The United States has one of the lowest voter turnout rates in the developed world. In 2016, for instance, just under 56 percent of the voting age population cast a ballot in the presidential election, according to the Pew Research Center. In countries like Denmark, Sweden and Belgium, by contrast, participation in the most recent national elections surpassed 80 percent.

But the United States is a big country, and there are huge regional variations in voter participation within our own borders. At the state level, for example, turnout in the 2016 presidential election ranged from less than 40 percent in Hawaii to nearly 70 percent of the voting age population in Maine and Minnesota.

Drilling down to the county level, as we’ve done in the map above, shows even more variation. To calculate county-level voter turnout, we relied on two data sets: total votes cast in the 2016 presidential election, compiled from official sources by the MIT Election Data & Science Lab, and total citizen voting-age population as calculated by the Census Bureau. The final turnout figure is simply the percent of the citizen voting-age population in a given county who cast a presidential ballot.

The data shows that in the average county, 59 percent of adult citizens voted in 2016. Turnout rates ranged from less than 20 percent in one Georgia county to nearly 100 percent in a number of small southwestern ones.

What’s driving those differences? While there’s no magic formula that can predict voter turnout in a given community, there are a number of factors that researchers have found, time and time again, to be closely associated with who turns up at the polls. We’ve discussed a number of them below.

State policies matter

One of the most striking features of the map above is the effect of state borders on turnout rates. States like Tennessee and West Virginia, for example, appear in much darker shades of red than their neighbors. Colorado, on the other hand, is much bluer than surrounding states.

Some of these differences can be traced back to state-level laws on voting access. Tennessee, for instance, has a strict photo ID requirement and allows for vote by mail only under certain circumstances. Colorado, on the other hand, mails out a ballot to every voter in the state and requires only an ID — the non-photo kind — of voters who choose to cast a ballot in person.

Electoral competitiveness matters too

Some state-level differences can’t be easily traced back to voting regulations, however. In presidential elections, turnout tends to be higher in battleground states where voters may feel that their votes count for more than they would in a non-battleground state, where the result is essentially a foregone conclusion. You can see this most clearly at play in the map above in Vermont and New Hampshire. Turnout was higher in the latter state, despite its having much stricter voting laws than its neighbor to the west.

Demographic differences

Demographic variables, like race, age and education, also play a significant role in turnout. In 2016, for instance, turnout was nearly 30 percentage points higher among Americans age 60 and older than among the 18-29 age group. Whites and blacks were about 15 percentage points more likely to vote than Hispanics. Americans with a postgraduate education were more than twice as likely to cast a ballot as those who hadn’t finished high school.

Factors like those create the patchwork of turnout rates observed in the map above.

Don’t forget about culture

But demographics aren’t destiny. Culture and community-level norms regarding civic engagement and voting all play a significant role, too.

One way to indirectly measure the strength of these norms is via Census return rates, which researchers use as a proxy for social capital under the assumption that a person’s likelihood of responding to a Census questionnaire is “associated with prosocial, trusting, civic-oriented behavior,” as a 2015 study explained it.

At the county level, social capital (as measured via Census return rates) is about as closely correlated with 2016 voter turnout as income or education. This is particularly true in the upper Midwest, in states like Minnesota, which has both high Census response rates and high voter turnout.

What does it mean for 2018?

If researchers know one thing about turnout, it’s this: It plummets during midterm years. The plot of U.S. election turnout over time works out to a wild zigzag, with swings between presidential and midterm election years in the neighborhood of 20 percent. The differences are observable across all demographic groups, but they’re particularly pronounced for young voters.

How that will all shake out at the county level remains to be seen. County-level turnout is more complicated in midterm election years, since there’s no single national-level contest common between all voters. Some states have big Senate and House races drawing a lot of attention this year, while others don’t.

One thing that’s certain is that interest in the election is running high this year. Google searches for “voter registration,” for instance, are at a level that more resembles a typical presidential election cycle rather than a midterm year.