Save for the 1960 election in Chicago or the 2016 election in the murky, conspiratorial sections of President Trump’s rhetoric, dead people don’t vote. It is the only demographic group that politicians truly don’t care about; after all, even people younger than 18 will likely end up voting someday.

Dying, however, is of immediate concern to political campaigns. Death tends to happen disproportionately to older Americans, for obvious reasons. About three-quarters of deaths in 2017 were among those 65 and older, according to Centers for Disease Control and Prevention data. And people who are 65 and older tend to 1) vote a lot and 2) vote Republican. Hence the concern: What happens to a party when its base dies off?

Look, the GOP isn’t going to literally die off anytime soon. But when you’re a political party that sneaked into the White House on the strength of 78,000 votes in three states — Michigan, Pennsylvania and Wisconsin — having a significant portion of your voter base in that 65-and-older group has to be somewhat worrisome. Younger voters are overwhelmingly Democratic but also tend to be lackadaisical about casting ballots. That foundation of older, conservative voters is essential to the GOP.

We wondered, then, how much death had reshaped the electorate since the 2016 election. Was it the case, for example, that the Grim Reaper is conducting the Final Get-Out-The-Vote in red states more than blue ones? What does that mean for 2020?

Thanks to data from L2, a nonpartisan voter data firm, we can report that in nearly every state, the number of new voters added since the 2016 election is larger than the number of voters removed from the voting rolls because of being dead. But the variation in the relationship between those two groups of voters (living, dead) is itself not even — nor is the partisan composition of the new-voter pool.

Contrast West Virginia with Texas on the map below, for example. There’s a lot of data there, so just focus on two circles: The black one (showing deaths since the 2016 election relative to the total current pool of voters in the state) and the multicolored pie chart (showing new registrants by party, also relative to the total current voter pool).

In West Virginia, the number of deaths is slightly higher than the number of new voters; the black circle is larger than the pie chart. In Texas, there are a lot more new voters since the 2016 election — and they are registering as Democrats (blue) more than Republicans (red) or independents/third party (gray).

Both of those states voted for Trump in 2016, as you probably know. The data above reinforces how they otherwise differ over the long term. West Virginia is getting redder and older, while Texas gets more blue with a booming population.

Red states, in fact, have seen a lot more deaths of registered voters than blue ones since the 2016 election. Blue states have lost a little over 2 million voters since the 2016 election, compared with 3.9 million in red states. (For the purpose of this article, we’re considering D.C. to be a state, though it still will not receive any senators.) Of course, there are more red states than blue ones. On average, red states saw 30,000 more voters cross the River Styx than blue states.

The category of 2016 states that had the heaviest average loss? Those three states that flipped from blue to red. They lost an average of 207,000 voters, slightly more than the other states Trump won more narrowly. Those states (which exclude the three that flipped) lost an average of 197,000.

But again, most states also saw a surge of new voters. Here, the gap was narrower, with blue states adding about 8.8 million voters (419,000 on average) and red states adding 11.4 million (380,000 on average).

Notice the two categories that added the most average new voters: states that voted heavily for Democrat Hillary Clinton and states that voted more narrowly for Trump. Dark-blue states adding more voters is a central reason the electoral vote favors Republicans: Those votes are added to states without shifting the number of electoral votes they’re worth (for now). And that’s the central question here, right? Shifts in votes?

That brings us to those states that Trump won more narrowly. It’s the category that saw the biggest net change when comparing deaths to new voters, with a net gain of 570,000 voters on average since the 2016 election.

It’s these states and the states that flipped that Democrats need to eat into if they’re going to win in 2020. The states that flipped lost more older voters but still added voters on net. The lighter-red states added a lot of voters, relative to other states.

There’s a very small correlation between the 2016 margin in a state and how its voter pool changed since that election. The states that saw the biggest net increases tended to vote more heavily Democratic (as the graph above also shows). Down in the bottom-right corner, below: West Virginia — dark red but with as many deaths as new voters, according to our data.

What matters for determining how the states are shifting is how those new voters plan to vote. Happily we can get some sense of that, by looking at their stated political parties or, in states without declared party affiliations, how L2 figures they are likely to vote. (Did they request a Democratic primary ballot in 2018, for example?)

In most cases, there are more independents than members of either party that were added to the voter pools in these states. But in the seven groups we’ve been looking at, six saw more new Democrats than new Republicans in the aggregate. The only group that saw more Republicans was the group of the most heavily pro-Trump states.

You can compare that with how those states voted in 2016, as below.

This dynamic also makes sense. A lot of new registrants are people who haven’t registered before: teenagers and young people just getting into the habit. And that’s a group that, again, skews more heavily Democratic.

It’s the flip side of the Republican concern about losing older voters. A lot of the new voters they’re getting aren’t Republicans.

Of course, all of those independents muddy the water a bit. Independents tend to vote heavily along consistent partisan lines, but we can’t see which party just from voter registration. So let’s consider each state on two dimensions: the gap between Democratic and Republican registration and the density of independent or third-party registrants in the state.

We get this.

It’s shaped something like a triangle, because the less dense the partisan registration in a state, the smaller the margin between the parties can be. But let’s consider three states in particular on this metric.

First, there’s Illinois. The L2 data suggests that most of the new additions are independent or third-party, so the state is up there at the top of our chart. But The Washington Post is predicting that Illinois will [checks notes] stay blue in 2020. With some states, we feel pretty comfortable in making assumptions about the voter pools.

Then there’s Texas. We noted before that there were more new Democrats there than Republicans in our data. We’ll note here that we’re still 14 months from the general election and these numbers will shift as more people register. That said, Texas is down in the lower left part of this graph, meaning more identifiably partisan registration and more of that registration made up of Democrats. More so than California!

Interesting — but not as interesting as Michigan, where the new registrations are heavily Democratic. That’s in keeping with the average for those three states that turned red in 2016. They may have been won by Trump narrowly in 2016, but, since then, they’re registering a lot more Democrats.

Since 2016, those three states have lost 621,000 voters to death, may they rest in peace. They’ve added 1.4 million voters. L2′s data tells us that 752,000 of those new voters are Democrats and 247,000 are Republican — a difference of 505,000.

These are the states that elected Trump by a combined 78,000 votes.