All this guy says is that it was a 50/50 probability, and then mocks everyone else who might say otherwise, which is basically everyone else except this guy. A 50/50 probability is a meaningless non-statement, and not what Silver, nor any campaign, nor any polysci academic creating a model for elections would find to be a reasonable, nor helpful, predictive statement.
A poll aggregation is a method based upon the accumulation of direct evidence of voter intent on a state-by-state and national basis. When there is a candidate as disruptive to the system as Trump, one can make a valid argument that poll aggregates are going to provide a better picture of reality than the generic models used in academia that rely upon traditional candidate templates. Per your link:
Economic effects
never achieved statistical significance in open-seat contest, and Presidential approval is weaker, and far less influential indicator. Huh, those are exactly the things you said a good method
requires to predict the 2016 election! How about that! And of the 2016 models, only 3 out of the 10 were more accurate than 538 concerning total vote percentages, which had a median error of .7 according to James Campbell (
https://www.cambridge.org/core/jour...on-forecasts/47D0EEDD5030B5F152AEB9B92A94DCE1).
And what's more, the vast majority of these models only predict popular vote totals, which as we can was not a useful metric for this particular election. All but two of them had Hillary as the favorite due to those popular vote totals. To be snarky, 2/10 models in favor of Trump puts his odds even lower than 538 had them ;D Let's get back to your original claim:
You're implying the modelling and analysis was trying to give an "impression" in some nefarious way. Define impression. A motivated impression? Motivated by what? Polling averages, based on past accuracy, period. They weren't saying "Trump is going to fail in his bid for the White House." They said "according to a reasonable interpretation of
polls, Trump has
less chance than Hillary to win." This is you inserting words into their mouths to create a strawman. You can say people shouldn't put their trust in polls to determine the Presidency to begin with, but that is neither here nor there as 538 is clearly appealing to people who are interested in polls and value the aggregation of polls into broader claims. And aside from this, the OP was about a single approval rating poll, and 538 is clearly a better option in determining the current approval rating of the President.
You also have literally no evidence to show 538 "eroded people's trust in poll analysis." A near 1 in 3 chance is
not a determination of loss, which they had clearly stated multiple times. I went into election night thinking Trump had a decent chance - not great - of winning. I knew which states were weakest and most likely to flip for him. How misleading, eroding my trust! Jesus, dude.
Your claim that they "misused statistics" is also conveniently vague. Exactly how did they do this, when they made it clear they were a
poll aggregate based strictly upon
polls and
poll accuracy?
