I used to be an admirer of Nate Silver and his empirical approach to covering elections. Not that the horserace ought to be the center of campaign journalism, but since media are going to focus on predicting who’s going to win, it seemed like Silver was approaching it as public opinion research rather than tea-leaf reading. When he left his perch at the New York Times to launch 538 as a freestanding enterprise, I wished him well in his pursuit of journalism that was based on testable information rather than on the opinions of powerful people.
My first hint that all was not right in Silverland was when he confidently declared, despite Donald Trump’s high poll ratings in 2015, that he would not be the Republican nominee (538, 8/11/15): “Our emphatic prediction is simply that Trump will not win the nomination.” Polling more than a year before the election famously doesn’t mean much, but this is a reason to not make predictions, not to predict that the opposite of the polls will happen. But not making predictions is hard to do if you’re in the prediction business, and so in the absence of useful data Silver and his crew substituted their own punditry–with embarrassing results.
As this campaign has gone along, it seems to me that the 538 crew have at times gone beyond the realm of punditry into the realm of hackery—that is, not just treating their own opinions as though they were objective data, but spinning the data so that it conforms to their opinions.
Take a 538 piece the other day (5/25/16) by Silver lieutenant Harry Enten, headlined “Sanders Isn’t Doing Well With True Independents” and arguing that “there is no sign that true independents disproportionately like Sanders.” A “true independent,” in this usage, is one who doesn’t lean toward the Democrats or Republicans; the idea that Sanders does relatively well with such voters is part of the argument that Sanders would be more electable than Clinton in a general election. Citing a recent Gallup poll (but using numbers beyond those available at the link provided), Enten reported that while Sanders does better than Clinton among Democratic-leaning independents, the same is not true with the true neutrals:
In the Gallup poll, Sanders had a 35 percent favorable rating among independents who don’t lean toward either party. Clinton’s favorable rating with that group was 34 percent.
Well, that seems very similar, doesn’t it? Enten added a caveat:
One could argue that Sanders has greater potential with these true independents than Clinton: Just 63 percent of them had formed an opinion of him, according to the Gallup poll, while 83 percent had done so for Clinton. But it’s also possible that these true independents will turn against him in greater numbers as they learn more about him.
Wait a second—comparing the percentage of independents who expressed a favorable opinion about the candidates with the percentage who expressed any opinion, you can calculate what Enten doesn’t give you, which is the percentage of unfavorable opinion for Clinton and Sanders among non-leaning independents. For Sanders, it’s 28 percent; for Clinton, it’s 49 percent. This is what pollsters would refer to as a “sign that true independents disproportionately like Sanders.”
I put it to you that if your headline is “Sanders Isn’t Doing Well With True Independents,” then concealing the fact that he has a net favorable rating among those voters of +7 percentage points, compared to his opponent’s -15 percentage points, is an attempt to deceive your readers.
Another example: a 538 piece headlined “The System Isn’t ‘Rigged’ Against Sanders” (5/26/16), in which Enten and Silver crunch some numbers and claim that Sanders would be doing worse if all states had primaries open to independent voters, compared to the actual mix of caucuses and closed and open primaries. The implication is that this is all Sanders supporters are talking about when they talk of the system being “rigged.” But take a look at how 538 was talking about Sanders before anyone got a chance to vote. Here’s Enten again (6/17/15):
Let’s imagine a case where Sanders wins Iowa and New Hampshire. In that world, you’d likely see the Democratic establishment rush in to try to squash Sanders, much as Republicans did to Newt Gingrich in 2012 after he won South Carolina….
Not only are early endorsements well correlated with the eventual outcome of the primary; in many cases, early state endorsers played a key role in helping faltering campaigns by providing strategic advice and organizational strength.
So from the beginning, 538 argued, Sanders had very little chance of getting the Democratic nomination, because if he showed any signs of winning, a Democratic establishment united against him would step in to “squash” him. If that’s not the definition of a “rigged” system, what is it?