When it comes to politics, I love to get all geeky. Maybe the word is wonky. Or nerdy. I’ve thoroughly enjoyed the last 24 hours. I’ve learned more than I need to know about North Dakota and Connecticut. In addition to politics, I also like figuring out the odds for games with a random factor such as poker, craps, or Risk. I’ve even spent hours calculating things like this: If my baseball team is three games out of first place with twenty games to go, but we’re playing bad teams and the first place team is playing good teams, what are our chances of catching up?
A few weeks ago I wrote a diary about the history of what happened in past midterm elections (Looking at Midterm Elections). I thought it would be fun, in a geeky way, to look ahead to the upcoming Senate elections and figure out the mathematical odds of winning or losing seats in the Senate. Then I turned the boring lists of numbers into pretty graphs.
Follow me under the fold as we peer into the future.
Before I begin, let me say that my analysis is a general overview based on mathematics and calculating odds. It’s macro-politics, not micro-politics. I’ll spend almost no time on specific candidates or specific issues. There are lots of other excellent diaries about that. If I say something like "The odds are good that we’ll lose a seat or two," please don’t ask "which seats?" It could be that we gain three and lose five. Or we might win two and lose three.
It’s like rolling dice or dealing cards. You don’t know for sure what’s coming next, but you can use math to figure out the odds of which events are more likely to happen.
The 36 Senate Elections (plus one)
Here’s a list of the Senate seats being contested in 2010 and how the Cook Political Report (CPR) categorizes them (ranging from Solid Democratic at the top to Solid Republican at the bottom):
Category | 18 | Democratic Senators |
Solid D | 8 | Inouye (HI), Bayh (IN), Mikulski (MD), Schumer (NY-A), Wyden (OR), Leahy (VT), Murray (WA), Feingold (WI) |
Likely D | 1 | Gillibrand (*NY-B) |
Lean D | 2 | Lincoln (AR), Boxer (CA) |
Toss Up | 6 | Bennet (*CO), *CT (Dodd), *DE (Kaufman), *IL (Burris), Reid (NV), Specter (PA) |
Lean R | 1 | *ND (Dorgan) |
. | 18 | Republican Senators |
Toss Up | 4 | *KY (Bunning), *MO (Bond), *NH (Gregg), *OH (Voinovich) |
Lean R | 1 | Vitter (LA) |
Likely R | 2 | *FL (LeMieux), Burr (NC) |
Solid R | 11 | Shelby (AL), Murkowski (AK), McCain (AZ), Isakson (GA), Crapo (ID), Grassley (IA), *KS (Brownback), Coburn (OK), DeMint (SC), Thune (SD), Bennett (UT) |
Note 1: Asterisks* appear next to states in which the winner of the 2004 election is not running in 2010. In some cases, there are several candidates and we don’t know who the party nominee will be (and the current Senator is listed in parentheses). More about this below, in Graph C.
Note 2: Some of you might argue with me about this, but I removed Massachusetts from the list. The special election to replace Ted Kennedy is less than two weeks away (January 19) and polls show that the Democrat, Martha Coakley, is ahead by nine to eleven points. I am assuming that Democrats hold on to Massachusetts. If you’re wondering what CPR thinks, they list MA as Strong Democrat.
Note 3: Obviously, the Cook assessments of the Senate races will change in the future. After Dorgan announced his retirement, CPR moved North Dakota all the way from Strong Dem to Lean Rep (which is bad for Democrats). After Dodd’s announcement, Connecticut went from Lean Rep to Toss Up (which is good for us). And from what I’ve heard about the attorney general who’s running in CT, his chances of winning are better than a mere "toss up."
There are 36 races on the list. Normally there would be 34 in the year 2010, but there are special elections scheduled for Delaware (Joe Biden’s old seat) and New York (Hillary Clinton’s old seat). Also, as I mentioned two paragraphs above, I’m going to blithely assume that Coakley will win the special election for Teddy Kennedy’s old seat in two weeks.
What-If Graph A: Flip A Coin
What if we just flipped a coin 36 times? There are almost 69 billion possibilities (2^36). Fortunately, two tools – Pascal's triangle and a spreadsheet – allow us to easily calculate the results. Here are the odds if everything is totally random:
With 18 Democrats and 18 Republicans, you’d expect something like this. Staying even (60 Democrats) is most likely, plus seven and minus seven are highly unlikely (less than one percent). It’s a bell curve that gradually fades to zero at both ends.
But flipping a coin is silly. That’s too random. Let’s add a factor.
