What's interesting is that these favorability ratings predicted the correct winner in each election and came pretty close to predicting the pop vote margin.
Since everything but 1992, 1996, 2000 Gallup Poll and the 2016 Monmouth poll was an actual exit poll, I would hope the correct winner in each election would be chosen as well as the popular vote margin.
Obviously, the polls in 1992, 1996 and 2000 were the final polls of the cycle, not August.
These are favorability ratings, not actual "who did you vote for" polls, though. I mentioned in a previous thread that favorability rankings are more predictive than actual polls right now - and as you can see above, the favorability ratings come pretty close to the actual pop vote margin.
In this cycle Hillary seems to run 8-12 points more favorable than Trump.