/ About / Archive / Other media
The Trump confidence game analysis has no basis in fact - 1/8/16
Vox has published another piece predicting that Trump will lose. This latest variation on the theme, written by David Roberts, will likely be new to liberal readers - but he is really just rehearsing an analysis that has been floating around the American right for months. Here's The Federalist, predicting The Beginning of the End for Donald Trump:
Trump’s candidacy has been propelled by a kind of macho bluster, with Trump portraying himself as a winner who is amazing—terrific!—at everything he does, who can afford to laugh off all those other losers...But what if Trump stops winning? ...Trump’s campaign has so far been based on the old adage that nothing succeeds like success. But the flipside is that nothing fails like failure.
 That was back in September after the first Republican debates. Here's what's happened since then:

One can persuasively argue that Trump lost several of the early Republican debates, and that those losses indeed impacted his lead. This seems particularly clear in the case of the Carson surge in October, which saw not only a narrowing lead for Trump but an actual drop in his numbers.

But the rest, of course, is history. Even as the Federalist was declaring the end of Trump, the Matt Bruenig Election Team was predicting an end to the Carson surge - and that's exactly what happened. At his best, Carson could only cut Trump's 12 point lead down to 7 before he ran out of steam.

The incident is instructive because it illuminates the essential analytical problem all of the pundits are running into: Trump is much, much more popular than his opponents. If you really wanted to, you could try to nickle-and-dime his advantages away with a lot of small-bore, circumstantial data analysis - but 22 points are a lot to chip away at. So the temptation is to just develop an easy silver-bullet argument that could nullify his entire lead.

Unfortunately, it's hard to do that without inventing an argument so powerful that it could vaporize any lead. This is where analyses like Roberts' run into trouble:
Trump will lose something — maybe Iowa, maybe New Hampshire, maybe just a couple of news cycles...He'll enter a negative spiral as self-reinforcing as his rise has been.
Obviously, this sort of "negative spiral" argument could hypothetically cancel out anything from a 22 point lead to an 88 point lead. Roberts does nothing to quantify or at all constrain the impact of this potential effect, even though it seems ridiculous to suppose that it could actually topple any imaginable advantage in the polls. And this becomes a particularly difficult hurdle for Roberts because we've seen identical predictions before: we've seen Trump faced with bad news before, and it turns out that the "negative spiral" effect barely cost him five points.

The Matt Bruenig Election Team's proprietary models account for this sort of negative feedback effect, but we simply note that the larger Trump's lead gets, the less likely this sort of thing is to nullify it. And at this point, Trump's lead is so enormous that neither a few primary losses nor a few bad news cycles are at all likely to break it.