Why Predictions Matter (Even the Wrong Ones)
At the end of the year, it has become something of a tradition for columnists to reflect on the things they got wrong over the past 12 months. This year, however, we all got the same damn thing wrong. At times I thought that Donald Trump might really be elected president, but I always got sucked back into the groupthink of the pundit class. Trump had no ground game, his campaign was incompetent, he had absurdly high unfavorables, and his fundraising was weak. We were right about all of these things; we were simply wrong that they were a decisive bar to the presidency -- at least when facing a candidate as bad as Hillary Clinton.
So instead, I’m going to reflect on something I got right a few years back -- which is, funnily enough, the same thing I got wrong.
In 2013, while temporarily between Newsweek and Bloomberg, I tossed off a quick blog post on my own site about whether Democrats should get rid of the filibuster. I thought this was crazy: Assuming there was a 70 percent chance that Republicans would control the White House and both houses of Congress in 2017, I said, the Democrats would be hanging them a weapon they would then gleefully use to push through their agenda.
Commenters went wild with incredulity. A 70 percent chance? More like a 0 percent chance, I was assured. Certainly, much less than 50 percent. Where are you getting that number?
The answer: my gut. “Something well north of 50 percent, but not a virtual certainty” would have been just as accurate. And I explained my reasoning in a follow-up post. This was the core of my argument:
- The most likely Democratic presidential candidates were Clinton and Joe Biden, because either of them would suck all of the oxygen away from other candidates -- and both had grave weaknesses.
- It’s hard for a party to hold onto the White House for more than two terms. Since World War II, only Ronald Reagan has managed to bequeath a third term to his party without either dying in office or resigning.
- It looked like 2014 was going to be a good year for Republicans in the Senate, and a White House win would give them a boost even in 2016, when they faced a tough Senate map.
I thus predicted high odds that Republicans would take the White House, keep the House, and through presidential coattails, achieve a majority -- though fragile and vulnerable -- in the Senate.
And boy, I got a lot of pushback. This wasn’t scientific (indeed it wasn’t, and the post was appropriately bracketed with notes that this was “entirely provisional” and quite possibly wrong). I had failed to understand that Republicans were basically barred from the White House because of their failure to connect with Latinos and other voters of the future. Republicans were clowns. And so forth. Nate Silver argued that the likelihood of Democrats taking the White House for a third term was probably more like 50-50; Nate Cohn put the odds of a trifecta at more like 15 percent.
This may be the point where you expect me to say “Ha, ha, I told you so!” Instead, let’s consider the reasons that, even though I got the right outcome, that post might have been wrong.
Take, for example, this paragraph:
Democrats who think they’re a shoo-in seem to be unaccountably banking on the GOP nominating some tongue-tied wingnut who will spend the campaign discussing the scientific evidence that women can’t get pregnant from rape. But as Joe Scarborough argued in 2012, this is wishful thinking . . . in his words, “The GOP doesn’t nominate crazy.” In 2012, out of an incredibly weak field filled with tongue-tied wingnuts, they nominated the moderate with the best public policy chops and solid debating skills. In 2016, they will have a much more attractive bevy of candidates from which to choose someone electable.
Now, my Trump-supporting readers will be angry to hear me say I got this wrong. But let’s leave aside the debate over whether or not he’s actually crazy, because the point is that if you had described his electoral strategy to me in 2013, I would certainly have identified it as crazy. So I got two things wrong in this post: First, the GOP did nominate someone pursuing a crazy strategy, and second, that was not the bar to the presidency that I would have assumed three years ago.
Now I still think that I was right about some things: the weakness of Clinton as a candidate; my skepticism of the folk version of the “emerging Democratic majority” thesis; the idea that it gets harder to hold onto the White House as time goes on, simply because voters blame whoever’s in it from everything that goes wrong in their lives. But if you get the outcome right but the mechanism wrong, you have to wonder if your prediction’s success is due to genius, or dumb luck. Even a stopped clock is right twice a day, and this could be one of those times.
Unfortunately, we’ll never quite know. My seat-of-the-pants prediction, and Silver and Cohn’s more carefully reasoned ones, cannot be falsified by looking at the outcome of an election. One-in-six events do happen every six times (on average), and cannot be distinguished from a two-in-three event if you only have a single data point. And this was a close race, not a blowout. I might have gotten the right answer with flawed reasoning, they the wrong answer with sound logic. Since each set of electoral predictions is unique and hard to compare to the others, you can’t even amass a series of such predictions to test pundit power.
Which is why judging pundits on predictions is a mug’s game.
In truth, punditry is not prophecy. It is not our job to tell the future; if we could do that, we wouldn’t need jobs, having already retired on our fabulous wealth. What pundits do is walk people through their reasoning about the world, and the evidence on which they make their decisions. Sometimes we’re seeing phantom patterns in noisy data; sometimes the pattern we see is real enough, but we lack crucial data to contextualize it. Nonetheless, we can provide information readers didn’t have, and a different way of looking at a familiar problem.
And that’s why I think, for all the embarrassments that ensue, it’s a good thing that pundits make predictions. In the run-up to the Iraq War, for example, it was useful for voters to be able to see why some people thought it would succeed, and others that it would fail. But predictions are also good for us, the pundits, because they teach us humility. My confidence in the success of the Iraq War taught me to be wary of my own confidence -- to understand that no matter how much data I have, I still see but through a glass, darkly. I have never been so confident again in any judgment I made, and I think that my judgments, and my readers, have been the better for it.
I still make predictions, of course -- sometimes after careful analysis, and occasionally, when I’m writing for nothing but my own pleasure, by the seat of my pants. But the record of my past mistakes forces me to note uncertainties carefully, and to remember that in the end, I can only make educated guesses about a world that’s under no obligation to live up to my expectations.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the author of this story:
Megan McArdle at email@example.com