Most are much more bullish on Hillary. This is the NY TIMES model which also has links to several other models.
http://www.nytimes.com/interactive/2016/upshot/presidential-polls-forecast.html
First, these are all statistical models based on assumptions and poll data. So no one is doing anything. They developed models, and now it's just math. They input poll numbers, and then run simulations on it (usually 10,000) where they simulate the election. The percentages are the number of times each state went to a candidate in those simulations, with the percentage chance of winning overall being the number of times a candidate eclipsed 270 Electoral Votes in those simulations.
The reason that is important is these are assumptions made months ago. No one is intentionally doing anything from a modeling standpoint to get clicks. Now, the polls might be and these models rely on polls, so you could have the same effect, but it's not Nate Silver or anyone else on the statistical modeling side doing it.
Now, why is Nate Silver's model showing a closer race? He's explained this several times but he much more closely correlates the states, where other models do so less. What does that mean? Here's a dumb/overly simplistic example....
Let's say the election comes down to Hillary needing to win either Ohio or Pennsylvania, and your percentages of Hillary winning the states are 50% in Ohio and 75% in Pennsylvania. And she only needs to win one.
Now, if you don't correlate the states at all and treat voting in each state as independent events, the chances that Hillary loses both states (meaning Trump wins) is 12.5%, so her win percentage would be 87.5%.
But, if you say the states are correlated, and that Ohio is basically a redder, more Republican version of Pennsylvania, meaning that if Hillary loses Pennsylvania, there is no way she is going to win Ohio, than the Pennsylvania percentage is really the only relevant one, so if Trump wins Pennsylvania (25%) then he also wins Ohio (because the universe where Pennsylvania goes red and Ohio blue does not exist) and the presidency, so that percentage is 25% and Hillary's win percentage is 75%.
The real models are SIGNIFICANTLY more complex than that, but that example is a simply way of looking at how 2 models can look at identical data and come up with pretty different percentages.
It's also why the 538 model is "more aggressive." If you highly correlate states, a bad poll in one state makes the win percentage go down in other states, and vice versa. Meaning you will see percentages drop and rise much more rapidly. If Hillary starts doing better in Ohio polling, for example, then her Pennsylvania percentage goes up without even waiting for a Pennsylvania poll. Most of the other models also have some form of correlation, but a much weaker relationship so you just don't see the big moves, and because there is less uncertainty when you have less correlation, her win percentage is much higher.
Finally, the whole map thing is misleading. For example, that NY Times link I posted has all the models with their state percentage, but if it is under 65% win for both candidates, they color it yellow, which is more indicative of toss-up. 538 uses a map and colors it shades depending on strength. The problem with that is that when states hover near 50% (which he currently has NC, FL and NV at), then it is easy for it to change colors and look like a swing when nothing actually changed, just because visually, it's striking. For example, right now, FL is 50.4% Trump in that model, but if they re-ran the simulators, out of 10,000, Hillary might win slightly more than 50, and then it would change colors, even without a data change.
= new reply since forum marked as read
you'll see that this person constantly is attacking other board members for being trolls.
