I stumbled upon this article about Who Will Win the 2008 U.S. Presidential Election?. And realized that what they we claiming here was not a predictive method, but merely these people’s attempt to fit, to pattern match, ‘Presidential Electability’ factors to winners of elections. And making a half hearted attempt to apply these historical factors to up coming election. I’ll give you some examples, these are the factors they used in a previous ‘prediction’ Who Should Win the 2004 U.S. Presidential Election? they used factors like
if the candidate served as Director of Central Intelligence (e.g., George H. W. Bush), was a four- or five-star general officer in the United States Armed Forces (e.g., Dwight D. Eisenhower), or ordered the combat use of nuclear weapons (e.g., Harry S. Truman),
as you can see such ‘factors’ could only be applied to demonstrate why Truman beat Dewey, not as a general rule that proves the model.The factors used in the 2008 ‘prediction’ only suggest combinations of Presidential candidates , but can not predict them, because the factors are based on historical artifacts as part of the pattern matching, factors that can not be predicted in advance.
With that in mind there are no factors for emotional content, or future historical events.
They further go on to make this statement, We assume that major party primary voters are rational and will understand the empirical power of our algorithm. and we all know that voters for both nominees or candidates are not rational. And with a model that does not factor the level of ‘rational’ or irrationality of the electoral process, they can not forecast or predict. In other words, they can’t put their money where their mouths are.
They make another bold statement Readers of this paper could make a lot of money at Intrade.com if our assumption that major party primary voters are rational is correct. but you don’t see them making money.
The old adage is this, never take investment advice from a poor broker.