538 isn’t a poll. It is a model that uses polls as the input. I agree with the point. It’s more useful for congress than president, given the larger number of events. There is a lot of noise in a single prediction. Still a lot of noise in Congress but less
30/70 is not bad at all. Either outcome is not unlikely under those odds. However, a large majority of predictions I saw were more in the 10/90 range, and those are the ones whose accuracies and objectivity are questionable.
Well done ladies… I’m sure your analysis was rigorous and free of bias.
Lets get rid of discrimination, by discriminating!
- Libs
After Nate Silver of 538 earned minor fame from his pretty accurate modeling of the 2008 and 2012 elections, a whole load of new rivals popped up in 2016 and some of them like that HuffPost model were terrible and riddled with bias. There was one from a Princeton academic too that gave Hillary at a >99% chance of winning which was just objectively absurd. If anything, even though the 538 model had Hillary favored and hence was ‘wrong’, I think they came out with quite a lot of credit for saying openly and consistently that Trump had a realistic chance of winning.
As for this year, 538 says the Democrats will probably win the House and the Republicans will probably keep the Senate but that there is a significant probability that one party gains/retains control of both.
Best models are prediction markets because the data is compiled from people actually putting money at risk and using their best judgment as to the outcome of the election. It’s like saying whats the best valuation assigned to a NYSE stock some valuation model or the price it’s trading on an exchange. Also, polls are often wrong because they don’t ask who do you think is going to win, they ask who will you vote for in this point in time – no wisdom of the crowd. Lastly, most of the presidential polls in 16 were just popular vote polls (so basically meaningless in a tight race) and were actually right.
But as I have stated earlier… prediction markets also seemed to give Hillary a huge positive bias. In contrast, polls generally placed the outcome within the margin of error. So, people saw the poll data and largely ignored it, resulting in biased model adjustments and even biased markets.
Nate silver actually has discussed that he considered overriding the model because it was giving Trump too high a percentage
I don’t see how any sort of serious quant can be performed, based on polling the masses, with people living in fear of liberal/corporate totalitarianism. We all know everything we say is being recorded by corporate tech, and if we don’t think the liberal/corporate way, they will be coming for us later.
But take me. I’ve never voted, because it’s a con, and accomplishes nothing—yet I already placed my vote for all republicans in my massively-democrat state. Call it an “F-you authoritarians” vote. I’m not sure what that says, but am one of the most anti-voting people in existence.
I think if i was in a swing state I’d vote. Actually never mind. Life too lit to give a shit.
That’s pretty cynical. Not voting, not participating, and still calling it a con or whatever…is like punting on third down.