Because being depressed about the future is cool on the internet, many people are saying that Donald Trump will probably win reelection. In reality, we don’t know if Donald Trump will win reelection. The signs are bad for Trump. But predictions are hard.

“But incumbents always win.”

Since Clinton’s election, three incumbents have won, and none have lost. So one way of looking at this is that, in the modern era, incumbents always win.

Since Nixon’s impeachment, four incumbents have won, and three have lost. So another way of looking at this is that, in a slightly longer modern era, incumbents lose about 42 percent of the time.

But the best way of looking at it is to say is that a sample size of three or even seven is not very many. Also, some of those elections were either weird or long ago, so it’s hard to know how useful they are for predicting things.

“But many presidents have had their approval ratings as low as Trump and still won.”

Clinton’s approval was near Trump’s current approval at this point in 1993. By 1996, Clinton’s approval rating was 60 percent. If Trump’s approval in 2020 is 60 percent he will almost definitely be reelected.

Will Trump’s approval go up that much? At this point in 1993 the economy was bad, so it’s reasonable to suspect that people in 1993 were upset about the economy. Right now, the economy is good, so maybe today people just don’t like Trump. Maybe that means Trump’s approval won’t go up much. But maybe it will. Predictions are hard.

Clinton was unusual because his approval rating went up a lot over time. Usually, presidents’ ratings go down over time. Here are some other approval rating anecdotes:

• Jimmy Carter’s approval in 1980 was a little lower than Trump’s now and Carter lost.

• Gerald Ford’s approval in 1976 was higher than Trump’s now and Ford lost.

• The first Bush’s approval in 1992 was almost exactly where Trump is now and Bush lost.

• W. ’s approval in 2004 was about ten points higher than Trump now and W. won, but not by very much.

“But he was equally unpopular in 2016 and he won anyway.”

This is true. It’s hard to predict how many votes someone will get from their popularity alone, especially when you don’t know who the other person on the ballot will be.

One thing we can do is just poll presidential matchups directly. People did that a lot in 2016 and Trump was losing by a moderate amount. That gap got smaller right before the election, probably because of James Comey, which is why Trump was able to win (but not by very much).

People are polling presidential matchups right now between Trump and plausible contenders. In most of them, Trump is losing by much more now than in 2016. If the election was today he would probably lose by a large amount. This is a bad sign for him, but things also can change in 3 years.

Here are some other things that can happen in 3 years that would probably matter:

• The economy could get even better than it is now

• One or both houses of Congress could change hands

• Trump’s commission could purge millions of voters from the rolls

• Trump could negotiate bipartisan deals with Democrats

• Congress could start investigating Trump for corruption and other things

• More cabinet secretaries could quit in disgrace

• Jared Kushner could bring peace to the Middle East

• An asteroid could strike the earth

• Trump could solve the North Korea crisis

• Trump could quit or decide not to run for reelection

• Trump could start a nuclear war with North Korea

• Trump could start a nuclear war with France

• There could be a huge terrorist attack

• There could be a recession

• Robert Mueller could find out Trump was working for Russia

Some of these things are probably more likely than others, but I don’t know how likely any of them are because predictions are hard.