Flimsy analysis? How did he find out about Atlas?
Yeah, it's pretty funny that this thread has devolved into the exact sort of flimsy analysis that Silver is railing against. Just completely unsupported assertions that a handful of cherrypicked data points have strong predictive power.
Has no-one ever considered that just maybe Nate Silver has done a tiny bit more research than you into what has predictive power in elections and what doesn't? And if the "zomg Dems have won the PV in 5/6 elections" had any significance, past election results would showcase it? (Hint: they don't. You don't have to be Nate Silver to work this out, have a look yourself by looking at election data and see if you can find any significant relationship between past performance and future results).
Looking at past results to predict future one is about all of what Silver does.
Not sure why you don't think past results have predictive value. A growing majority of voters are very consistent in what party they vote for in presidential elections, more so than in decades past. Recent presidentials suggests more people are likely to vote Democratic than Republican and the GOP has a higher bar to clear as far as turnout and winning over the shrinking pool of undecideds.That would matter only if a history of success tends to be correlated with further success. But it isn't.
Because (and he and Sean Trende have gone into this in more depth in other articles), we don't have any good reason to believe that this advantage will continue. Historically, there hasn't been any consistent pattern, the "theoretical electoral college advantage" flips between the parties basically at random.
He's made his name at that, but the fundamental tools he uses are broadly the same. And much of that is recognising statistical relationships when they exist and discounting spurious or weakly supported evidence. And a ton of the stuff that passes for "analysis" here falls into the latter category.
Has no-one ever considered that just maybe Nate Silver has done a tiny bit more research than you into what has predictive power in elections and what doesn't? And if the "zomg Dems have won the PV in 5/6 elections" had any significance, past election results would showcase it? (Hint: they don't. You don't have to be Nate Silver to work this out, have a look yourself by looking at election data and see if you can find any significant relationship between past performance and future results).
Of course past election results can play a role in predicting future results unless you think the electorate makes its decisions based on a coin flip. As has been pointed out in this thread, we now have a deeply polarized electorate with very few true swing voters. So the fact that Dems have done better in recent presidential election cycles can serve as an indicator that they are heading into 2016 with an advantage as well, assuming the economy doesn't take a complete nosedive.
About past election results and their ability to predict future ones:
The correlation between President Obama’s margin in 2012 and his margin in 2008 across all 50 states and D.C. is .96. In other words, you can closely predict Obama’s margin in 2012 almost perfectly from his margin in 2008; his drop from 2008 to 2012 was fairly uniform, and limited the number of electoral votes he lost from 2008.
http://www.centerforpolitics.org/crystalball/articles/12-from-12-some-takeaways-from-a-wild-election/
I'm not disputing that state results tend to be highly correlated with past results. I'm saying that the winner of the whole election isn't correlated with who won the last handful (the exception being: incumbents have an advantage).
