My interpretation of the 2016 election is pretty clear, and stated repeatedly on this blog. James Comey's announcement of the "re-opened investigation" moved the polls enough to hand the White House to Trump, in an environment that naturally favored the GOP by the forecasting model I generally prefer: Alan Abramowitz's "Time for a Change" model, which predicts presidential elections using GDP growth in the second quarter of the election year, the incumbent president's popularity regardless of whether or not he is actually a candidate, and whether or not one party has already won two terms in a row (a third term is a rarity). That model favored the GOP. Trump was an historically weak candidate, tilting things towards the Democrats, but Comey tilted them back to where they "should" have been. And then got fired, with the initial excuse being that Trump didn't like how hard he was on Clinton.
And now, we have Clinton's book, indicating that she is still pissed at Sanders, and holding a grudge about the fact that Sanders could attack her, and she couldn't hit back because she knew she would win the primary, and would need the precious, little snowflakes who looked at Sanders as he-who-walks-on-bongwater. As it turns out, there's political science here!
Helmut Norpoth! There's a name that sounds like a Trump-supporter, right? Anywho, while my favorite presidential election forecasting model is the Abramowitz model, Norpoth has a competing model based on the contentiousness of the primary. Whoever has a more contentious primary should be at a disadvantage. And, in 2016, Norpoth's model predicted... Trump. And he published it in the October 2016 issue of PS: Political Science & Politics. Here's an ungated link.
Now, Helmut brags about the success of his model, but he does incorporate the "pendulum," which is the Abramowitz observation, and frankly, all of these models work pretty well, or they wouldn't keep getting published. They only really conflict in the very close elections, like 2016. (Or, wait, did you think it was some sort of "landslide?" Where would you have gotten that idea?) Also, Abramowitz's model called 2016 correctly too. Although, to be fair, good ole' Alan lost faith in his own model in 2016.
Regardless, if Hillary wants to hold a grudge against Bernie for helping Trump, she can cite Helmut Norpoth.
That said, as much as I fuckin' hate Bernie Sanders, his goo-goo* nonsense, his economic illiteracy, and plenty more, Helmut Norpoth's model has never stood up to the other predictive models, as far as I'm concerned. We have a limited data set, and multiple, competing models explain the limited data set in comparable ways, while I just don't see the "causal mechanism." You may have seen surveys about Sanders voters defecting to Trump, but did you also notice the surveys about Cruz/Rubio/other GOP candidates' voters defecting to Clinton? It works both ways, and in similar proportions. Sanders deluded himself for longer in the process because Democrats use different rules for allocating delegates, allowing Sanders and his drugged-out supporters to delude themselves for longer about whether or not the idiot socialist actually had a chance. (He never did. Not once. Ever.) That meant more focus on Sanders and his supporters than on the more diffuse non-Trump Republicans. And, that kind of thing is consistent from year-to-year.
As far as my critique of the Norpoth model goes, when it really matters is in a year with an incumbent on the ballot, where the incumbent faces no real competition most of the time, and there is a real contested primary on the other side. Combine that with the fact that incumbent presidents rarely lose and you have some illusory effects in the Norpoth model.
That said, 2016 was close. In a close election, anything can affect the outcome. In a close election, everything matters, in which case nothing matters because everything is random noise, except the big stuff. What was big? The Democrats had won two terms in a row. That was big. What else was big? Not Trump's hands, certainly. James Comey's announcement. That was big. How do we know? The polls moved. Big...ly. Sanders? Eh... He's an obnoxious, self-righteous fool who doesn't understand a single thing about politics or public policy, but I doubt he mattered much, in the scheme of things. If he had acted more like Ted Kennedy after he challenged Carter, and failed to endorse Clinton, maybe then one could make a case, but Sanders did eventually endorse and campaign for Clinton. So, I just don't see it, as much of a tool as Sanders is.
*Derisive term for one who thinks that "good-government" reforms, like campaign finance reform, will fix everything.