Donald Trump recently claimed that the Democrats could not “legitimately win” the 2020 election. As with Trump’s 2018 midterm election predictions, this comment seems more influenced by wishful thinking than any real grasp of reality. More than likely, the 2020 election will turn out to be a train wreck for Trump, and he will be the first president since George H.W. Bush in 1992 to lose his bid for re-election.

According to fivethirtyeight.com, which averages other polls, the percentage of Americans who disapprove of Trump’s performance as president is 53.2%, while the number who approve is 42.5%, putting him at a net -10.8% on approval. True, his numbers have improved a bit since the government shutdown ended, but they remain dismal. Approval numbers more than slightly below 50% are always worrisome for politicians and 42.5% is dismal. A poll by Politico/Morning Consult from early February showed that 52% of voters would definitely or almost definitely vote against Trump in 2020.

Of course, all of the above polling is on the national level. It is the Electoral College that decides elections in the United States, after all. Trump lost the popular vote in 2016, but still pulled out a victory. Shouldn’t he be able to do the same in 2020? Not necessarily. A Feb 6 Morning Consult map of polling in every state shows Trump above water in only 17 states, down from more than 40 in January 2017. States that now disapprove of Trump include not only the three usually-Democratic states that went for him in 2016 (Michigan, Pennsylvania and Wisconsin), but also Iowa, Arizona, Ohio, North Carolina, Florida, Georgia, and even ruby-red Utah and Texas. To be sure, Trump will not lose in Texas or Utah, whatever the polls may say, and he  does not need to win all of these states to be re-elected, but he must win the majority. Given that his disapprovals exceed his approvals by 10% in many of these states, this seems wildly unlikely.

Some people have argued that Trump must be the favorite because he is the incumbent. Yet an analysis of the 2018 election results by fivethirtyeight.com showed that incumbency was only about 3% in the 2018 election, and there is no obvious reason to think this will be different in 2020.

Of course, the election is a long way away. Trump could still turn things around, right? Certainly, but there’s no particular reason to think this is likely to happen. Unlike other presidents whose parties took a shellacking in midterm elections, Trump has made no effort to let the country know that he has gotten the message they were trying to send. Nor has he tried to reach out from his base, and most of his recent actions have been highly unpopular.According to a recent NPR/PBS/Marist Poll, only 36% of Americans approved Trump’s decision to declare a national emergency to build a border wall. Earlier polling, such as a Pew poll taken in mid-January showed that 62% of Americans either strongly or somewhat disapproved of Trump’s handling of the shutdown.

An event with even greater potential to impact Trump’s re-electability is the Robert Mueller investigation, which is expected to be released in the next week or so. If the report were to find evidence that Trump actually had committed a crime, the Democratic-controlled House of Representatives might well bring impeachment charges against the president. Though it is unlikely that the required 2/3 of the Republican-controlled Senate would go along with removing him, recent polls show that, if Mueller’s report were sufficiently damning, a majority of both Democrats and Independents would support the charges. This situation is very different from that of 1998, when the majority of voters believed that the GOP decision to impeach Bill Clinton was motivated mainly by politics.

By now, any Trump supporters will be screaming that almost all of the evidence presented here is based on polling, and that polls proved spectacularly wrong in 2016. However, the polls in 2016 were not nearly as far off as some people like to claim. They predicted a narrow Clinton win in the Electoral College based on a narrow Clinton advantage in the popular vote. The advantage in the popular vote predicted actually occurred, but failed to translate into an advantage in the Electoral College. That did not exactly make the polling wrong.Besides, even if it had been off, it does not prove that the polls will miss again, any more than a weather prediction that doesn’t come true disproves all weather predictions forever.

Perhaps the most convincing argument piece of evidence against a Trump repeat is the 2016 election. An almost perfect storm of events converged to benefit Trump, but he managed to lose the popular vote and nearly the Electoral Vote nonetheless. His opponent, Hillary Clinton, was both extremely unpopular and a mediocre campaigner. Her overconfident campaign took some key Midwestern states for granted and failed to dedicate the necessary resources to keep them in her camp. According to all U.S. intelligence agencies, Russia was doing its best to help Trump. Finally, FBI Director James Comey made the catastrophic mistake of reopening the investigation into the email scandal less than two weeks prior to election day. Yet Trump still managed to lose the popular vote by three million and barely managed to pull out the Electoral College. To an extent far greater than he or his supporters are willing to admit, Trump has Lady Luck to thank for his 2016 election. In 2020, she is not likely to be so kind.