A Reponse to KK

One inaccurate prediction and the critics are ready to jump! This from my QJPS co-editor-in-chief Keith Krehbiel (originally posted as a comment):

Nolan’s update makes me wish I had taken the time to write up the various reasons I disagreed with his original, refreshingly out-on-a-limb comment. But now I’m thinking, ‘Prediction is easy, especially about the past, so should I bother?’

Well, even with 100+ days behind us, there is still plenty of future ahead of us, so I’ll venture a somewhat different genuine prediction than Nolan’s (“genuine” as in, about events yet to come). I would be interested in seeing counterpart expectations generated from alternative theories that take a “strong parties” perspective.

Here is a strict version of my expectation. New Specter will be just like Old Specter, voting with the Democrats sometime and Republicans others, and with a much more even split than most/all other Senators. In short, the Specter switch won’t matter at all. This is what the pure version of Pivotal Politics, for example, would say. Not many people would buy this, however, so let’s take it a step farther by trying to incorporate Nolan’s electoral observations into my essentially take-elections-as-given theory.

Specter switched parties for transparent electoral reasons, so to see what difference it makes in governing, one has to size up whether and how his electorally induced preferences will change as a result of the fact that he now has to win his seat from the left side of the Pennsylvania electoral median rather than from the right side. Models of electoral competition don’t speak very directly and generally to this situation, however it is difficult to concoct a plausible scenario in which New Specter ends up right of Old Specter, and it’s easy to do the opposite, so let’s just accept the assertion (nowhere disputed to the best of my knowledge) that New Specter IS now playing to a more leftish audience than Old Specter. (This quasi-theory based assertion is certainly consistent with Nolan’s data, too.) How does this parties-in-electorate induced preferences shift affect what policy comes out of the Senate, the Congress, and ultimately the Government?

Returning now to the theory of government, suppose Old Specter was THE filibuster pivot (probably true on a few issues at least). Now, in light of his left-shifted electorally induced preferences he is no longer the filibuster pivot. Who is? Again, hard to say definitively, but it’s easy to characterize qualitatively: a Senator, such as Snowe or Collins, who has preferences very close to Old Specter — so close, in fact, that it’s hard to imagine there being any measurable impact on policy outcomes. In short, the Specter switch will matter a tiny bit at most. All it does is shift the filibuster pivot “one person” to the left, and even with an ostensible vanishing moderates problem in the Senate, there are still enough moderate Senators that there just isn’t that much policy space between, say, a Old Specter and a Snowe.

My prediction in brief. We’ll see plenty of gridlock — much more than the 100-day-high pundits have prognosticated. Furthermore, if and when major legislation is passed, it will happen via major compromises that make the final product look much more like what Old Specter wanted than what Speaker Pelosi and Majority Leader Reid want.


Footnote: In case anyone wonders what role party discipline plays in this forecast, the answer is none. The idea of Harry Reid disciplining Specter strikes me as comical on its face. I am reminded of an anecdote from a Republican Leaders Office staff meeting I attended in the early 1990s. Minority Leader Bob Dole ran the meeting. Arlen Specter had placed a hold on a bill that Dole was ready to move, given his negotiations with then Majority Leader George Mitchell.
Dole: Still stalled?
Staff: I’m afraid so, Senator. You know Specter: he’s an independent thinker.
Dole: He’s independent.

The problem with prediction is that an ex ante good one can turn out to be wrong and a lousy one can turn out to be just as accurate as a broken clock is twice daily. So let me try to defend my original prediction as the better one ex ante even if may eventually concede that Keith’s was more accurate.

As I originally pointed out, Keith’s prediction that Specter would not move much isn’t really borne out on data from other party switchers. Of course there are many reasons to be skeptical of the data. Perhaps the best is that almost all of the switchers are Southern Democrats who become Republicans. So it might be incorrect to forecast Specter’s positions from the behavior of politicians in a completely different context. But I would maintain that the contexts are quite similar where it is the Northeastern Republicans on the short end of a regional realignment. Keith, on the other hand, finds the forecast objectionable mainly because Specter is one ornery cuss (I’d love to include that as a variable).

But Keith’s real objection is the insinuation that it matters whether Specter becomes Diane Feinstein’s ideological soul mate. He points out that the preferences of the filibuster pivot are likely to shift only a small amount, say to those of Olympia Snowe. If it were just a straight-up Arlen-for-Olympia trade, I agree it would not be a big deal. But let me raise two objections. The first is a quibble: once Al Franken is seated, the NOMINATE data show that the switch is Olympia for Ben Nelson. A bigger jump, but, I concede, not a huge one (but to paraphrase a Mel Brooks movie, it’s good to be Ben Nelson).

