Controlling the Agenda
Owen Rice has a series of cool charts that show optimal classification location of members of congress and “cutting lines” on various votes. The way it works is first you locate members, based on their votes, into a two-dimensional ideological space. Then on any given vote you can create a “cutting line” across the ideological space that does the best possible job of correcting sorting members into yeas and nays. That helps you get a sense of the underlying dynamic of the issue.
Here’s an example of a vote that broke down on pure party lines:
Kevin Drum wants to know why the optimal classification comes out this way:
Actually, though, I think I’m more interested in the placement of senators themselves. Democrats are almost all bunched into a single grouping, with only four outliers. Republicans, by contrast, are spread through considerably more space on both the economic and social dimensions. That doesn’t seem intuitively right to me, but it strikes me as more complimentary toward Republicans than Democrats. So tell me again why they want to defund pointy-headed political scientists?
It’s not intuitively right. What I think it is is an illustration of the importance of setting the agenda. The Democratic leadership has only brought to a vote bills that unite the overwhelming majority of Democrats. Consequently, a visualization based on votes of the 111th Senate shows the Democrats as enormously bunched-together. If you look at the House where Nancy Pelosi doesn’t need a unanimous caucus to pass bills, you see that Democrats and Republicans are about equally dispersed. If Republicans were to capture the House and pick up some Senate seats in 2010, then legislation would more often be focused on issues that split the party caucuses (education, immigration) and the visualizations would look different.
16 Responses to “Controlling the Agenda”
Register
RSSLogin
Topic Cloud
ArchivesBlog RollAbout Matt YglesiasContact Me
Donate
November 7th, 2009 at 1:24 pm
“The Democratic leadership has only brought to a vote bills that unite the overwhelming majority of Democrats.”
You are right about this Matt.
That is why ‘elections have consequences’ and real power is in setting the agenda.
November 7th, 2009 at 1:37 pm
The Democratic leadership has only brought to a vote bills that unite the overwhelming majority of Democrats.
Or, Democrat Senators just do what they are told to do.
November 7th, 2009 at 1:50 pm
Democrat Senators just do what they are told to do.
You’ve got to be a joke troll.
November 7th, 2009 at 2:14 pm
Neither you nor the source explains the “Social Issues” axis. Is it a libertarian vs. authoritarian axis? If so, what does the upward direction represent?
November 7th, 2009 at 2:45 pm
Or, Democrat Senators just do what they are told to do.
My irony meter just broke.
November 7th, 2009 at 2:59 pm
If this is based off of DW-NOMINATE (which I think it is, because it looks similar, and has a link to voteview at the bottom), then I think that by “social issues” they actually mean geographic issues. Poole, who invented DW-NOMINATE, believes that, while about 90% of votes can be explained on a liberal to conservative scale, about 10% can be explained by a scale of regional voting (think civil rights, free silver, cap-and-trade, etc). Higher up on the vertical axis means more “southern” on regional issues (this doesn’t translate exactly, as people in parts of the midwest and appalachia often vote in a “southern” way on regional issues) and lower down on the vertical axis means more “northern” on regional issues (which, again, is not a perfect description, as the west tends to vote in a northern way).
November 7th, 2009 at 3:28 pm
There are any number of lines that divide the set of points in half. What decides which line is chosen? Is it maximum average distance from the line?
November 7th, 2009 at 4:02 pm
Ah, another “who flung do” chart. Impressive.
November 7th, 2009 at 4:03 pm
What decides which line is chosen?
The usual goal is to minimize classification error. If the classes are linearly separable, as in the plot above, then there are various ways of picking the “best” line among those that classify the points correctly.
November 7th, 2009 at 4:42 pm
Thanks Mac, that’s helpful. I’ve seen these charts before and thought they’d be more helpful if I knew what the axes meant. Now that I know, I’m not sure I agree with the choice of axes, but at least I sort of get what they’re trying to say.
November 7th, 2009 at 5:25 pm
So, because the study ignored all the votes where Republicans voted together, they look better.
Gosh! Who’d a thunk!
November 7th, 2009 at 6:35 pm
4/6/10: I think Poole usually says that, throughout most of the US’s history, the second axis represents attitudes toward race. Here’s a quote that I just found: “A one-dimensional model typically provides a good fit to the data, with a second dimension needed in periods when race issues are distinct from economic ones.” So “regional issues” here usually means “civil rights”; though I guess there might be times when it means “free silver.”
