About this Blog
About The Author
Email Me

RealClearPolitics HorseRaceBlog

By Jay Cost

« The Hill Committees at the Five Month Mark | HorseRaceBlog Home Page | An Incipient Realignment? A Response to Judis and Teixeira, Part 2 »

An Incipient Realignment? A Response to Judis and Teixeira, Part 1

This essay is a response to a fascinating article by John B. Judis and Ruy Teixeira entitled "Back to the Future: The Re-emergence of the Emerging Democratic Majority." Judis and Teixeira argue that the 2006 election signals a realignment that favors the Democratic Party. I think their theory is underdetermined, and in this essay I shall offer my justification for that position.

First, let me make clear that what follows is a non-partisan critique. I am not going to try to convince you that the facts point toward the opposite of what Judis and Teixeira argue. I'm not going to try to convince you that the GOP is on the rise, and that 2006 was an aberration. I am, rather, going to argue that the epistemological foundation of their entire project is unstable, and - as a consequence - Judis and Teixeira fail to support their hypothesis. Simply stated, this kind of argument is very tricky, and Judis and Teixeira fail to make it.

I shall spread my response over today and tomorrow.


What is a political realignment? To put it succinctly, a realignment is a broad-based, deeply-felt, and long-lasting change in the partisan orientation of a segment or segments of the electorate. Realignments can be one of two types: secular, or slow-moving evolutions in partisan preference that occur over a period of time; critical, or sudden shifts to one party or another that do not have a historical antecedent.

Lots of people like to predict realignments. Every post-election political science conference you attend, you will find at least one panel with at least one person pushing the idea that the last election was a signal that a new realignment might be a-comin'. I avoid those panels. I think that predicting realignments as they are starting to occur, as opposed to identifying them after the dust has largely settled, is exceedingly difficult. I think that you are far too likely to make a major inferential error when you argue for a prospective realignment. Judis and Teixeira make at least four.

(1) The great Harvard political scientist V.O. Key differentiated four types of elections. On the one hand is a realigning election, along the lines of what we just discussed. On the other hand is a set of elections that revolve around realigning elections: reinforcing elections in which the party favored by the alignment maintains its position, deviating elections in which the party not favored by the alignment makes gains without bringing about a new alignment, reinstating elections in which the party favored in the alignment returns to its position after a deviation.

How can you differentiate a realigning election from the rest? A realigning election is, as I said, broad-based, deeply-felt, and long-lasting. All of these imply a temporal distinction. Many voters must make the same kinds of political choices in multiple contests in cycle after cycle. How do we know if a given event was indeed broad-based, deeply-felt and long-lasting? We have to wait and see if large numbers continue to vote the same way over many offices again and again.

How can you do this before you have observed multiple elections? What would a critical test be, right now in 2007, to corroborate the theory that 2006 was a realigning election? I cannot think of one that does not require the assistance of Doc Brown, Marty McFly, and a flux capaciter. Maybe one is possible, but a prospective argument for a realignment is, I think, a dangerous undertaking. Minimally, you have to offer some fairly compelling evidence in lieu of future returns.

Obviously, since Judis and Teixeira think this Democratic majority is "emerging," they see a realignment that is just starting to happen. Of course, they do not have future election returns to offer, so what is their compelling. alternative piece of data? They argue:

We take a different view [than the view that 2006 was an "an event-driven election"]: that this election signals the end of a fleeting Republican revival, prompted by the Bush administration's response to the September 11 terrorist attacks, and the return to political and demographic trends that were leading to a Democratic and center-left majority in the United States. In 2006 the turn to the Democrats went well beyond those offices directly concerned with the war in Iraq or affected by congressional scandals. While Democrats picked up 30 House seats and six Senate seats, they also won six governorships, netted 321 state legislative seats, and recaptured legislative chambers in eight states. That's the kind of sweep that Republicans enjoyed in 1994, which led to Republican control of Congress and of the nation's statehouses for the remainder of the decade.

This argument is an insufficient way to deal with our lack of knowledge about the future. Just because many seats flipped from one party to another does not mean that they will stay flipped. In 1974, for instance, the GOP took hits nationwide from the top to the bottom of the ballot in every region of the country, but the damage did not last. Extent of damage does not imply permanence.

Even if we assume the effects of 2006 are permanent (which, I expect them to be to an extent), it does not mean that it was induced by long-running "political and demographic trends" that are consistent with realignments. Eventful political years can be significant and long-lasting for a party even if the "political and demographic trends" still favor the other party. There are other features to realignments other than length of effect. They must be broad-based and deeply-felt - which means not only that voters need to vote for the candidates they elected in the first instance of the realignment, they must confirm that decision in other elections as they acquire the opportunity. The fact that officials elected in an electoral "shock" are largely retained in office is not sufficient to claim that a realignment has occurred. For instance, the GOP made slow-but-sure gains in the Mountain West and the South in the postwar period, even though 1958 was a long-lasting setback in the Mountain West and 1974 was a long-lasting setback in the South. Neither of those events were realigning events, even though their effects were long-lasting.

