Tracking the Senate races

Two years ago I published a bunch of diaries on Senate races, with elaborate graphs.  

They’re BAAAAACK!

Method:

 1.  I gathered polling data from Pollster.com.  I only included polls in 2008.

 2.  I assigned the undecided in two ways: 75% to the challenger (per general practice) and 50-50.  The first method is represented by solid boxes.  The latter method by hollow dots.

 3.  I then added loess lines to all 4 series of boxes: Solid lines for the solid boxes, dotted lines for the hollow dots.  Good descriptions of loess (aka local regression) are here  (technical notes on local regression)

Some notes:

 1.   For any poll and for either method, the percentages add to 100.  

 2.  If the hollow box is far from the solid box, then there are a lot of undecideds.

pollster.com has its own charts, and there’s nothing wrong with them.  I like mine better.  First, they have more information. I show different outcomes with different divisions of the undecided.  Second, I like my smooths better.  Third, I assign the undecideds, making things a little less choppy.  

there are  races, below the fold

There are 13 states where the Senate race is judged by 538 to not be “safe” for the incumbent party:

Nate Silver (aka poblano) rates these as

Likely Dem: CO, LA, NH

Lean Dem: None

Tilt Dem: AK

Tilt Repub: MSB

Lean GOP: OR, KY, MN

Likely GOP: TX, ME, NC, OK, ID.

Note that 12 of these (all but LA) are currently Republican held; Idaho did not have enough poll results to track

Since I have little interest in figuring out whether, e.g. John Kerry wins by 30 points or 40 points, I confine my analysis to these 13.

Alaska

If the undecideds break for the incumbent (which seems unlikely, here), the race is even.  If undecideds break evenly, then Begich is up by about 5 points.

Colorado

A steady trend towards Udall, who is now ahead, even if the undecideds vote mostly for the incumbent.  If they break evenly, then Udall is up by 10 points, and his lead is growing

Kentucky

Although the incumbent (Mitch McConnell R) is still ahead, Lunsford is closing the gap.  The increase in undecideds is encouraging — that has to favor a challenger, I think

Louisiana

A worrisome race, in which Landrieu (D) seems unable to gain a solid lead.  

Maine

Allen (D) was doing better earlier in the year; recent polls have been all over the place.  Why Collins (R) remains popular in Maine is hard to figure.

Mississippi (B)

I don’t think a lot of people expected this to be close, but it is. Don’t put too much stock in the lines — there are very few polls.

North Carolina

Relatively little movement, here, and most of that was in the first part of the year.  

Oregon

Here, on the other hand, there is dramatic movement.  If you compare the change in undecided to the proportions endorsing each candidate, it appears that Merkly (D) is winning nearly every person who has made up his or her mind in 2008.  

Texas

Another race where few expected the Republican to be in any danger; lines also based on few polls