare Part D would have passed under a filibuster, his broader claim that the filibuster "tends to pull legislative efforts toward the center, rather than killing them off entirely" is at least partly wrong.

Let's bring some political science to bear on this question using the gridlock zone model popularized by Keith Krehbiel, David Brady, and Craig Volden. In this model, we assume a one-dimensional political space along which all legislators have ideal points ranging from liberal to conservative. We also assume the status quo and proposed legislation can be located as points on this dimension and that the president is near the liberal or conservative extreme. In this world, the relevant players are the veto pivot (the decisive voter on whether a presidential veto is overriden) and the filibuster pivot (the decisive voter on invoking cloture to end a filibuster), as in this figure from Krehbiel's Pivotal Politics:

Omitting technical details, the key point is that much depends on the location of the status quo. If it is located between the veto and filibuster pivots, then gridlock will prevail -- the president will veto any movement away from his ideal point (and his veto will be sustained), while the filibuster pivot will block any movement toward the president's ideal point. If the status quo is located outside the so-called gridlock zone, then it is possible to pass legislation. However, the content of the bill (i.e. its location on the ideological dimension) will be determined by the relevant pivotal voter, who must be at least indifferent between the bill and the status quo. As such, the more extreme the status quo is relative to the relevant pivot, the more centrist the resulting legislative outcome will be.

Without the filibuster, the relevant pivots constraining legislative movement toward the president becomes the median voters in the House and Senate. As such, bills on which the status quo falls between the median voters and the filibuster pivot can now be acted upon and will be moved close to the medians' ideal points.

Using this framework, we can show that the second part of Douthat's statement is false. The filibuster kills all bills in which the status quo falls between the median voter and the filibuster pivot. I posted a list of failed cloture votes that attracted majority support from recent sessions of Congress back in December, and there are presumably many more issues on which such votes aren't held given the likely outcome.

As for the first part of Douthat's statement, it depends on how you define "the center." Getting rid of the filibuster would obviously move outcomes toward the ideal points of the median voters in the House and Senate, who are by definition the central voters in their respective chambers, so Douthat is presumably appealing to a broader notion of "the center." Looking at the list I put together of the pivotal voters in the Senate with and without the filibuster, it's not always obvious which is closer to the center of the national ideological distribution.

Perhaps the best approach to this question is the one taken by my future Dartmouth colleagues Joseph Bafumi and Michael Herron, who asked voters about key Congressional votes in order to place them along the same dimension as legislators. They found that the House and Senate medians in the 109th Congress of 2005-2006 were more conservative than the national median voter, but the results of the 2006 election moved them much closer to the national median (PDF):

These results suggest that the filibuster pivot might have been more representative of the national median than the Congressional median in the 109th Congress (i.e. Louisiana Democrat Mary Landrieu instead of Ohio Republican Mike DeWine; see my December post), but in the 110th Congress the Congressional median was almost surely closer to the national median than the filibuster pivot. (Caveat: The extent to which actual legislative outcomes will correspond with the filibuster pivot's preferences will vary by issue depending on the location of the status quo -- see above.)

In short, the extent to which the national median voter's preferences correspond with the relevant pivots in Congress will vary over time. There's no guarantee that the filibuster pivot will be closer to the national median than the Congressional median.

Update 1/7 8:17 AM -- Krehbiel sends an additional response that I've posted with his permission. In particular, see points (a) and (b), which make a distinction that I failed to explain sufficiently:

Interesting post, Brendan. I think, however, you are too easy on Douthat (which I confess I have not read) if he said, as you quote, that the filibuster "tends to pull legislative efforts toward the center, rather than killing them off entirely."

In the pivotal politics model, the effect, if any, of the filibuster (as opposed to no filibuster) depends on one and only one thing: the extremity of the status quo point relative to the median and filibuster pivot. Specifically, either:

(a) the status quo is extreme in which case the filibuster pivot does not constrain convergence to the median, i.e., it has no effect, or
(b) the status quo is outside the gridlock interval but not extreme, in which case the filibuster pivot dampens convergence to the median -- policy will converge/moderate, but not all the way, or
(c) the status quo is inside the gridlock interval in which case the filibuster pivot implicitly freezes it, because a 60-vote majority cannot be obtained.

That said, (re)consider Douthat's claim, and then current healthcare politics:

To see that Douthat is wrong, note that the filibuster has no effect in situation (a): the same thing happens whether 60 votes or 51 votes are needed. This leaves only two possibilities. In situation (b), the filibuster dampens convergence rather than "tending to pull legislation toward the center" as Doubthat claims. And, in situation (c) the filibuster precludes convergence, i.e., causes gridlock which is exactly opposite the claim. In short, the argument cannot be sustained (within this framework).

The pivotal-politics interpretation of the ongoing healthcare lawmaking, similarly, depends upon where the status quo lies. If we are in situation (b), existing bills have converged towards but probably not all the way to the median voters' ideal points in their respective houses. If we are in situation (a), however, the bill/bills will ultimately be weakened to the point that leaders come to the realization that the supermajority cannot be constructed, i.e., gridlock happens.

My only disagreement with Krehbiel concerns the interpretation of Douthat's statement (i.e. whether he is "wrong" about the filibuster "pull[ing] legislative efforts toward the center"). As noted above, I suspect that Douthat is referring to the "center" of a broader ideological spectrum rather than the median voters in the House and Senate. As such, he may think that the dampening outcome under scenario (b) above is likely to be closer to the national median than the Congressional median outcome that would otherwise prevail. However, the Bafumi/Herron results suggest that this claim is likely to be false at least some of the time.

Update 1/7 9:37 AM -- Krehbiel clarifies:

You are correct. I interpreted "pulls to the center" as "pulls all the way to the median." My reasoning is that it would be sort of silly for one to credit (presumably) the filibuster for pulling policy [part way] to the center, when abandonment of the filibuster would allow policy to go all the way to the center (a better thing, presumably), wouldn't it?

Comments

[Sarah]

Putting the science back in political science. Love. It.

[ClayDavis]

Pft, but the all-knowing Tom Coburn already convinced me that political science is a waste of time and money.

What's there to study? Much better to let the Beltway gasbags call their status-quo preferences "moderate," "centrist" and "serious" without having to justify this, even if our nation remains unable to deal with serious problems . . .

[David]

Brendan claims to have shown that a particular claim by Douthat is at least partly wrong. The ordinary reader would take this comment to mean that Douthat's claim is wrong in the real world. However, Brendan has has only shown is that Douthat's claim is wrong within the context of a particular model.

I ran into this distinction while creating an actuarial exam based on Samuelson's economics book. Some committee members submitted questions based on statements that were apparently true because they were quotes from the book. However, some of these statements were true only within the context of a particular economic model. As stand-alone statements their validity couldn't be evaluated.

[Jinchi]

However, Brendan has has only shown is that Douthat's claim is wrong within the context of a particular model.

By that standard, shouldn't you believe that Douthat's original post is pointless? He hasn't attempted to prove that his argument is right in any world (he's just stated it as fact) and you claim that it could never be proven at all.

[David]

The current health reform bill is the most significant, ultra-partisan law in the last 60 or 70 years IMHO. It's being passed only because the Dems have the Presidency and a veto-proof majority. That's only one instance, but it's an important one, and it supports Douthat's position.

For me, Brendan's model-based approach has three glaring problems. First, the simplified assumptions are obviously quite far from reality. Second, the technical details have been omitted from Brendan's blog post, presumably because they're too complicated and/or too lengthy. Third, the fact that one model comes up with a different conclusion seems inadequate to prove that Douthat is wrong in the real world. For all I know, other models might support Douthat.

Jinchi raises an interesting point. It suggests the following: If the only real political truths are those that can be scientifically proved, but if Policial Science cannot adequately prove real-world propositions, then there can be no real political truths. I tend to think that's more or less the case. Politics doesn't have laws with the degree of certainty of the law of gravity.

However, academics may have a different definition of "truth" -- namely, publishability. Douthat's claim is wrong in the sense that a Political Science journal will publish an article saying that Douthat is wrong based on a model like this one.

[Rob]

In fairness to the Bafumi-Herron model, their graph does show eight humps, and that much humping seems like a valid depiction of Congress. Marshall McLuhan was right: the medium is the message.