The calculation itself is fine, but I don't think the interpretation is correct.
One issue is that you're treating the county results as coming from a random process when they don't. There are trends that underlie these counties' voting results, including the way voting behavior correlates with ethnicity and education levels. These counties tend to be whiter and have lower education levels than the nation as a whole. Groups like that used to vote in line with the whole country, but now they don't. See this article for details..
Another issue is that the set of bellwether counties itself is a result of selection effects. In 2016, there were 35 such counties (according to the article I linked). Nearly half of them (16) voted for Clinton and lost their status as bellwethers. So the 19 counties you're looking at are the ones that made it through a filtering process that only let through counties that like Trump. It shouldn't be that surprising that those counties liked Trump again four years later.
The calculation itself is fine, but I don't think the interpretation is correct.
One issue is that you're treating the county results as coming from a random process when they don't. There are trends that underlie these counties' voting results, including the way voting behavior correlates with ethnicity and education levels. These counties tend to be whiter and have lower education levels than the nation as a whole. Groups like that used to vote in line with the whole country, but now they don't. See this article for details..
Another issue is that the set of bellwether counties itself is a result of selection effects. In 2016, there were 35 such counties (according to the article I linked). Nearly half of them (16) voted for Clinton and lost their status as bellwethers. So the 19 counties you're looking at are the ones that made it through a filtering process that only let through counties that like Trump. It shouldn't be that surprising that those counties liked Trump again four years later.