Seeking An Oracle

Nate Silver argues that the Democrats’ odds have been overstated all along, and it was never her election to lose.

Incidentally, Clinton slightly outperformed the “fundamentals” according to most of the political science models, which usually forecast the popular vote rather than the Electoral College. For instance, the economic index included in FiveThirtyEight’s “polls-plus” model implied that Trump would win the popular vote by about 1 percentage point. Instead, Clinton won it by roughly 2 percentage points. That’s not a huge difference, but it’s something to consider before assuming that Clinton must have been an exceptionally flawed candidate.

It’s also possible, of course, that Clinton and Trump were both “bad” candidates but that their flaws mostly cancelled one another out. This idea would seem to be supported by their record-low favorability ratings, for example. One bit of pushback to this theory: Politicians are an unpopular lot nowadays, and most of the other men and women who ran for president in 2016 wound up with bad favorability ratings too. So Trump and Clinton’s unpopularity may have been partly an artifact of the partisan political climate.3 Either way, it’s rarely easy to win a presidential election, and Clinton was trying to win hers under more challenging conditions than what Democrats faced in 2008 and 2012.

I agree in part, and disagree in part. I think a lot of us – even people like me who have always been skeptical of the blue wall – overstated the inherent Democratic advantage. I thought Generic Democrat beat Generic Republican, most likely. I also thought Hillary’s favorables would improve in a way they never did, and I thought she would get better as a candidate. But mostly I thought the Democrats pretty cleanly had the baseline, and I no longer think that was the case. If Hillary Clinton had lost to Generic Republican, I’m not sure I would be fair in criticizing her for the loss.

That said, I think Silver is too dismissive of the “cancel each other out” theory. Yes, you can add Ted Cruz’s name to Clinton’s and Trump’s as being unpopular, but come on: He’s Ted Cruz. What all three of them have in common is that they were varying degrees of unpopular from the start. Bernie Sanders and John Kasich never had that problem. Kasich and Sanders may have skated a bit because they were not viable nominees, but Mitt Romney was more popular (or less unpopular) the day he lost than any of those other three were at any point in the race and he certainly faced a lot of head. And while all this was happening, Barack Obama’s approval ratings were inching ever upward. So, not all politicians. These two (three, counting Cruz). Neither was able to capitalize on the unpopularity of the other because they themselves were unpopular.

The single biggest reason Clinton lost was that she was unusually unpopular. She'd have done 3% better if she had even favorability ratings. pic.twitter.com/0hAzLMEQ10

— (((David Shor))) (@davidshor) November 16, 2016

I don’t believe that the election was always Clinton’s to lose, but do believe that it was Clinton’s to lose as soon as Trump got the nomination (arguably, as soon as it was a Trump-Cruz race) unless she was (roughly) as bad a candidate as Trump was. To believe otherwise is to believe that candidate quality and behavior simply doesn’t matter. I don’t think anyone believes that’s true. At least, except in this narrow context in which we try to explain an unexpected result, we certainly don’t act like it. And if we’re going to believe that about the election that just happened, we should believe it about elections going forward. Are we going to commit to the notion that Trump’s approval rating doesn’t matter? That it doesn’t matter how he does? That it’s partisanship and the economy all the way down?

The only out here is to say “Actually, Trump was never a weak candidate and his liabilities were compensated for by his strengths, any Republican nominee would have had similar difficulties, and his low approval ratings were not actually the product of his behavior and candidacy.” There is likely some truth here. I certainly underestimated his strength as a general election candidate. When I look at 2020, this is what I am most worried about, because it means Trump stands a pretty good chance of being re-elected. But, again, I don’t believe it’s a theory that we are willing to commit ourselves to. Hillary Clinton being weak candidate who blew the election is actually among the more optimistic scenarios.

With remarkable haste, the conversation among many political types pivoted back to Received Wisdom and how this race – when you really look at it – wasn’t that remarkable. This race has been playing wack-a-mole with Received Wisdom from the start. The wisdom of the primaries was The Party Decides. Endorsements matter. The establishment wins. Then boom. No worry, though, because as we know there are still the polls. People were wrong in the primary because they ignored the polls. The polls pointed to a Clinton victory. So there you go. Then boom. And now we’re talking about the political science fundamentals and economic fundamentals.

Something, somehow, must explain this and future elections. If we have been led astray, it is because we were following the wrong oracle. This other oracle was right. So now it seems like we’re going to follow this new oracle for four years until it turns out like 2000 when the economic models pointed heavily in an election that ended up so close we didn’t know the outcome until a month later.

Polls can be indicative. Economic models can be indicative. Political science fundamentals can be indicative. So can gut, for some people. And these things are indicative until they’re not. The polls are right until they’re not¹. Economic models tend to be all over the place and are sometimes wrong. We often flail from one oversimplified set of pat explanations (“It’s all about appealing to the middle”) to another set (“Actually, there are no undecideds anymore”).

It’s deeply unsatisfying not to say that we’re all guessing. There is going to be a tendency, going forward, to try to snap everything back into place. To find ways to say that the polling wasn’t really wrong, the political science was right, and the things that we think we know going into 2020 are going to be correct. All of these things are reasonable indicators. Just not good predictors. It’s extremely difficult to come up with good models with such a small sample set. We know a lot less than we thought we did, and sciencing it out is mostly going to be trial and error, and error, and error…

All the same, I do expect in about three years we will be hearing quite a bit about how smart the new, new wisdom is, and that we should trust the Oracle.

1. Even if you think 2016 was pretty accurate on the whole, that would actually make it unusual for the 2014-16 cycle.