How to Sort Out Nate Silver’s FiveThirtyEight Data Leading Into the Midterm
Nate Silver is a talented man. Talented enough that what started out as a political statistics blog, FiveThirtyEight.com, has been bought and sold and expanded and fooled with a few times since its inception during the 2008 election.
His schtick, for lack of a better phrase, involves taking polling data, applying a bunch of additional statistical data to it (e.g. how far from the final election result each polling “house” tends to be, what the demographics and past voting performance in a particular state/district have been, what broader polling data can tell us about narrow questions, etc.).
The problem is, most people take a look at the “probability” percentages he assigns any outcome, and go into immediate excitement or despair. So let me break down what Nate is telling you about Tuesday’s election:
Think of his probabilities as the sides on a die
Right now, according to Nate, Republicans have a 4-in-5 chance of holding the Senate, and Democrats have a 7-in-8 chance of taking the House. What does that mean, exactly?
The easiest way to think about this is, for the Senate, take a 5-sided die (yes, I checked, they exist), and color four of those sides red. Then take an 8-sided die (again, yes), and color 7 of those sides blue for the House.
You still have to roll the die.
That’s how probability works. Selecting the number of chances, the number of chances each outcome has, and then taking the chance.
The Democrats have a slightly better chance of winning the Senate than Donald Trump did of winning the Electoral Vote while losing the Popular Vote
In 2016, the final probability Nate calculated for Donald Trump winning the election without taking the Popular Vote was 10.5%.
That…doesn’t sound like a lot, does it? Well, tell that to President Trump.
70 percent doesn’t mean squat
If you look at Nate’s ratings of various races, 70 percent (or thereabouts) probability comes up in a lot of races. That’s roughly the point at which a “normal polling error” (that is, an error the size of a poll’s “margin of error”) can throw it the other way.
If you want to think of something as, more-or-less, “certain,” that race needs to be north of 90 percent in Nate’s ratings. But…
85 percent tends to mean multiple things have to go wrong
And Democrats are sitting at 87 percent probability, as of this writing, to take the House. There are still scenarios where Republicans hold on (Nate explains all this here) but it would take an unusual polling error for it to happen.
Susan Collins and Lisa Murkowski may become the most powerful politicians in America for the next couple years
If all goes exactly as Nate predicts, down to even the slight advantages (and it wouldn’t be the first time), we’re looking at a 50-50 split in the Senate, with the Vice President casting votes constantly.
That means Republican Senators Susan Collins of Maine and Lisa Murkowski of Alaska, the last real “independent-minded” Republicans left in the body, will be calling all the shots. If Republicans want to get anything passed, they will not be able to lose a single vote (as the Vice President only votes in the event of a tie).
And, if Nate hits 100 percent again, Democrats are looking at losing North Dakota, while picking up Arizona and Nevada. Republican efforts to pick off any other Democrats, including in Indiana, Missouri, Florida, West Virginia, and Montana, will come up short…again…if Nate hits 100 percent.
If there is a systemic polling error, Democrats have the momentum
Let’s get something out in the open real quick…
The polls have been accurate. Even in 2016. They were as accurate as they usually are. So what happened?
We happened. We couldn’t make up our mind of who we liked enough to turn out and vote for (making “Likely Voter” modeling by the pollsters more difficult).
But the final election results came in within that general margin of error. And things get messy when that margin of error crosses the victory line. And it crossed for Trump in 2016. Trump had the momentum leading into Election Day.
This time, Democrats do. For the past few days, Democratic probabilities have been on a slow-and-steady rise in both the Senate and House.
Midwestern state capitols may throw a collective big, blue switch
Right now, the only state in the Midwest to have a Democratic governor is Minnesota.
However, if all pans out the way Nate’s probabilities show (again, a very very big “if”), Iowa, Illinois, Wisconsin, Michigan, and Ohio will join them. And if you throw in the thought of a slight polling error again, Kansas and even South Dakota may go blue.
But, almost all of those races are super-super tight. Of those eight races, the leader in four of them (Ohio, Wisconsin, Iowa, Kansas) has a probability below 60 percent. That means you can more-or-less throw a dart at the board with those.
Don’t expect Nate to land another 100 percent performance
It’s not his fault, really. It’s that Americans are growing increasingly polarized, and polarization affects turnout. For some, polarization energizes. For some, polarization alienates.
When you add in unlikeable candidates running campaigns (or even careers) chock full of controversy, that throws a wrench in public opinion making polling more difficult.
And if YOU can’t figure out who you like more, how the heck is Nate supposed to do it?
This really clarified the probability/ percentage issue for me. Thanks!