Tuesday, November 26, 2013

Read Stuff, You Should

Happy Birthday to Mark Margolis, 74.

A little good stuff:

1. Sometimes, PolitiFact does a good job

2. I'm really not a big fan of the DW-Nominate assessment of presidents, but kudos anyway to Chris Cillizza for taking the best evidence around, whatever its limitations.

3. Seth Masket on House/Senate. As I said yesterday: guilty, but with an explanation.

4. Sarah Kliff reviews the ACA points you might need for Thanksgiving arguments.

5. Speaking of stuff I don't really agree with...I need to link to this one, from Alan Jacobs and Scott Matthews, although it would tend to undercut the arguments I've been making about the long-term public opinion effect of ACA success or failure. I'm not convinced, but if you're interested in the discussion, be sure to read what they have.

6. And Julia Ioffe explains what happened to Heritage.


  1. Whenever I hear about social scientists who attempt to make quantifiable assessments of a politician's liberalism or conservatism, my immediate question is: how are they defining the concepts? While I roll me eyes at Republicans who describe Obama as a radical socialist, or at the occasional left-wing troll who considers Obama slightly to the right of Reagan, I'm perfectly willing to admit that there's a great deal of subjectivity in these classifications, and I'm skeptical of attempts to quantify them scientifically. What usually happens is that the conclusions end up defying common sense.

    The National Journal rankings, for example, are based on nothing more than how closely you adhere to the party line. If you're a Democrat, then voting in favor of a Democratic Party initiative is counted by the NJ rankings as a "liberal" vote; if you vote against it for any reason, that makes you more conservative. For example, in the 2007 rankings (the same ones in which Obama was ranked the most liberal Senator), Dennis Kucinich is ranked among the 30% least liberal Democratic House members! The probable reason is that he tends to vote against Democratic initiatives on the grounds that they're not liberal enough, which isn't factored into these rankings.

    What are DW-NOMINATE's criteria for what makes a pol more liberal or more conservative? I've searched high and low and have been unable to get any clear explanation--all I see is summaries of their statistical formulas. But I am skeptical of any system that ranks LBJ as more moderate than Carter and Clinton, not just because of how far it departs from the conventional wisdom (even taking Vietnam into account), but because of the many, many differences of what it means to be a "liberal" or "conservative" in different eras.

  2. I think Julia Ioffe really should have asked someone not connected to Heritage about how much is was esteemed as a think tank in the old days. More than it is respected today is not a very high bar.

  3. DW-NOMINATE is just a statistical average that measures the key factor of polarization in congressional votes (I am not a statistician, so I may be off on the technical terminology). Roughly speaking you can call this "liberalism" and "conservativism" but when you extrapolate this over time you run into trouble. Before the New Deal, Southern Democrats were more "liberal" than Northern Democrats according to DW-NOMINATE. You could probably concoct some story about Southern populism, but I
    would bet anything that the real story is that sectional issues were more likely to divide Congress than things we would clearly label "liberalism" or "conservativism". (But then again, there has to be some reason why labor and the New Dealers ended up aligned with the Dixiecrats, rather than the party of Lincoln and Teddy Roosevelt.)

    Taking all this into account DW-NOMINATE can be very useful, but all too often people just act as if it's the last word on how "liberal" a given politician is.

  4. To answer Kylopod's question, it's both tough and not so tough. The procedure finds "dimensionality" in the voting history of Congress. It doesn't just run in one Congress, but in fact uses the idea that we can use those who served in multiple Congresses as barometers for the others. All fine and good you say, but can't they move over time? Well, yes. But, they don't end up doing so much in the NOMINATE data. (I have some problems with this, but think the overall idea that most MC's remain relatively fixed for their careers, particularly for short periods of time, is a good one).

    Now, you say "how do we know who's liberal or conservative?" Well, we don't. What we get are their estimated positions in multi-dimensional space. In practice, two of those dimensions are usually enough to explain over 90% of the variance in voting behavior (maybe even 95%), with a few periods in history (like now) where one dimension really soaks up almost all the variance. What are the dimensions? Well, that's tougher, but it can be done. What you do is look for votes that neatly divide along only one dimension, and then look at the debates surrounding those votes. Poole & Rosenthal have done this and claim that the first dimension is "liberalism/conservativism", essentially defined as the amount of government involvement in the economy. The second dimension has varied over time: slavery, racism, modern social conservativism, etc....

    Here's my problem with it, but I'm not mathy enough to really handle it: if votes increasingly TIE issues together, then the chamber could LOOK single-dimensional, when really it's just our party system collapsing the issue space into two dimensions. Take welfare, for example. Ostensibly, this is just a pure question of how much safety net there should be. However, opinions on welfare are linked to that AND to racial resentment..hello, second dimension! Or Tea Party support: Strongly related to conservativism.....and racial resentment! Yet, the modern Congress reduces to almost entirely one dimension....because the dimensions have become more highly correlated in the public (which is to say slightly) and the elites (which is to say strongly).


Who links to my website?