Friday, August 5, 2016

UPDATE/EXPLANATION: If Election Held Today, Clinton Would Have 55.5% Chance Of Winning Arizona; Trump Has 69% Chance Of Winning On November 8

It is somehow especially appropriate that on this morning where monsoon storms have been popping up around the state, we would be discussing how election forecasts work. The Arizona Democrats sent out an email today touting yesterday's FiveThirtyEight forecast showing that Hillary Clinton had a better chance of winning Arizona than did Donald Trump.

Yesterday's forecast showed her with a 52% chance (Trump had 48% chance). TODAY, that "Now-cast" shows Clinton with a 55.5% chance. Let's unpack what is going on here.

Here is what Arizona's Politics added to its article last night (in response to a couple of emails and phone calls):

FiveThirtyEight has three different forecasting models. The model now giving Clinton the edge is their "Now-cast", which forecasts "who would win an election TODAY" (emphasis added). Their "Polls-plus forecast", incorporating "polls, the economy and historical data" gives a forecast of who will win on Election Day. At this point, the latter forecast - the one that ultimately matters - still gives Trump a 69.6% chance of winning Arizona on November 8. The headline and lede have been adjusted accordingly.

So, first and foremost, we have to recognize that today's Now-cast is NOT predicting that Clinton will get 55.5% of Arizona votes - either today or on November 8. It is meant to indicate that if the election were actually held today, there is a 55.5% chance that Clinton would get more votes than Trump (majority OR plurality), and thus receive all of Arizona's electoral college votes.

The Polls-plus forecast today favors Trump. It indicates that, as of today, he has a 68.6% chance of winning ON ELECTION DAY. (Not that he would get 68.6% of votes.)

Obviously, as each day slips by, the Now-cast and the Polls-plus forecast models will get closer and closer to matching - on Election Day, they should be the same.

So, the Now-cast is much more "certain", and more volatile. View the charts below: there is a 56% range on the Now-cast, while there is less than an 18% range for the Polls-plus forecast.


I watched FiveThirtyEight's Now-casts on NCAA Tournament (mens' and womens') basketball games. When an underdog opened up a double-digit lead, a graph - the equivalent of the Polls-plus - would often still show the favorite favored to win in the end.

Oops, a few minutes ago, the skies opened up and the rain came down.  There was only a 30% chance. But, that "Probability of Precipitation" was "The probability that precipitation will be reported at a certain location during a specified period of time."  It does not mean the PoP was wrong. My office just "beat the odds".


We welcome your comments about this post. Or, if you have something unrelated on your mind, please e-mail to info-at-arizonaspolitics-dot-com or call 602-799-7025. Thanks.

No comments: