Slicing and Dicing the Senate
Wednesday, August 9th, 2006Below are two syncronized animated charts. The file containing them is large, a half a Meg (sorry about that).
Each point on the charts represents one US Senator; they don’t move. The idea of this technique, “Optimal Classification Estimates”, is to reduce each legislator to two numbers and them pin them onto that graph. It’s extremely reductionist, but it works. The lines slicing thru the chart illustrate how voting proceeds on various bills. A line running from top to bottom reveals that a bill was decided largely on economic issues; while a line running left to right is a bill that was decided on social issues. The democrats on the left are economically liberal, i.e. they tend to look out for the weaker and more numerous economic actors. The Republicans on the right are economically conservative; look after the economically large, but few.
The chart on the right shows how well the model works. The hand full of points shown are Senators who votes didn’t fit the model.
I grab’d that chart from here.
The model is extremely accurate; around 95% these days. Amazingly you don’t actually need two axis; you will get 90% accuracy with a single axis that runs almost top to bottom, but slices slightly at an angle. You can see the entire Senate sorted into that ordering here. For example Joe Liberman isn’t the most conservative democratic Senator, there are a handful who are more to the right than he is.
I’ve writen about this model before; and I keep coming back to it because it totally changed the way I think about politics. It’s all economic; all the noise about social issues never actually flows thru into the legislative agenda.
If you download the chart and stick it into the right program (Quicktime player works for me) you can single step thru the votes. You can then go look up particular votes. That facinating because the topic of a bill may appear to be along one dimension but the vote shows clearly that it was entirely decided by another dimension. For example, the votes around now failed plan to create a guest worker program are a good example of social and economic conservative issues playing off against each other.
Why do I keep expecting John Robb to point out that rust is yet another example that validates his 
AOL recently released a huge sample of search engine queries. In a highly questionable move they tied these queries to reasonably anonomous user identifiers; for example we know that user known as #724 searched for “how to install a glue down floor”, as well as “carbol tunnel” etc. He did 366 searchs between March 1st and May 5 2006.
Links