Skip to content

More evidence of growing nationalization of congressional elections

The other day I posted some evidence that, however things used to be, congressional elections are increasingly nationalized, and it’s time to retire Tip O’Neill’s slogan, “all politics is local.” (The discussion started with a remark by O.G. blogger Mickey Kaus; I also explain why I disagree with Jonathan Bernstein’s disagreement with me.)

Alan Abramowitz writes in with an analysis of National Election Study from a recent paper of his:

Average Correlations of House and Senate Votes with Presidential Job Evaluations by Decade

Decade House.Vote Senate.Vote
1972-1980 .31 .28
1982-1990 .39 .42
1992-2000 .43 .50
2002-2008 .51 .57

This indeed seems like strong evidence of nationalization, consistent with other things we’ve seen. I also like Abramowitz’s secret-weapon-style analysis, breaking the data up by decade rather than throwing all the data in at once and trying to estimate a trend.

P.S. Another interesting comment came from Sean Scanlon, responding to my post on fivethirtyeight. Scanlon wrote:

As a BC grad who spent many hours studying politics in a library named after Tip, I must say I think you misunderstand the saying “All politics is local” in the context in which he originally coined it. The phrase, if I remember correctly (though I’d have to dust off my copy of his memoir “Man of the House”), comes from the first time he ran for city council in Cambridge. Tip made the decision not to spend a lot of time in his neighborhood assuming that, since he knew nearly everyone there, he did not have to work for their votes. On Election Day, he saw a woman who lived near him named Mrs. Cleary if she voted for him. When she replied that she had in fact not voted for him, he was appalled. After all, he lived next to her for years, had shoveled her driveway, etc. When he asked why she didn’t voted for him, she responded: “Because you never asked. People like to be asked.” It was based on that experience that he coined the phrase, not because he felt that local concerns outweigh national ones but rather that never forgetting your roots and taking things for granted is a recipe for disaster in politics. To me, Tip meant the phrase to mean a lot more than simply being well connected in the clubby world of local politics (though he certainly placed high stock in that); it meant staying on top of things in your district. In doing so, you are not untouchable in wave elections but it certainly can make a difference as was the case in several House districts last fall.

I’d like to make a few points in response:

1. I agree that “all politics is local” when you’re running for city council! My problem is the common use of O’Neill’s quotation to imply that local politics is all-important even for U.S. representatives.

2. I agree with Scanlon that the “local” in the O’Neill phrase relates to direct work with constituents rather than backroom dealing. That’s why I thought Bernstein’s objections to my argument were beside the point.

3. I agree that local politics can make a difference–all the evidence shows that (a) there is an incumbency advantage and (b) this advantage varies, with some incumbents being stronger than others. But local politics doesn’t matter as much as it used to, and national factors seem to me to be much more decisive.


  1. Gaurav Sood says:

    FEC data on fund-raising would likely provide a stronger picture of the nationalization of elections.

  2. numeric says:

    Congressional elections used to follow national trends
    very closely. Elections where things "changed" somewhat dramatically at the Congressional level were
    1946, 1948, 1952, 1958, 1964, 1966, 1974, 1980, 1986,
    1994, 2006, 2008. These elections do not seem to be
    particularly associated with any decade. Aren't you the
    one who has talked under other circumstances about the
    variability of small estimates and how they often produced
    erroneous estimates (type M, type S, etc)? And I seem
    to recall the praising of a regression for attractive
    offspring as the more usual analysis rather than taking
    one particular quintile, and a figure (Figure 2, as I
    recall) showing how putting standard errors on the regression lines could produce type S errors. I guess
    I'd like to see a regression also for all elections, and then for each decade, and whether one can get type S errors.

  3. Andrew Gelman says:


    I have a graph here showing jumps in the national House vote over time, and indeed previous decades have shown some large jumps as well.

    I get frustrated when people look at changes in seats as a reflection of the mood of the people. Changes in seats are an outcome, and they're politically important, but it is votes, not seats, that measure opinion. (This paragraph is not a response to your comment, just a general thought that came to mind.)

    Finally, regarding your comment above: I'm just reporting that Abramowitz's results are consistent with other things we've seen. This is a little different from the sex ratio estimates which were not in accord with other things we know about human sex ratios. I would have no problem with someone redoing Abramowitz's analysis in other ways and with other data.

  4. Steve Roth says:

    > secret-weapon-style analysis

    I find it an especially useful method as well (as you've seen), being able to eyeball trends and judge whether they're consistent or have weird outliers. (This one's consistent trending is quite convincing to my human judgment, thumb held up and squinting.)

    But why do you call it secret-weapon style? Is there more to know there? One reader is very interested.

  5. Steve Roth says:

    re: secret weapon, never mind. I looked it up. Sorry.

  6. Dan says:

    More to know about what Steve?