Election Update: How Our Model Works
By Nate Silver
Filed under 2016 Election
Want these election updates emailed to you right when they’re published? Sign up here.
I’ve been getting a lot of questions lately about why our forecast model is so bullish on Trump.
For reference: our polls-only model currently puts Trump’s chances at 33 percent, while our polls-plus model puts it slightly higher at 34 percent.
This means we’re giving Trump significantly more favorable odds than the poll aggregators that do not have someone’s name in their title. For example, look at the Princeton Election Consortium, which gives Hillary Clinton a 99 percent chance of winning, or The New York Times, which has her at 86 percent.
Even our own Harry Enten (who is not famous at all) claimed just last month that Trump has to win almost every swing state to have a chance at victory. He had solid data to back that claim up, but there’s one key factor Harry and the other sites haven’t taken into account:
As you can see in the above graphic, our model predicts that me fucking this election up would be “very embarrassing” at best, although a career or life-ruining public backlash also falls within the margin of error.
And we have data to support this. When comparing adjectives that come up when searching for my name on Twitter now, and comparing them to the same search on November 9th of 2012, we see some unsettling trends:
Internal polling done at FiveThirtyEight revealed that just thinking about this data gives me diarrhea 100 percent of the time.
My editorial staff aren’t helping either:
- A Gallup poll released last month showed a 28-point increase in FiveThirtyEight employees savagely owning me in private Slack channels.
- A recent poll from YouGov showed a 12-point bump (from the same time last year) in those who would love to watch my lifeless body slide down the side of a mountain.
- And although there are no polls that have been fully conducted since 2 minutes ago, when an intern from ESPN took a full piss on my neck and shoulders as I wrote this blog, there’s empirical evidence to suggest that the majority of likely voters in Nebraska’s second congressional district would agree the situation was deeply fucked.
For more evidence of this kind of shit, just look at when our senior political writer Clare Malone published a piece called “The End of the Republican Party” almost four months before election day. Does she not know what kind of position that puts me in? I’d invite her to analyze some data from an ABC News/Washington Post poll that was taken earlier today:
I was on Stephen Colbert’s live post-debate show a couple of weeks back and another fucked up thing happened.
You may remember this as the interview where 32 percent of vowels in the words I used were “e”, which is a 2-point advantage over Colbert whose vowel distribution was larger. Anyway, between commercial breaks he leaned in at a 22-degree angle and whispered these 51 words into my ear:
“Don’t fuck this up, famous writer, analyst, and statistician Nate Silver. If you fuck this up they’ll put you in jail. You nailed all 50 states plus DC in 2012, and the number of those you get wrong this time around correlates directly with how many times I’ll kill your family.”
I feel like I can summarize all this with an interesting bit of data that’s been appearing in our model recently.
In our simulations, which forecast each state and tally all the electoral votes each candidate could win, we’ve found that in 83 percent of possible outcomes I wind up dead in a ditch underneath a pile of burning tires. We’ll be monitoring this as Election Day draws closer and the final polling data comes in.
Nate Silver is the founder and editor in chief of FiveThirtyEight.
