[muawiyah]

Usually no one WINS by getting less votes than his same party predecessor ~ McCain got fewer votes than "W" ~ but in this case Obama has to lose 10 million votes before he gets down that low so there is a cushion of sorts. (correction)

Don't just tut tut models that show Obamugabe winning ~ show the specific errors ~

[Cruising For Freedom]

Don't just tut tut models that show Obamugabe winning ~ show the specific errors ~

You've got it backwards. Proving that the data models use isn't predictive isn't necessary. The models need to rigorously why the data they have constitutes a correlation.

This is no better than layman stock market technical analysis or data mining. Coming up with models is a hard science for quants getting paid 400K/year, one well beyond the means of polisci majors.

The only predictive power is sampling the actual votes (by the means of asking people who they will poll), and even then, there is a lot of art and error in sampling that.

Other than that? Unless some megabucks hedge fund firm is kind enough to open-source a model they spent millions and millions on, it's all bull. You are, quite seriously, no better off using the models cited in the article than using the "Halloween Mask Purchase" and "Cow Poops On An Obama/Romney Square" predictions.