People vote at a polling place set up at the Kenter Canyon Elementary School in Los Angeles. (AP Photo/Nick Ut)

In every election, there are people who go to the polls to cast a ballot but who don't vote in every race. Usually, those "undervotes," as they're called, happen down-ballot, resulting in fewer votes for, say, county commissioner than, say, president of the United States. But in every election there are also people who skip the presidential ballot for whatever reason. It happens.

With 2016 pitting two historically unpopular candidates against one another, we were curious about the extent to which undervoting occurred in the marquee contest this year. We found some evidence that it was occurring a few weeks ago, but set out to tally the undervote more deliberately, pulling in data from every state to figure out how many people skipped the top of the ticket.

We were able to compile data from 33 states and D.C. In those states in 2012, there were 754,000 undervotes at the top of the ticket — about 0.9 percent of all ballots cast. In 2016? At least 1.75 million people skipped the presidential contest, 2 percent of the total. In other words, in these states, one out of every 50 people declined to vote in the presidential contest.

That undervote varies by state.

In only three states was the undervote percentage down. In states where it was up, it was up by an average of 2.5 times as much as in 2012.

The way this is tabulated is fairly simple. In states that compile the total number of votes cast, we compare that number to the number of votes in the presidential race. The difference is the undervote. We checked every state for this data, but some haven't yet posted overall vote totals, and others use the balloting in the presidential race as their total statewide count for the purposes of estimating turnout. (Which, of course, means that the "overall total" is the same as the presidential total.)

No state had more undervotes than California — nearly 470,000 — which makes sense because it's the most populous state. But California also had one of the higher undervote percentages. The highest was in Montana, though that only added 20,000 undervotes to the total.

In several states, the number of people who didn't vote was near or greater than the eventual margin of victory.

  • Arizona. Margin, 91,234. Undervote, 88,332 -- 96.8 percent of margin.
  • Maine. Margin, 22,142. Undervote, 23,965 -- 108.2 percent of margin.
  • Florida. Margin, 112,911. Undervote, 160,450 -- 142.1 percent of margin.
  • Michigan. Margin, 10,704. Undervote, 75,335 -- 703.8 percent of margin.

Notice that all of those states are ones that would count as "swings." Had one-seventh of those Michigan voters who decided not to vote in the presidential race cast a ballot for Hillary Clinton, she would have won the state.

The extent to which this effect can be blamed on the candidates is hard to say with certainty. It's clearly the case that a greater density of voters in 2016 were willing to pass on the presidential race than were in 2012, but any number of factors could be at play beyond just the people on the ballot. There's no strong correlation between turnout changes since 2012 and the undervote; if anything, turnout was more likely to be up in places with a greater percentage of blank presidential ballots.

It is clear, though, that people deciding on one candidate or another instead of skipping the presidential race could have made the difference in the election. Flip Michigan and flip Florida — both states where the undervote was bigger than the margin — and Donald Trump drops to 261 electoral votes.

But between those two states, more than a quarter of a million people preferred to pick nobody than anyone on the ballot.