Four years ago (more or less) -- Wishful Analytics

You would be hard-pressed to find a group of reporters and analysts who have more to answer for than data journalists on the political beat in 2016.

The dean of the group, Nate Silver, turned in an embarrassingly bad series of analyses during the primary, but to his considerable credit, acknowledged his errors and greatly upped his game during the general election. For this reason, he was perhaps the only major poll watcher to have reasonable forecasts on the eve of the election. The rest of the group, however, learned absolutely nothing from the nomination and continued making the same egregious errors through the general.

Since then, the people who screwed up have pushed the self-serving idea that, while there may have been mistakes, they were things that are only apparent in hindsight.

In cases like this, you simply cannot beat contemporary accounts.

Friday, August 21, 2015

Wishful Analytics

As mentioned previously, Donald Trump's campaign has definitely strained the standard assumptions of political reporting, Though this is an industry wide problem (even Five Thirty Eight hasn't been immune), it is nowhere more severe than at the New York Times.

The trouble is that the New York Times is very much committed to a style of political analysis that takes the standard narrative almost to the formal level of a well-made play. The objective is to get to the preassigned destination with as much craft and wit as possible. Nate Silver's problems at the NYT generally came from his habit of following the data to conclusions that made his editors and colleagues uncomfortable (by raising disturbing questions about the value of their work).

Cohn's articles on Trump have been an extended study in wishful analytics, starting with a desired conclusion then trying to dredge up some numbers to support it. He really, really, really, really, really wants to see Trump as another Herman Cain. Other than both being successful businessmen, the analogy is strained -- Cain was a little-known figure who surged well into the campaign because the base was looking for an alternative to an unacceptable presumptive nominee – but Cohn brings up the pizza magnate at every opportunity.

In addition to reassuring analogies, Cohn is also inclined to see comforting inflection points. Here's his response to the McCain dust-up.

The Trump Campaign’s Turning Point

Donald Trump’s surge in the polls has followed the classic pattern of a media-driven surge. Now it will most likely follow the classic pattern of a party-backed decline.


Mr. Trump’s candidacy probably reached an inflection point on Saturday after he essentially criticized John McCain for being captured during the Vietnam War. Republican campaigns and elites quickly moved to condemn his comments — a shift that will probably mark the moment when Trump’s candidacy went from boom to bust.

Paul Krugman (like Silver, another NYT writer frequently at odds with the paper's culture) dismantled this argument by immediately spotting the key flaw.

What I would argue is key to this situation — and, in particular, key to understanding how the conventional wisdom on Trump/McCain went so wrong — is the reality that a lot of people are, in effect, members of a delusional cult that is impervious to logic and evidence, and has lost touch with reality.

I am, of course, talking about pundits who prize themselves for their centrism.

...

On one side, they can’t admit the moderation of the Democrats, which is why you had the spectacle of demands that Obama change course and support his own policies.

On the other side, they have had to invent an imaginary GOP that bears little resemblance to the real thing. This means being continually surprised by the radicalism of the base. It also means a determination to see various Republicans as Serious, Honest Conservatives — SHCs? — whom the centrists know, just know, have to exist.

...

But the ur-SHC is John McCain, the Straight-Talking Maverick. Never mind that he is clearly eager to wage as many wars as possible, that he has long since abandoned his once-realistic positions on climate change and immigration, that he tried to put Sarah Palin a heartbeat from the presidency. McCain the myth is who they see, and keep putting on TV. And they imagined that everyone else must see him the same way, that Trump’s sneering at his war record would cause everyone to turn away in disgust.

But the Republican base isn’t eager to hear from SHCs; it has never put McCain on a pedestal; and people who like Donald Trump are not exactly likely to be scared off by his lack of decorum.

Cohn's initial reaction to his failed prediction was to argue that the polls weren't current enough to show that he was right. When that position became untenable, he shifted his focus to the next inflection point:

Mr. Rubio, the senator from Florida, has a good case to be considered the debate’s top performer. A weaker Mr. Bush probably benefits Mr. Rubio as much as anyone, and if Mr. Bush raised questions about whether he would be a great general election candidate, then Mr. Rubio added yet more reason to believe he could be a good one. Mr. Rubio still has the challenge of figuring out how to break through a strong field in a factional party.