What-If Graph B: Incumbent Advantage
In reality, there are red states and blue states. Incumbent politicians have an advantage against challengers. So I took the information from the Cook Political Report, which is based on PVI’s and past results and recent polls and whatever. I changed Pascal’s Triangle to a Markov chain, which is similar to the triangle but with percentages tossed in (and multiplying instead of adding). I arbitrarily assigned the following values to each Senate race:
Solid | 95% chance the incumbent party holds the seat. 5% chance of an upset. |
Likely | 85% hold. 15% upset. |
Lean | 65% hold. 35% upset. |
Toss Up | 55% hold. 45% upset. |
Note 1: Yes, I know that "toss up" means equally even chances (50/50), but I wanted to give a slight advantage to the incumbent party, so I made it 55/45 in favor of incumbents. Which is still pretty even.
Note 2: Everything is in a spreadsheet, so it’s quite easy to change Strong to 100% or 82% or anything else. Changing these numbers doesn’t affect the results a whole lot (the bell curve just gets slightly shallower or steeper). The fact is, according to CPR’s analysis, there are more Republicans in Solid seats and more Democrats in Toss Up seats (plus poor little North Dakota, which seems to be leaning Republican after Dorgan's exit).
Note 3: The odds of winning don’t equal poll results. If pre-election polls show an advantage of 75% to 25%, that doesn’t mean that the candidate has a 75% chance of winning. It’s probably closer to 100%.
Here’s the graph:
Look at all that red. According to my calculations, the Republicans could easily pick up one or two seats. The grand totals from this graph are: 70% chance that the Republicans gain one or more seats, 14% that the Senate doesn’t change, and 16% that the Democrats gain.
What-If Graph C: Open Seat Adjustment
There are twelve "open" seats (marked with an asterisk in the CPR list above) where the candidate who won the Senate seat six years ago in 2004 is not running in 2010. Here’s a list of who was originally elected. Some have already resigned. Others have announced they won’t seek re-election.
Six Democrats:
CO (Salazar), CT (Dodd), DE (Biden), IL (Obama), NY (Clinton), ND (Dorgan)
Six Republicans:
FL (Martinez), KS (Brownback), KY (Bunning), MO (Bond), NH (Gregg), OH (Voinovich)
Technically, there are two other open seats. I mentioned above that I’m assuming that Coakley will win Ted Kennedy’s seat in Massachusetts on January 19. And some people count Pennsylvania as an open seat because Specter was elected as a Republican in 2004 and now he’s a Democrat.
We don’t even know who will be nominated for some of these open seats. Maybe a lunatic teabagger will get the Republican nomination somewhere. Maybe a nasty primary will split one party (or both) in some state.
Just to see what happened, I removed the incumbent advantage for these twelve races and made them all exactly even (a 50/50 chance for either side). Here are the recalculated odds:
This one looks a little better for the Democrats, although I’ll freely admit it’s extremely silly to count Kansas as 50/50. They’re so red, someone should call them "Bloody Kansas." The last time they elected a Democrat to the Senate was 1932. But it’s also kind of silly to consider New York a toss up.
What-If Graph D: Presidential Results
I wanted to end with an optimistic graph to give my fellow Democrats a tiny bit of hope. So let’s just pretend that all of the people who voted in the 2008 Presidential election show up at the polls and loyally vote the same way for their Senator in 2010. A few days before the Presidential election, CPR ranked the 50 states according to how they’d vote for President (using their categories of strong, likely, lean, or toss up). Here’s the graph based on 2008 Presidential results:
I like these odds – even if they’re somewhat less than reality-based.
Here a list of the states that voted for President in the opposite way of their current Senate seat (and notice that four of them are open seats):
Republican Senators:
*FL – Obama 51-48%
IA – Obama 54-44% (Grassley)
*NH – Obama 54-45%
NC – Obama 50-49% (Burr)
*OH – Obama 52-47%
Democratic Senators:
AR – McCain 59-39% (Lincoln)
*ND – McCain 53-45%
A Few Non-Conclusions
The election is still many months off. We don’t even know, for certain, which candidates will be nominated by either party. There might be another retirement or two. We might see scandals or intraparty squabbles. Teabaggers might somehow sabotage the Republicans (either with primary fights or a Ross Perot variety of third party). The final health care bill and the economy and various other issues will pull voters one way or the other. Cook will publish new Political Reports (and I might even update these charts in the future). Many things could happen. Many things will happen.
I do have one strong conclusion. When a right-wing talk show host or a blogger predicts that the Republicans will take over the Senate in 2010, they’re full of crap. The odds of that are nearly zero.