That was such a joke. American polling sites need to stay out of European elections. 538 also though the UK 2015 election would see the birth of a six party system. Lol.
But what's the alternative explanation for why all the polls agreed so much, if not herding? Polls agreeing with each other by much better than the margin of error says they should is highly suspicious, regardless of what methodology they're using. That's not some clever Nate Silver punditry. That's just basic math.
Well I am by no means a polling expert and I generally avoid the analysis associated with them, but from what I understand, French polling is much more accurate than American methods and even when they are wrong they are very much within a small margin of error; 2002 being the most obvious example which showed that error translate into a shocking upset.
You can't beat the margin of error. I mean, you can beat it once in a while, but not on a consistent basis. Every poll has a statistical margin of error that's derived from the number of people being sampled. On top of that, there's systematic error that stems from whether you're constructing your sample well or not. Now maybe French polls are really good at beating the systematic error down to nothing. But statistical error is impossible to beat. Again, this doesn't require any specialized local knowledge of France or French polling to say that. It's just stats. So if that's what Silver was saying (and that's what he seemed to be saying) then I think he's absolutely right.