The second objection is a more serious methodological one. On any given issue, there is considerable uncertainty about how an individual senator might vote (especially the moderate ones). So the deterministic predictions of the Pivotal Politics model may understate the effects of the shift I had forecast for Specter. Let me clarify with a simple example. Suppose Specter had not shifted. After Franken is seated, support for a key piece of Obama’s agenda is as follows: 57 support it with probability one, 47 oppose it with probability one, and 6 support it with probability .5. This is a fairly plausible scenario (it is close to what the Stimulus bill would have been if Franken had been in the Senate). The probability that the bill achieves cloture and passes is equal to the probability that at least 3 of the undecided support it or .656. Now suppose that Specter had become one of the core supporters so that passage requires only the support of at least 2 of the 5 remaining undecideds. Now the probability of success is .813. While the probability of change we can believe in only goes up about 15%, it is certainly more than “a tiny bit at most.”

There is one thing that Keith and I do agree on. The effects of internal party pressure (i.e. whipping) on Specter’s behavior are likely to be second order or non-existent (that was the main punchline of the paper with the data on party switching).

Soon I hope to share some thoughts about what political scientists might learn about the role of parties from Arlen Specter.

4 thoughts on “A Reponse to KK

  1. carl pinkele

    While not on the point re the discussion of congressional voting patterns and political science and/vs politicos, journalists and consultants I suggest that in the areas of voting behavior, elections, and public opinion political science has much to offer the practical arena players. Political scientists, as a group, might not be skilled tacticians but they are often fully equipped to add very constructive material to strategic observations outside the congressional confines.

  2. Keith Krehbiel

    Nolan makes some good points (as always) and perhaps misses one or two (a much rarer event). Brief replies to his paragraphs follow beginning with number 2.

    2. I don’t predict that “Specter would not move much.” I’m agnostic on this point and simply note that existing theory doesn’t give more than qualitative guidance. Nor do I dispute Nolan’s data, beyond sharing his own reservations about their relevance. Rather, what I’m saying is that even if Specter does move a lot, it won’t matter much in terms of policy. And, as for the “ornery cuss variable,” (a) it’s more of a constant actually, and (b) it plays no role in my prediction. Besides, I like Specter’s orneriness. Here I hypothesize that it’s a Kansas thing. He grew up in Russell—Dole’s home town, too—and not far from Colby, Mark Hansen’s stomping grounds. But I digress (it’s western Kansas after all)…

    3. The “quibble” Nolan raises involves Al Franken and Mel Brooks. I think we’re done with this one.

    4. Finally, Nolan’s “methodological objection” is a good point about the plausibly probabilistic nature of the relative paucity of Senators in today’s pivot ‘hood. I haven’t thought much about this, but my intuition for what it’s worth (about 2 cents) is that Nolan’s argument cuts both ways. Just as probabilistic behavior by near-pivots sometimes makes for a bigger than complete-information-predicted shift in policy toward the Democrats, sometimes it also makes for a bigger than expected shift towards Republicans. The only reason for possibly expecting asymmetry in outcomes distributed about the complete-information baseline expectation is that Republicans (when Democrats are in the majority) will usually be on the status-quo/obstructive side of the pivot(s). What needs to be determined is how this backs into optimal proposal behavior under probabilistically voting proximal pivots. This seems well worth modeling, but, alas, probably not in the near term by two guys who spend an increasing fraction of their days & nights trying to further propel a startup journal into prominence. On the other hand, now that readers of Nolan’s blog know something about the QJPS editors’ interests, maybe someone else will take this bait and illuminate us all not only about the specter of Arlen abut also, more generally, about the policy consequences of uncertain pivots.

    5. Yes, definitely, Nolan and I agree on the inefficacy—OK, “second-order effect,” as he puts it—of internal party pressure in the US Congress. Dissenting opinions welcome.

  3. Joshua Tucker

    Keith: I’m a bit confused about the filibuster pivot point. Let’s assume that Franken gets seated. Then let’s assume we’ve got an issue where old Specter was the filibuster pivot point (e.g., the 60th vote). If new Specter has to move to the left for electoral concerns – as you’ve made an assumption of your set up – doesn’t that mean the Senator immediately to his left becomes the new pivot point (e.g., the new 60th vote), not the Senator immediately to his right (e.g., Snowe or Collins). In that case, it seems that the new pivot point would be someone like Ben Nelson. Now perhaps your point is that Ben Nelson doesn’t look all that different from the old Specter, but it does strike me that having a Democrat be the pivot point – provided it is not Specter himself – could lead to different outcomes than having a trio of moderate Republicans (Snowe/Collins/old Specter) as the pivot point.

    1. Keith Krehbiel

      Yes, Josh, you are basically right, if the parties are indeed perfectly bifurcated in the issue space (and Nolan’s Franken-Brooks analysis wasn’t a complete throw-away). I was a bit sloppy about this because of my belief that, on any given issue, the primary dimension of conflict is characterized by at least some party overlap in the center, even though general ideology measures (DW-Nominates, for instance) indicate otherwise in recent years. I think such bifurcation tends to be overinterpreted, however, and this would seem to be affirmed in roll calls, which usually include defections. Then again, Nolan could just as easily look at the roll calls in which there are a few party defections on either or both sides of the aisle and say this is corroboration of his uncertainty hypothesis. As indicated in my reply to him, I think that’s a valid point, too, and that it serves as more evidence that lots of interesting possibilities remain to be explored from current and forthcoming events. Thanks for flagging this.

Comments are closed.