Every presidential election year, voter turnout goes up. And every midterm, it goes down again.

This means that in presidential years (the ones divisible by four, such as 2016), large vote totals are needed to win just about any office. But in midterm years, such as 2018, much smaller vote totals suffice.

For example, Republican Ryan Zinke got less than 205,000 votes in Montana in the 2014 midterm, and won his re-election race in a double-digit landslide. But the candidate who got 205,000 votes in 2016 — Democrat Denise Juneau — was losing in a double-digit landslide when she did so.

Bearing that in mind, you can imagine why the same line of thinking arises ahead of every midterm election: Hey, if we could just get all those same people who voted for us in 2016 to come out and vote for us in 2018, even a losing performance in the presidential year would become a huge win in the midterm. This is an especially tempting line of thought for Democrats, who under normal circumstances tend to do a little bit worse in keeping up their party's midterm turnout than they do in presidential years. (Republican turnout also falls off considerably in midterms, only not quite as much.)

If you go back, you can see people raise the possibility of squaring this circle again and again and again. Just get those same people to vote for us every two years instead of every four — how hard can that be?

The problem is, it just never works out that way. It's not easy, and it's probably not even possible. And that could be a problem for Democrats, according to a new poll ahead of next year's race.

The latest Washington Post/ABC News poll shows the Democrats, as you might expect ahead of Trump's first midterm, leading on the two-party generic ballot. That is, when asked which party they plan to vote for in 2018, 53 percent of registered voter respondents who said they are "certain to vote" chose the Democrats, compared to just 40 percent who chose the Republicans.

But here's the thing: If you filter out those who were old enough to vote in the 2014 midterm but failed to do so — about half of all the adults polled — the Democrats' advantage almost vanishes. Within the smaller universe of proven midterm voters, the Democrats get 48 percent to the Republicans' 46 percent.

Might things be different this time? Could it be that President Trump has changed the equation, such that Democrats who never voted in a midterm before will come crawling out of the woodwork just because he's president? Maybe. Then again, consider this: According to the poll, 24 percent of the registered voters say they'll be casting votes to show support for Trump, compared to 27 percent who say their midterm vote will represent a show of opposition to him. That's only a 3 point net negative, with another 49 percent stating that Trump is irrelevant to how they will vote for U.S. House next November.

It's a good reminder that even though Democrats have the wind at their backs and Trump is unpopular going into 2018, it is a mistake to assume that they can win easily, and probably an even bigger mistake to assume that they can win by simply tying their opponents to Trump.

One further note about quixotic ideas about expanding the electorate: If you study the exit polls from 2006, the last time Democrats won a midterm, they didn't actually do it by expanding turnout at the margins, or significantly changing the election's demographics to something more favorable to them. In fact, combined Black and Latino turnout did not change significantly between 2006 and 2014, when Democrats were clobbered (18 versus 20 percent). Rather, Democrats won in 2006 because they got more votes from the groups that Republicans usually win. In 2006, Democrats won 47 percent of the white vote for U.S. House and 44 percent of the white male vote, instead of the 38 and 33 percent that their House candidates won within those two groups (respectively) in 2014, which happens to be exactly what they got in 2016 as well.