In my quick and dirty summary of 2016, I made the observation that anyone who just went with Alan Abramowitz's "Time for a Change" predictive model did far better than anyone who followed the twists and turns of the presidential campaign or the polls. That isn't unique to 2016, and to push that point further, I've spent a bunch of time going through the shit I've said here. I got a hell of a lot right before the general election campaign got going. In other words, before I got distracted by the polls. Back when I relied on political science. Political science beat the polls.
In March, I declared the Republican primaries over. While many of my contemporaries were still in the throes of Trump denialism, I was pointing out his inevitable nomination, and describing the contours of a Clinton-Trump general election contest. Here are the main things I said.
1) In a March 17 post, I basically called it a toss-up. I'm not sure why I didn't mention that the Democrats had already won two contests in a row, which tilts things Republican, and informs the Abramowitz model, but whatever. I pointed out that the economy was in shaky territory, Trump's ideological incoherence could help him, slow mobilization of minorities could hurt Clinton (and we now suspect did), Trump supporters would be enthusiastic, leading to turnout effects, and money would be a wash, as it always is. Therefore, a toss-up race. Tied at the national level. When undistracted by polls, as I was back in March, I kind of nailed it.
2) In a March 21 post, just four days later, I wrote that polls might systematically underestimate Trump's support because of the "Bradley effect," in which respondents don't want to admit their support for a candidate. Trump's record of, let's admit it, racist and misogynistic behavior make him someone with whom you might not want to be associated, if you don't want people to think you are OK with racism and misogyny. As we now know, the polls were systematically wrong. Nailed it.
OK, to be fair, we don't know yet that this was a Bradley Effect. A lot of work is left to be done on what happened to the polls. There is also the possibility that the pollsters just got the "likely voter screen" wrong, and failed to figure out who was likely to show up on election day. The issue I have with that possibility is the consistency of the error rates. Some states had bigger error rates than others. Wisconsin!!! But, there was a lot of consistency between most states' error rates in the polling averages. There is work left to be done, but I suspect some Bradley effects somewhere in the data. Dissertations and books will undoubtedly be written on the topic.
3) In a March 25 post, a few days later, I got something very, very wrong. Why? I left political science and started bullshitting. I speculated that Trump, as an entertainer and businessperson, might display some chameleonic tendencies and change for the general election. Nope. He won anyway. As Trump would say, "WRONG!" Why was I wrong on this? It wasn't political science. It was just punditry. Do you see what happens, Larry?
There's plenty more, obviously, because I post something every morning. Why? This whole thing is basically a writing exercise for me. Thanks for indulging my narcissism. Regardless, that set of stuff in March jumped out at me as worthy of revisiting. Political science works. Bullshit doesn't. Oh, and yes, Alan Abramowitz is a badass, while Nate Silver just does guitar face. No, I'm never letting up on that.