In Oklahoma, for example, the results of the 2016 primary gave Sanders a 10% lead over Clinton, while in Minnesota, the gap was even larger at 23%. In his home state of Vermont, Sanders won by over 70%.
This election cycle, however, Sanders has only just scraped a majority of the votes in Vermont, while outright losing Oklahoma & Minnesota by 13% and 9% respectively. (source)
Can this disparity be explained using polling data or other analysis?
The simple answer is there was more competition.
In 2016, it was mostly Sanders vs Clinton for the Democratic Primary. She had several potential scandals brewing (her private email server, questions about the Clinton Global Initiative, etc) and Sanders was the only outlet for people who did not want to see Clinton win.
Sanders had plenty of healthy competition this time
Elizabeth Warren - She overlapped Sanders on a number of issues, and may have been a spoiler for him
Liberal candidates, such as Elizabeth Warren and Bernie Sanders, have embraced ambitious spending programmes to provide universal health care, increase access to higher education and deal with America’s worsening housing crisis.
Joe Biden - The former Obama VP, he has been the "please not Bernie" moderate option
If Sanders were to win the nomination, his proud connection to socialism and his 50-year record of extremism would be enough to doom his chances. Add to that his unpopular campaign proposals to double the size of the government, sharply raise middle-class taxes, and take private health insurance from 180 million Americans, and you hand the GOP a deadly arsenal of material to use against both Sanders and other Democrats on the ballot.
But the political winds may have shifted. On Saturday, roughly half of South Carolina Democrats voted for Biden, with Sanders almost 30 points behind. That matters, because it will dampen Sanders’ momentum heading into the massive Super Tuesday primary. And it might get voters in those states to do what the moderate candidates refuse to do: settle on one to carry the banner forward.
Michael Bloomberg - Probably more of a spoiler to Biden than Sanders, he, too, was an "anyone but Bernie" candidate who spent over $500M in Super Tuesday
Former New York City Mayor Mike Bloomberg's more than $550 million in campaign advertising and exclusive focus on Super Tuesday states were supposed to give him a big night Tuesday night, making him the default candidate for establishment Democrats to rally around to stop Bernie Sanders. Instead, he had just 44 delegates as of Wednesday morning, landing him in a very distant fourth place.
It likely has something to do with who he was running against in 2016. It's hard to make a "head-to-head" comparison when the opposition was different.
There are more candidates at this point in the race, so votes are going to be more split. Expecting % differences to remain constant doesn't make much sense when at least 4 candidates got non-negligible support in most places.
In Minnesota, an endorsement of one of Sanders' opponents by a popular senator and former presidential candidate Amy Klobuchar, as well as strong state support for another progressive candidate Elizabeth Warren, both likely hurt Sanders.
Additionally, primary elections are highly dependent on voter turnout. Voter turnout was low among younger voters, and excitement for Sanders might have been dampened in places like Vermont where he was expected to win easily - his supporters there may not have felt as compelled to vote there as in the previous cycle.
FiveThirtyEight has some thoughts overall on the trends on Super Tuesday including the direction that late deciders went.
Minnesota switched from a caucus in 2016 to a primary in 2020. Turnout increased from 200,000 votes to more than 700,000, so we can't say that Sanders lost his base of support from 2016 into 2020 (in fact, he got almost twice as many votes in 2020). One theory is that Sanders has a "more enthusiastic" base, and caucuses draw the most enthusiastic voters because of the time burden a caucus requires. Relieving this burden and increasing voter turnout is likely to blunt Sanders' advantage.
Also, 8% of Minnesotan ballots in 2020 went to candidates who had left the race as of election day, mainly because of mail-in voting. Another 13% went to the 3rd and 4th placing candidates, because in 2020 there were two more major candidates (with >1%) than in 2016. This complicates a comparison to the two-candidate race in 2016.
And as others have made clear, Sanders' main "moderate" opponent is different in 2020. Clinton's net favorability among Democrats in April 2016 was +36%; Biden's net favorability among Democrats in February of 2020 was +49%.
While it would take either mass mind-reading or polls asking the right questions (which I can't find) to produce a definitive answer, I would theorize that it is simple practicality on the part of voters. Their primary (no pun intended) objective is to win the November election. Going by current odds (e.g. https://www.actionnetwork.com/politics/2020-presidential-election-odds-donald-trump-bernie-sanders-joe-biden ) Sanders has less than a 10% chance of beating Trump, while Biden is nearly even.
Realistically, any candidate who hopes to win the general election has to attract votes from the ~40% or so of independent voters. Sanders describes himself (and is described by many others) as a socialist. Socialism is viewed unfavorably by independent voters, and especially so by older voters, who are more likely to actually vote: https://www.npr.org/2020/02/19/807047941/poll-sanders-rises-but-socialism-isnt-popular-with-most-americans
So even if a large fraction of Democrats who vote in the primaries actually like Sanders better than the other candidates, many dislike the prospect of four more years of Trump - particularly a Trump winning by a large margin - even more. So they vote for the candidate they view as most electable instead of the one they like best.
In 2016 this wasn't as great a factor, as few people expected that Trump would actually win the nomination.
