Election Projection Statistics: Do your own analysis

Introduction

Imagine if you could run analysis of the odds of which Presidential candidate was likely to win like the pro’s. Without a pencil sharpener or even a calculator.

Thought Experiment…
First off, let’s understand how the election is won. It’s not won on popular vote, it’s won on the electorate college count. If you win a state, even by a hair, you get ALL of the states electorate college votes. Despite there being 50 states in the US, only 7-10 states determine the election. That’s because the other 40+ lean so far Republican or Democrat that their outcome is known well in advance. California will vote democrat, Texas will vote Republican.

So if there are only 10 states, couldn’t we see who’s ahead in those states and reasonably guess who’s likely to win? Yes, but let’s start on a more simple level.

Imagine the whole election hinged on one swing state, Nevada, and polling shows both the Republican candidate and Democratic candidate in a tie. What are the possible outcomes? It could go R or could go D. Easy. Two possibilities.

Now imagine there are two swing states. Possible outcomes: RR, RD, DR, DD. Not too hard. It’s like math problem from middle school.

Three swing states grows 8 possible outcomes. At then states, there are 1024 possible outcomes.

Figuring out all the outcomes gets a bit hairier when each state has a different value of electorate votes (Florida is 29, New Hampshire is 4), you’re trying to reach the finish line of 270 electorate votes and one candidate is already starting ahead of the other, and there is a different likelihood of your favorite candidate winning some of these swing states. Meh. That’s a mathematical story problem that sounds more fun to just take a -1 red check mark on your homework and skip.

The Solution…
Enter, Christian Anderson, a Harvard post-doc. I proposed a problem to him:

“There are 10 swing states left. I want to know how many different pathways Obama/Romney could win. Ideally, I’d like to be able to assign probabilities (ie, Michigan has an 85% of going to Obama, North Carolina has a 75% chance of going to Romney) to more accurately help determine each candidate’s probability of winning.”

He is a brilliant solver of mental problems. He made me a computer program that answers both questions. I’m making it publically available for politicos to download who want to figure out the chances of their own candidate winning while the computer does the heavy mental lifting.

Here’s how…

1. Download the main program, R, which is math calculating freeware, available here from UC Berkley. Install the program.
2. Right click this link and select Save Link As. Download the political calculator data file “2012ElectionAnalysis” (recommend saving to desktop to make it easy to find).
3. Go to desktop, and double click “2012ElectionAnalysis”. It will load the program interface.
4. Type: z and the program will show you the list of swing states and the current probabilities assigned to each (.50 probability each)
5. Copy and paste: elect(c(Dem=221,Rep=191)) and the program will now show you the probability for the democrat, Obama, winning; and the probabilitiy for the Republican, Romney, winning assuming 50/50 odds for each swing state left.

Walah! You effortlessly calculated the statistical odds for the election, given each swing state has equal odds of going to Romney or Obama. Now you can see the math, it’s an interesting outcome, eh? But swing states aren’t equal…
If you are feeling curious and want to adjust the odds – say you think North Carolina is more likely to go to Romney, or New Hampshire and Michigan are more likely to go Obama, what would the odds be then?

Copy and paste: “>z<-edit(z) This will allow you to easily edit the data.

The far left column is the state, the next column is the number of electorate votes it carries. The far right column are the odds. .5 is 50/50 odds. .6 means 60/40 odds for Obama, .9 means 90/10 odds for Obama. .4 means 40/60 odds against Obama or said differently, 60/40 odds for Romney. .1 means 10/90 odds against Obama, or 90/10 odds for Romney.

Enter your predictions. Close the window. Re-run the elect formula, elect(c(Dem=221,Rep=191)), and now you can instantly see what the odds are of your candidate winning given your new assumptions.

You can find some good “odds” information at the New York Time’s 538 blog, on the right hand side, under State By State Probabilities.

My favorite is to see how the odds change from the original 50/50 for all swing states, to giving Obama Ohio, which he currently leads.

Conclusion
I’ve run this with a variety of different scenarios that I think are pro-Romney, neutral, and Pro-Obama to get a feel of the reasonable odds of either candidate winning. It’s quite illuminating and definitely not 50/50.

If you do this, private message me what you think the odds will be and what your assumptions where (copy and paste the values from z to me). We are also doing a small group beting on the election outcomes. If you’d like to play, let me know. 🙂

Special thanks for Christian for making this possible.