Could we see a streak of presidents who get reelected by a smaller margin than first time?

[President Johnson]

Barack Obama's reelection in 2012 remarkable in history, since it was the first time a president received a renewed mandate for a second term with a smaller share of both popular votes and electoral votes. You actually have to go back to the Founding Fathers for a similar pattern. Franklin Roosevelt's margins decreased in 1940 and 1944, but that was in his runs for a third and fourth term, respectively. Andrew Jackson in 1832 received a slightly smaller vote share than in 1828, but more electoral votes. Woodrow Wilson lost several electoral votes in 1916 (277 vs. 435 in 1912), which was due to a split in Republican votes in 1912. Other than that, all presidents who went on to win reelection expanded their electoral votes and popular vote shares. In all other cases, the presidents who saw their support decrease failed to win reelection at all.

I think it's fair to say in case Donald Trump gets lucky again, he wins 2020 with fewer than 306 electoral votes. And if Joe Biden wins and seeks a second term, he's probably favored for 2024, though I could see his margins decrease, like Obama. Especially if he wins over 350 electoral votes in November and beats Trump something like 53-45% in the popular vote. It would be a tall order to reach these numbers again unless the Republicans again screw up big time.

Up to 2050, could we see more presidents winning reelection by losing support compared to their first run? Or maybe even no president getting reelected by a wider margin? Somehow I think this could be the case.

[Del Tachi]

Ugh, observations like this are always kinda annoying because they combine a small sample size +  overfitting based on historical results to arrive at what mimics a predictive "pattern."

The fact that Obama was the first president since the Early Republic be re-elected with a reduced margin in the popular vote and electoral college is not especially notable.   There have been 18 elections in which incumbents were reelected (1792, 1804, 1812, 1820, 1832, 1864, 1872, 1900, 1916, 1936, 1940, 1944, 1956, 1972, 1984, 1996, 2004 and 2012.)  Of these 18, six saw the incumbent lose ground over his previous race in the electoral college or popular vote.  An event happening one-third of the time is not an anomaly.

You may make a stink about me including 1940 and 1944 here because they weren't proper "second" terms for FDR, but why should that matter?  You've arbitrarily attached significance to an incumbent's first reelection only because (in most cases) it's the only one they have.  FDR having 3rd or 4th reelections shouldn't except him from the historical "rule;" you're just redefining the rule to have a narrower application so that it better fits the data (i.e., overfitting) 

Finally, I think you're also engaging in what I'd call the "fallacy of binomial significance."  If we're talking about incumbents gaining or losing ground in their reelection, the distinction of whether or not they fell on either side of an arbitrary "270th" electoral vote isn't really that important.  You're attaching significance to an "on/off" threshold that (theoretically) could be set anywhere.  If you include losing incumbents who lost ground, then you can add 1800, 1828, 1840, 1888, 1892, 1912, 1932, 1980 and 1992 to the list.  That means that 16/27 reelections have seen the incumbent lose ground...which is anything but a historical anomaly.