Visualizing General Assembly Relationships

By: Kathy Gerber
Published On: 3/27/2007 4:07:57 AM

This outlines a method of creating a dendrogram of the relationships among members of the General Assembly based solely on voting record.  The demo here uses the data from votes on 12 bills by the Virginia House of Delegates.  This method can also be useful in other states and for Senate data and HOD and Senate combined.

The best place to start is with the results which are illustrated here.

The underlying data along with the R code used are available here.

A first pass using data from only 6 votes showed little differentiation among those delegates on the left of the dendrogram.  Bills were selected solely for their utility in the model.  Votes that were unanimous would give zero differentiation so were not used.  Otherwise, bills were chosen from a variety of issue domains.  Ideally the full universe of votes would be included, and a different clustering algorithm would be employed.

This is a useful tool for confirming certain aspects of voting behavior and for revealing some relationships that are not obvious otherwise.

Data was obtained from the Virginia General Assembly legislative information system.  Feel free to use the data and R code for your own purposes.  R is open source and may be obtained from http://www.r-project.org/

I doubt if there's huge demand for this information, so if you're not comfortable with running or writing your own program but you want to analyze similar data, let me know.  If you put the data in a spreadsheet coded as in the demo above - 0, 1, 2, 3 encoding with appropriate column labels - it should only take a few minutes to do.


Comments



Ouch! my head hurts (Josh - 3/27/2007 12:17:27 PM)
What does it all mean?

too much information for my tiny brain and limited attention span.

Thag need:  right bad - left good.

Frankenstein says:  Fire... Baaaad.  ;)



Me like fire. (Lowell - 3/27/2007 12:36:12 PM)
Grunt.


Interestingly, my 2 delegates matched votes (Josh - 3/27/2007 12:19:53 PM)
Brink and Toscano are my guys, and they're together.  fascinating.  Am I revealing too much about myself with that observation?


Didn't some guy last week say on this blog (PM - 3/27/2007 2:25:04 PM)
that women couldn't do Math?

My wife thanks you for the "R" reference for use in her own computations.  (She's a STATA and MATHLAB freak usually -- she actually has fun doing stuff like this.)

She may send it to her sister-in-law, the former head of the U.S. actuarial association.  (Also true)

Me, I'm going to move some heavy objects to feel better.

caveman

Questions: Why is Caputo next to Marshall on the chart?  Because they voted similarly on some budget items?  Their social votes are very different.  Would such a chart be useful if you looked at a wide range of votes, or would it look too cluttered?

Hoping your bro is okay . . .



Caputo (Kathy Gerber - 3/27/2007 9:12:28 PM)
is next to Marsden - sorry I couldn't get better resolution. The optimum is going to be data driven, so it depends. There's some balance between getting differentiation and showing some groupings. Those on the right of the chart tended to vote in a block so it took more votes to break them up a bit.

Your wife would probably like R - it's interface is no worse than Stata's and the modules install in a similar way.  You can install the binaries easily on just about any platform. 

Anyway, I'll do the Senate one of these days.



Aging eyes (PM - 3/27/2007 9:27:28 PM)
My aging eyes lumped Marsden and Hall together so it looked like Marshall to me at first read.  I can see it clearly now.


A site with really artistic polling graphics (PM - 3/27/2007 9:43:23 PM)
You might want to check out http://www.pollkatz.... for up-to-date polls, if only for the artistry.

http://www.pollkatz....  is an example.

They even sell tee-shirts with the polls imprinted on them:  http://www.cafepress...