Jay Cost: A Prediction for the 2006 House Contest

Most political commentators remain ambivalent about the chances of a Democratic capture of the House. Most are willing to say that they cannot capture the Senate – especially now that Lott has decided to return. It would require them to run the tables on every vulnerable and “vulnerable” incumbent Republican out there: Santorum, Dewine, Burns, Chafee, Talent. This would be at a time when incumbents are, historically speaking, as invulnerable as they have ever been, and Republicans are as invulnerable as they have been since Hoover. Most think this is impossible.

But what about the House? To take the House, the Democrats would have to net 15 seats. Can they do this? As readers of my previous work have undoubtedly noted, I believe the answer is no, and the firmness with which I hold this opinion stands as an exception to the current consensus. In the course of the next two columns, I intend to justify this opinion to you.

I think the reason that so many other analysts are ambivalent is that they are making one of two methodological mistakes. Some analyze the House elections top-down. In other words, they investigate the nationwide electoral picture and, from this, attempt to infer how many seats will change hands. I think this is the necessary way to estimate net seat changes, but I also think that too few deploy arguments that are informed by the best, or even a complete, theory of congressional elections. Thus, they look at incumbent retention rates, voter dissatisfaction with Bush, congressional approval ratings, generic congressional ballot, the economy – all of that macro-level data – but cannot really conclude with confidence what it implies. It is all there in a stew that seems to taste more like elephant than donkey, but who can say?

The rest have been analyzing the race from the bottom-up. In other words, they look at which members currently suffer from weak poll positions, which are facing stiffer-than-normal competition, etc. From this they develop a set of possible Democratic captures. All of this is well and good as a way to begin estimating what will occur in the aggregate in November. However, maximally, this mode of analysis will only give you a range of options. You need to look at the bigger picture to begin to determine which scenarios are more likely, which are less likely. For instance, the consensus among “trench” experts is that there are 21 Republican and 11 Democratic House seats that are legitimately contested, with the caveat that the number of contested seats could get bigger or smaller. From what can you conclude based upon this alone? The result can be any of 32 possible outcomes – from a Democratic net gain of 21 to a Republican net gain of 11.

To move from this range to a situation wherein we assign probabilities to certain outcomes requires us to analyze the bigger picture. I think that we can do a better job than most analysts have done thus far. I believe that one can estimate electoral results with a good degree of confidence and precision this far from November. This is where I stand in sharpest disagreement with other commentators. Most others I read tend to speak metaphorically, ambiguously, and always leave themselves an “out” clause. I think this is unnecessary. A higher level of precision in discussing the House elections is possible today. We do not need to wait until November 6 to speak non-metaphorically, to speak without “out” clauses, to speak unambiguously, to speak precisely. I believe that all of this is possible because I do not view House elections as Charlie Cook does: events in which sometimes “the political laws of gravity (are) suspended.” I think that the causal laws of American politics do not become suspended; otherwise, they would not be laws. Congressional elections are like most other political processes, characterized by regular, consistent features and stochastic, surprising features. You cannot predict aggregate electoral results exactly, but the process is regular enough that you can develop a precise prediction that will be very close to correct.

What we need is a good theory: one that makes intuitive sense, one that is historically accurate, one that is precise, one that can help us estimate the size and scope of its possible error. There is such a theory, and it is wickedly simple. It was developed first by Yale’s Edward Tufte in the 1970s and refined later by UC-San Diego’s Gary Jacobson in the 1980s and 1990s. It explains most of the electoral results we have seen between 1946 and 2002. It is the following: the party of the President tends to lose seats in Congress as his party is increasingly exposed in the election, as change in real income per capita stagnates or declines, as his job approval rating stagnates or declines. This theory does not hypothesize that the average voter thinks in this way – that individuals tend to weigh these three factors. As we have discussed previously, the average voter tends to ponder local issues when he casts his vote. This theory hypothesizes that the net effect of every voter is such that the electorate, as a whole, acts in this way. This theory does not contest Tip O’Neill’s wise-but-overused statement that “all politics is local.”

As a nation, therefore, we tend to punish or reward members of the President’s congressional caucus according to three independent variables: exposure, real disposable income per capita (RDI/cap) and presidential approval. Let us take each variable in turn – noting that, if we set all of them to zero (i.e. a party that is neither under- nor over-exposed, a president whom literally nobody likes, and an economy that has neither grown nor shrunk in real terms), we can expect the party of the President to lose a sizeable portion of his House delegation.

Exposure is defined as how many seats a party has above its historical average. For every “extra” House seat that the party of the President enjoys, it tends to lose a portion of its congressional caucus in the next election, all else being equal. The intuition behind this variable is pretty clear. It is like the stock market. When you see a stock performing above its historical average, all else being equal, you expect that stock to go down again. When you see it performing below its historical average, you expect it to go up again. It gets to the idea that there is some kind of “natural” level of value. The Republicans have, today, a “natural” position in the House and, all else being equal, we expect an adjustment toward that position in November.

Next is presidential approval. The intuition behind this is also fairly clear: there is a linkage between the President and the members of his party. This theory does not hypothesize where the linkage exists – does the voter think about the President, do political donors think about the President, etc. It does not hypothesize how the linkage exists – does the voter use his vote as a message, does an unpopular president bring out his opponents, etc. It only hypothesizes that it exists: a popular president somewhere and somehow helps his party; an unpopular president somewhere and somehow hurts his party.

The final important variable is real disposable income per capita (RDI/cap). RDI/cap is the average amount of inflation-adjusted after-tax income each person receives per year – it serves as an estimate for how the economy is performing. Just as with presidential approval, this theory hypothesizes that the economy matters. It does not hypothesize where or how it matters. The argument that voters in congressional elections practice economic rationality, a.k.a. pocketbook voting, has a hard time fitting the facts. So, this theory does not claim that the average voter checks in with the Commerce Department to see how RDI/cap is performing before heading to the polls. It only argues that, as a whole, the electorate tends to vote as RDI/cap tends to change, all else being equal.

And that is all there is to it. Three simple variables will get us extremely close to the correct outcome in 2006. It will not be completely accurate. There will be some error in the theory’s prediction. But, this theory is so good that any prediction about 2006 would be remiss not to make use of it. Thus, what we shall do tomorrow is, to quote my high school geometry teacher, “plug and chug”. We shall see what this theory predicts will happen in November, estimate the extent to which its prediction might be wrong, and see if we cannot get a sense of which party this error might favor.

Jay Cost, creator of the Horse Race Blog, is a doctoral candidate of political science at the University of Chicago. He can be reached at jay_cost@hotmail.com.

Garbage in Garbage out. The current "polls" are anyone over 18 with a pulse that answers the phone. Sorry 50% or so of them will not vote. Next, they skew the sample heavly to Dems and Indies. Wonder what happens to the current polls if you blanced, or favor the Repubs? After all, how do you justify a greater number of Dems and Indies then Reps? Repubs ARE the majority so it would suggest that Dems are at best an equal portion of the sample and should no longer be polled in greater percentages then Repbos

And finally the real kicker. Congressional Districts are not the same as a general pop sample. Does the Dems no good to just count noses. "National Polls" do not break down by region. Look at the 2004 map. MOST of the Congressional Districts are red. Does the Dems no good to win 70%+ in N.E., New York and Cali and lose everywhere else.

So what it boils down to? It's all nonsense. NO ONE can guess how next week will go much less 9 MONTHS from now. This is just more useless noise to give the Political Junkies to gab about.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.