No non-incumbent since Ulysses S. Grant in 1868 has lost less than three points off the prior re-elected incumbent’s showing.
One way in which this dynamic plays out is the tendency of the party holding the White House to lose seats in Congress in the midterm elections. Another is the tendency of the governing coalition to lose support to the party out of power—a centrifugal force so consistent that parties have historically held the White House for more than eight years at a time only when they have assembled a coalition broad enough to sustain consistent losses for multiple elections in a row without ending up with fewer votes than the other party—i.e., below 50 percent of the two-party vote.
About that two-party vote: in looking at the share of the vote won by a presidential candidate, you have two choices. You can use a percentage of the total votes cast, or you can throw out the votes for third-party candidates and use a percentage of the votes that were cast for the two major party candidates—thus, the two-party vote. Both measures have uses for different purposes, but because every election in U.S. history has been won by a major-party candidate, very few third-party candidates even win a single state, and almost no third-party movements persist beyond a single election cycle, the two-party vote is the better measurement of the two parties’ relative odds of actually winning an election. For the analyses that follow, therefore, I use the two-party vote as the relevant measure, except where otherwise noted, although the charts in this and the next, related article also provide the overall popular-vote percentages for comparison purposes.