I've been creating some new models for predicting elections. I didn't feel very good after the 2004 elections because my models weren't anywhere near 95% accurate (especially on state issues). I've been working hard with the numbers. I'm using models now that instead of predicting percentages of the vote (like my last batch) now predict the likelihood of passage.
Using these new models, I now predict that Referendum C has a 20% likelihood of passage. Referendum D has a 16% probability of passage.
The current models only explain 80% of the results, in the coming months, as I add in more data from prior years, the models will get better and the numbers will change.