The essential point is that we need to see what happens next. It could largely wash out, as some elections do. It could have a lasting effect that nevertheless does not amount to a realignment. It could be a realignment. We have one data point; to arbitrate between these alternatives, we either need more data points, or a compelling reason why we do not.

Incidentally, Teixeira and Judis are just flat out wrong to argue that we should expect the Bush administration's problems not to trickle down the ballot in a non-realigning year. National politics has that effect all the time. In fact, a party's fortune on one governmental level is highly and positively correlated with its fortunes on other levels. I do not know why they would imply otherwise.

(2) Within the last few cycles, there will be at least one election that does not fit your theory of realignment. That is, the party on the losing end of the supposed realignment will have scored a least one important victory. You must explain this in a way that is not reducible to special pleading. You can do that with all of the critical realignments. You can make a convincing case why 1928 and 1932 were fundamentally different, or why 1852 and 1860 were fundamentally different. You can do it with the secular alignments - noting the slow-but-sure trends in favor of the advantaged party.

Unfortunately, Judis and Teixeira fail to do that. To explain the failure of their hypothesis over the cycles preceding 2006 (remember, they initially offered it in 2002), they invent a psychological concept.

But there was also evidence of another psychological process, which might be called "de-arrangement." The focus on the war on terror not only distracted erstwhile Democrats and independents but appeared to transform, or de-arrange, their political worldview. They temporarily became more sympathetic to a whole range of conservative assumptions and approaches. In the past, voters had trusted Democrats to manage the economy, and in 2002 that preference should have been strongly reinforced by a recession that occurred on Bush's watch.

This is a textbook example of special pleading. To invent a psychological concept to explain data that cannot otherwise be explained is spurious reasoning. I honestly find it incredible that they think this is sufficient to resurrect their theory after the setbacks it suffered in 2002 and 2004. Personally, I would have loved it if I could - in my dissertation work early last year - insert an as-yet-undiscovered psychological concept to explain the vast divergence between my initial theory and what I found in the real world. But, unfortunately for me, those stingy profs wouldn't let me! So, I had to spend six months back at the drawing board. I think that Judis and Teixeira need to try a little harder to reconcile two major falsifying instances with their theory.

Minimally, if they are going to resort to a unique psychological concept, they need to find one that does not undermine the entire premise of their account. If we assume - for the sake of argument - the facts as they present them, and the existence of the psychological concept they create, what is not to say that recent, pro-GOP adjustments in worldviews were not permanent, and the 2006 election was the "de-arrangement?" Also, what is not to say that "de-arrangement" indicates that voters lack core beliefs on many of the basic political divisions of our day, and that recent variations in party support imply this lack of grounding rather than a "coming home?" The concept of "de-arrangement," in other words, is self-immolating. They have proffered an idea that they think serves their argument, but actually burns their idea as much as the idea they are trying to burn.

At a more general level, we should be able to see here the trap that lies in wait for us if we assume that current polling data about core beliefs/values are valid predictors of the partisan preferences that the public exhibits at the ballot box over the next ten years. I think it is highly problematic to assume this. Public opinion is ill-formed in many respects - which in turn means that analysts are highly susceptible to over-reading polling results. Thus, surveys are insufficient substitutes for future election results.

People do not pay much attention to politics, and their political opinions reflect that. Opinions, even on salient issues, are subject to fluctuations and seeming irrationalities that would surprise most political elites. Even if we can identify the public as holding a discrete set of values that seem to us to imply support for the Democratic Party - it does not mean that we can assume this support will be forthcoming. The connection between avowed core values, on the one hand, and issue/candidate preferences, on the other hand, is often quite tenuous.

Multiple problems present themselves when we try to estimate a voter's electoral behavior based upon his espousing a certain "core" value - which is what Judis and Teixeira consistently endeavor to do in this essay. Any decisive values we have identified might not be as central, and therefore not as decisive, to the respondent as we think they are. To assume that they are implicitly applies a kind of psychological value structure to the respondent that might not exist. At the heart of their argument about voter psychology is, I think, an unstated left-right ideological structure that is known not to exist in the mass public. Also, the values we have identified might be in conflict with other values that we have failed to identify. Respondents can hold contradictory values without knowing that they are contradictory (after all, knowledge of the contradiction implies political knowledge, which we know to be generally lacking). Contradictory values can incline a respondent in one direction based upon how an issue is framed - and framing is essentially what the parties try to do. Also, the salience of a respondent's values might wax and wane based upon the political mood - so that they emphasize their Republican set of values when the Republicans are up, and emphasize their Democratic set of values when the Democrats are up. And so on.

Generally, making a connection between a voter psychology that manifests itself only obliquely in survey responses and electoral results over the next two decades is a highly questionable endeavor. Judis and Teixeira seem to be aware of this at least in some regard - hence their sound inclination to be wary of over-reading the pro-Republican survey results of the last few years. But they fail to hold themselves to the same standard when the data favors their side. They thus do not adhere to a key rule that separates sound analysis from polemic: standards must be applied consistently and blindly.

I will have more to say on their argument tomorrow. Stay tuned!

-Jay Cost