Last Friday (7/22), Trump and Hillary both posted shareable graphics for facebook users to showcase who they're voting for in November. As of Monday (7/29) at 9pm, this is what they showed:
Trump's photo has been liked and shared by over 50% more people than Hillary's has been even though Hillary's specifically asks users to share the photo while Trump's does not.
This doesn't show up in polls* or polls-plus forecasts, though, so Nate Silver should just keep ignoring indicators like these. The endorsements of the political class, a group despised by the electorate, is a better measure!
Speaking of Silver, how dopey does he look for pairing the launch of his 2016 general election model with the headline "Trump Has A 20 Percent Chance Of Becoming President"? The release date was arbitrary. If he'd put it out several weeks earlier, when Trump was marginally ahead in the RCP average--as he is again today--he wouldn't have been able to dress up yet another punditry prediction of Trump's impending downfall in quantitative clothing.
As of today, Silver's model shows Hillary with a 54% chance of winning to Trump's 46%. Oops.
* Most poll samples are created by mirroring demographic population profiles or by using turnout from previous election cycles to estimate turnout for the election in question, and most of these are conducted on registered voters. But "likely voters" is a stronger predictor of the actual behavioral intentions of potential voters. These "likely voters" polls have consistently shown Trump doing several points better than "registered voters" polls do.