j reacts to the dangers of modelling with uniform swings while ignoring geography

this is exactly the kind of uniform swing distribution that i spent months yelling about last fall and in the end failed horribly. clinton's theoretical advantage in red states around the mississippi suggests that she should gain voters there much, much faster than she would in the western part of the red state block (rockies/prairies). so, a ten point lead nationally should give her a 15 point boost in mississippi (numbers are pulled from ass), enough to make it close, and almost no boost at all in utah. it's also ignoring what could be a significant third candidate.

i think it's too early to do this seriously, but i'll say the same thing i said about canada: you're better off trying to find a historical precedent than you are trying to find an algorithm to convert 2008 into 2016. and, one does exist. it wasn't even that long ago. and, clinton has a strong connection to him.

my best guess right now is to look at 1992 or 1996. perot got almost 20% in 1992 and less than 10% in 1996, but clinton got more electoral votes in 1996 despite winning less states. he won georgia in 1992 on the split, but won arizona in 1996 on his record. as we get closer to the election, johnson's numbers will tell us which one is a better model. but, i think that trump matches up better to bush than dole. yes, dole was seen as a doddering idiot. but, it was bush that was seen as the ideologically distant northeastern liberal.

that said, you'll have to make a couple of changes, too. it looks like she's way ahead in virginia, and she's not going to win west virginia.

but, insofar as the map starts to break? i think '92 is your road map.

http://fivethirtyeight.com/features/what-a-clinton-landslide-would-look-like/

it's a complicated debate about independent variables.

but, if you ask meteorologists, they'll tell you that you're better off finding a historical precedent than building your forecast on last season.