With some time to reflect on the problem, I also wonder if there’s been too much #datajournalist self-flagellation. Trump is one of the most astonishing stories in American political history. If you really expected the Republican front-runner to be bragging about the size of his anatomy in a debate, or to be spending his first week as the presumptive nominee feuding with the Republican speaker of the House and embroiled in a controversy over a tweet about a taco salad, then more power to you.The post is entitled "How I Acted Like A Pundit And Screwed Up On Donald Trump". That's a cop-out. He explains it thus:
Unlike virtually every other forecast we publish at FiveThirtyEight — including the primary and caucus projections I just mentioned — our early estimates of Trump’s chances weren’t based on a statistical model. Instead, they were what we “subjective odds” — which is to say, educated guesses. In other words, we were basically acting like pundits, but attaching numbers to our estimates.The numbers he attached were things like the quantity of political endorsements received; comparing Trump's poll numbers with those of other exciting early leaders who subsequently flamed out like Herman Cain, Newt Gingrich, and Rudy Giuliani; favorability ratings (Trump's were low while primary success putatively required them to be high), etc. In other words, quantitative stuff.
He says this wasn't formulated into any specific model like his polling predictions--which only differs from what Real Clear Politics does by assigning different weights to different polls instead of RCP's binary yes/no on whether or not to include a poll--the results of Silver's forecasts and RCP's are, not surprisingly, virtually identical. That's likely technically true, but based on the fact that all of these quantitative indicators suggested Trump wouldn't get the nomination, amalgamating them into a formula would have clearly yielded the following prediction: Trump will not win the Republican nomination.
Trump is going to win that nomination, of course. Silver didn't get it wrong because he wasn't meticulous enough with the way he balanced his numbers, he was wrong because he was looking at the wrong numbers, at least this time. Political forecasting is more marketing research than it is Science!--the need to adjust inputs and calibrate assumptions accordingly is a prerequisite to having a shot at getting it right every time rather than only getting it right when everything plays out exactly like it did the time before.
Because I bought in early on Trump, maintained Sanders never had a shot in hell because blacks weren't going to vote for a carpetbagging wonkish Jew from lily white Vermont, and wear a pseudonym that requires it, some modest suggestions for what Silver should have looked at:
- Immigration. Polls appear to show wide variation in public sentiment on the issue according to how the questions are presented, but the most objectively-worded polls have shown for decades now that immigration restriction and deportation are majority positions at the national level, and are overwhelmingly so among Republicans.
Trump made this his signature issue. With the exception of Tom Tancredo, who was at the time an unknown congressman from Colorado who lacked charisma, stature, and salience, no other presidential candidate since Pat Buchanan has paid it any heed.
Humorously, way back in September of last year, Silver contrasted Sanders' putative substance to Trump's alleged lack of it:
Sanders is campaigning on substantive policy positions, and Trump is largely campaigning on the force of his personality. I'm not sure this assertion requires a lot of proof, but if you need some, check out the candidates' websites. Sanders's lists dozens of specific policy proposals across a wide range of issues; Trump's details his position on just one, immigration.Indeed.
- Relatedly, I'll echo Steve Sailer and note that the success of "far right" European political parties, for which Trump is a closer American representative of than any of the other GOP presidential candidates are, should've tipped Silver off to the fact that immigration (and related issues like internationalism) was going to be a driving factor this time around.
- Nowhere in Silver's 5,000+ word post did he mention social media. Trump dominated the rest of the field, Republican and Democrat, when it came to generating interest on social media. Back in December, I quantified it graphically:
Months before any actual voting had taken place, Trump and Clinton had garnered the most followers. Several months of voting having now occurred, it is clear that Trump and Clinton are going to win their parties' respective nominations.
To emphasize how crucial this has become, consider that Trump is instantaneously able to communicate directly to more people than the audiences at any given time of Fox News, CNN, and MSNBC combined. I hear about every news item Trump's involved in from Trump before I hear about it anywhere else. By leveraging social media, he is able to assert more control over the narrative than the major media do.
- Nor did Silver mention rally/speech turnouts. When one guy is struggling to fill a high school gymnasium and the other guy is packing sports stadiums, it's probably time to reconsider favoring the former over the latter. If, on two-day notice, they'll skip work to travel a couple hours across state to wait a few more hours in line for a shot* at seeing a candidate speak, there's a reasonable chance they're going to drive a couple blocks to the local church or middle school on primary day and spend a couple of minutes voting for that candidate.
* Many people who've shown up for Trump rallies have been turned away because the massive venues were full; my wife and I got to the rally in KC a couple of hours before it started and would not have been able to get in had a friend not arrived two hours before us and saved us a spot in line.