And, how did my model do?
There was a lot of talk about California, and a lot of back and forth, but in the end I stuck with my original model to predict the outcome of that race. See the table above for the results, but the bottom line is that I predicted that Clinton would get 57 percent of the votes and Sanders 43 percent. It turns out that Clinton got 57 percent and Sanders got 43 percent.
Excuse me for a moment while I bask in the bright light of being-right-ness.
Thank you. Now, on to the details.
First, a quick, note on the numbers and methods. All my percents (for prediction and as reported for the outcome) are the proportions of each candidate’s take of the two candidates, so “other” or “The Lizard People” or anything other than Clinton or Sanders are taken out of consideration. In some cases this will cause the numbers to look different than those reported by the press. The awarded delegates I provide here are from the Washington Post, and often do not add up to my predicted proportionate amount. This is because the process of awarding delegates is complicated and bizarre. Eventually the numbers of proportionate delegates will settle to be very close to those you would get form using the percentage of votes for each candidates.
The outcome of yesterday’s primaries was pretty much as expected, but not exactly. Polls and my model both seemed to predict that Clinton would win New Jersey by a large margin, California by a good amount, likely New Mexico, and that Sanders would take Montana and the Dakotas.
Clinton ended up doing better in New Jersey than expected, but in the case of landslides, the final numbers are often a bit off probably because of some fundamental behavior of variance. California was as expected, as was Montana. Sanders did much better in New Mexico (a closed primary, by the way) than expected, but still did not win.
The Dakotas are the enigma. The expectation was that Sanders would do very well in both states, better in South than North. It turns out that South Dakota totally reversed, with Clinton winning by four percent. In North Dakota, Sanders wiped Clinton out, not only winning by a large amount as expected, but trouncing clinton with what must be one of the highest margins all season.
With respect to my model (detailed here), I think we are looking at sample size and a few other things. I was within a fraction of a percent in the largest state, and the smallest states were the oddest. But, I also suspect different campaign efforts by the different candidates played a role. Also, when we talk about openness of the primary (or caucus) it is important to note that not all contests have corresponding Republican contests going on at the same time. That may be a big factor in the Dakotas.
In the end, there are two big winners today. Hillary Clinton had a resounding victory in the largest state, and did very well across the board otherwise. This comes hours after the press deciding to declare her the Winner-Apparent based on math, and it verifies that math. Sanders has continuously said he would fight to the convention, attempting to overthrow the process using super delegates. He seems to have not noticed that the entire Democratic Party is mad at him, even former Sanders supporters, and the super delegates’ job is actually to make an effort to maintain the spirit of the process when something goes wrong. Sanders is the thing that is going wrong at the moment — with his effort to reverse the democratic process — so there is zero chance that the Supers are going to come to his aid.
The second winner is, of course, Science by Spreadsheet. I’ve been running spreadsheets on elections since spreadsheets were invented, and this is the best cycle I’ve had. I’m pretty sure my model out performed all the other models. Perhaps I will summarize all that in another post at some point.
Can’t wait to get started on the electoral map.
I should mention that DC still has a primary to go, and it will go overwhelmingly for Clinton.