It’s important to understand that they don’t start by trying to figure out where congressfolk are on a liberal/conservative or southern/non-southern axis and then drawing lines for the votes. Instead, what they’re doing is to say, “Let’s see if we can plot everyone on a two-dimensional space somehow, so that we’ll be able to draw a line across the space for every single vote so that the yeas fall on one side of the line and the noes fall on the other.” You need a lot of different votes for this to have any meaning; if you only had two votes, you could just put everyone in one of four corners, corresponding to the four possible combinations of yes-no votes. But what they’ve found is that if they plot everyone in two dimensions, they can do a good job of drawing the lines (and going up to three dimensions doesn’t really help any). So that plot is the two-dimensional plot that allows them to do the best job of carving up the votes. None of that requires any interpretation of what the axes mean; that interpretation is something they’ve added to a chart they derived without any interpretation.
In fact, right now you can do a pretty good job of predicting everyone’s votes by plotting them on a single liberal-to-conservative axis and finding a breaking point for each vote. The second axis doesn’t add that much accuracy. (That’s why the line is almost vertical.) Part of that is that the people who are most liberal on economic issues are now also most liberal on racial issues as well. Back before the Dixiecrats left the party, though, there were probably a lot of cases where, say, a vote on civil rights would give you a horizontal line (the northerners voted for it and the southerners against), and a vote on a tax bill might look more vertical (the liberals voting one way and the conservatives voting another).
Hope this is useful and accurate. (I did some programming work for Poole and Rosenthal one summer, and spent a lot of time looking at these charts.)
November 8th, 2009 at 12:31 am
Matt W is correct (as is Mac, although Poole’s Optimal Classification is different from NOMINATE). The axes resulting from this type of estimation are a product of the voting behavior, which in turn is indeed subject to the agenda. The “social issues” notion is Poole’s traditional label (”civil rights” in his classic work with Rosenthal). People not familiar with that work often confuse it with the social issues that are already captured by liberal/conservative. When just recent votes are used, the meaning may seem even more peculiar. All it really means though is “the dimension of the most systematic variation orthogonal the first dimension.” This just happens to be historically related to intra-party regional distinctions on social issues (possibly related to libertarianism in some cases) that result in a variety of relatively infrequent cross-party coalitions. It is true that the second dimension of voting is not that important for predicting votes overall or in placing members (if you just want to have a rank order of members based on current behavior you can find that at VoteView too), but it usually explains at least a few more voting decisions. Some votes are much better predicted when two dimensions are considered. This may be important if you are interested in the location of a given cutting line, rather than just on the locations or rank of the members.
November 8th, 2009 at 9:30 am
That doesn’t seem intuitively right to me, but it strikes me as more complimentary toward Republicans than Democrats.
That’s because he doesn’t understand what the chart represents. It’s not a measure of diversity of opinion. The chart is a tool to predict voting patterns. Each Senator is positioned on the chart in a way to optimize that prediction and are swept into position on the chart by votes actually taken.
If Harry Reid ensured that every vote got unanimous support from Democrats all the Democrats would plot on the same point in space. If he occasionally peels off random Republicans, they get more widely distributed in space.
November 8th, 2009 at 4:00 pm
I don’t think the interpretations expressed by either Drum or Yglesias are correct. It appears to me to be a natural outcome of what a majority left and a minority right party will look like in this graph.
The key measures are that while the Democrats form a tight clump, they are closer to the center of the diagram, and while the Republicans are more diffuse, they are flung to the far-right edges of the diagram.
The majority party will capture the center, and with with larger amount of averaging over more voters, you get a tight clump. Matt’s point about the House probably just reflects fewer constituents and less averaging. The Democrats still remain clumped and closer to the center while the Republicans are farther from the center.
Since the right hasn’t captured the center, they are on average farther away than the democrats (and towards the right). With lower numbers of voters for the minority (and a bunch of smaller states as well), you get a more diffuse cloud.
I predict that if the Republicans were to take the Senate at some point the graph would look like this one in reverse: Republicans tightly grouped nearer the center and Democrats flung out towards the left edge.
Now whether that will be a cause (Republicans recapture the center with centrist policy) or effect (the center of American politics drifts rightward) relationship, I can’t say.
November 9th, 2009 at 3:20 pm
Given Bernie Sanders
Therefore, this chart just not reflect actual political views.
:End