2008 Election Prediction
Like Alex I think trying to predict a presidential election at this point is quite difficult. What makes it hard to predict the presidential election at this point is the primary process and the problems inherent in any voting scheme. For example, the Iowa Caucus for both parties strike me as violating the concept of Independence of Irrelevant Alternatives (IIA). What the heck is that? Well it all goes back to some work done by an economist on voting processes, Kenneth Arrow. Arrow laid out several desirable criteria that any voting scheme should satisfy and he proved an interesting theorem called Arrow’s Impossibility Theorem which asserts that no voting scheme exists that satisfies all of these desirable criteria. These criteria are:
- The Pareto Principle (PP)
- Universal Domain (UD)
- Independence of Irrelevant Alternatives
The Pareto Principle holds that if everyone favors A1 to A2 then the voting scheme should select A1. The Universal Domain does not place any restrictions on voters’ preferences. That is we want the voting scheme to work in a variety of situations. Non-dictatorship should be fairly obvious as to what it is and why it is a good thing. Independence of Irrelevant Alternatives holds that if people prefer A1 to A2 then introducing (removing) candidate X into the mix wont change that ordering. We’d also like the voting scheme to lead to what is called a transitive social ordering. That is if grapes are preferred to oranges and oranges to apples, then grapes should be preferred to apples.
Arrow’s result is that no voting scheme satisfies all of these criteria. If a voting scheme satisfies, PP, UD, IIA and is transitive then it is dictatorial. If it isn’t dictatorial, satisfies PP, UD and is transitive then must violate IIA. And so on with any and all combinations.
The IIA assumption is the most frequently violated assumption and as such it tends to make results unpredictable. Remove a candidate, even one that doesn’t have too much support and things become rather unpredictable as this example shows. That example works off of a Borda Count like voting scheme and is based on candidates from the 2004 Presidential election. In that example the candidate with the least amount of support is pulled from contention and his supporters have to move to other candidates. After the first round the person placing second to last in the first round is now the winner.
As such, I think there is a great deal of uncertainty in who will be the candidates at this point. Of course, Iowa does not determine everything. When Bill Clinton first went on to win the presidency he “lost” in Iowa. Still, the results of the Iowa Caucuses are considered quite important and can have a “bandwagon effect” IMO.
Still, I’m supposed to pick a winner for each party so here goes. I’m going to type all the names into Excel, use the pseudo-random number generator and whomever gets the highest number for each party will be my two choices. So I pick Obama and Giuliani. It’ll be interesting to see how this “random” approach works vs. all the others.