...
Mr. Walker won by not losing. In a lot of ways, the moderators’ tough, specific questions played to Mr. Walker’s weakness. He didn’t have much time to emphasize his fight against unions in Wisconsin. But he handled several tough questions — on abortion; on relations with Arab nations; what he would do after terminating the Iran deal; race; and his employment record — without appearing flustered or making a mistake. His answers were concise and sharp.
...

Mr. Kasich also advanced his cause. He entered as a largely unknown candidate outside of Ohio, where he is governor. But he was backed by a supportive audience, he deftly handled tough questions, and he had a solid answer on a question about attending same-sex weddings. His answer might not resonate among many Republicans, but it will resonate in New Hampshire — the state where he needs to deny Mr. Bush a path to victory and vault to the top of the pack.



It was Donald Trump, though, who might have had the weakest performance. No, it may not be the end of his surge. But he consistently faced pointed questions, didn’t always have satisfactory answers, endured a fairly hostile crowd and probably won’t receive as much media attention coming out of the debate as he did in the weeks before it. If you take the view that he’s heavily dependent on media coverage, that’s an issue. Whatever coverage he does get may be fairly negative — probably focusing on his unwillingness to guarantee support for the Republican nominee.

You might want to reread that last paragraph a couple of times to get your head around just how wrong it turned out to be. Pay particular attention to the statements qualified with 'probably' both here and in the McCain piece. The confidence displayed had nothing to do with likelihood – all were comically off-base – and had everything to do with how badly those committed to the standard narrative wanted the statements to be true.


This attempt to prop up that narrative have become increasing strained and convoluted, as you can see from the most recent entry

Yet oddly, the breadth of [Trump's] appeal and his strength reduce his importance in shaping the outcome of the race.


If Mr. Trump were weaker, or if his support were more narrowly concentrated in either New Hampshire or Iowa, he would play a bigger role in shaping the outcome. In that scenario, a non-Trump candidate might win either Iowa or New Hampshire — and he or she would be in much better position than the second-place finisher in the state where Mr. Trump was victorious.



If Mr. Trump were to win both Iowa and New Hampshire, the second-place finishers would advance as if they were winners. Assuming that one or both of the second-place finishers were broadly acceptable, the party would try to coalesce behind one of the two ahead of the winner-take-all contests on March 15.



In the end, Mr. Trump almost certainly won’t win the Republican nomination; the rest of the party will consolidate around anyone else. He can influence the outcome only if his support costs another candidate more than others. But for now, he seems to be harming all candidates fairly equally.

First off, notice the odd way that Cohn discusses influence. If I asked if you would like to “play a bigger role in shaping the outcome” of something, you would naturally assume I meant would you like to have more of a say, but that's not at all how the concept of influence is used in the passage above. Cohn is simply saying that a world where Trump was behind in one of the first two primaries might have a different nominee but since Trump wouldn't get to pick who would beat him, it's not clear why he would care and since there's no telling who would win in Cohn's alternate reality, it's not clear why anyone else would care either.


But even if we accept Cohn's framing, we then run into another fatal flaw. Put in more precise terms, “harming all candidates fairly equally” means that each candidate's probability of becoming president would have been the same had Trump not entered the race. This is almost impossible on at least three levels:

Trump has already produced a serious shift in the discussion, bringing issues like immigration and Social Security/Medicare to the foreground while sucking away the oxygen from others. This is certain to help some candidates more than others;

For this and other reasons, the impact on the polls so far has been anything but symmetric;

And even if Trump's support were coming proportionally from each of the other contenders, that still wouldn't constitute equal harm. Primaries are complex beasts. We have to take into account convergence, feedback loops, liquidity, serial correlation, et cetera. The suggestion that you could remove the first two primaries from contention without major ramifications is laughably naive.


Finally there's that “only.” Even if Trump isn't the nominee (and I would certainly call him a long shot), he can still influence the process as either kingmaker or spoiler.


While Cohn's work on this topic has been terrible, what's important here is not the failings of one writer but the current culture of journalism. This is what happens when even the best publications in the country embrace conventional narratives and groupthink, adopt self-serving but silly conventions and let their standards slip.