"It will all come down to turnout" is an oft-ridiculed cliche in election circles. But there's some truth to it: In a highly polarized political environment like the one we're in, election results may be determined more by changes in turnout patterns than by changes in partisan preference. In other words, political parties not only need to convince voters to agree with them, but also to convince voters who already agree with them to vote.
In light of this, political strategists on both sides of the aisle had their eyes on the November 2023 elections as an indicator of who might vote in the 2024 presidential race. So, now that those elections are in the books, what can we take away from them?
538 analyzed turnout patterns in five states with high-profile elections in 2023, looking at both trends over time and differences between counties with different demographic compositions to see where turnout was highest. (You can download the data we used in this analysis on our GitHub page.) Overall, off-year turnout in these states was at or near its highest level in more than a decade, suggesting that we are still in the high-turnout environment that has characterized U.S. elections since 2016. We also found that turnout rates tended to be higher in suburban counties than in urban and rural ones, and that the pre-2016 conventional wisdom about off- versus on-year turnout has changed — which could pose both an opportunity and a danger to Democrats in 2024.
Since former President Donald Trump took office in 2017, turnout in U.S. elections has been off the charts. According to U.S. Census data, 53 percent of the citizen voting-age population voted in 2018, the highest midterm turnout since at least 1978. The 2020 election broke another record with 67 percent turnout, the highest since 1992. And midterm turnout in 2022 was almost as high as in 2018, at 52 percent.
Turnout rates in off-year elections during the same span have also been elevated. And if the 2023 elections are any indication, turnout in the 2024 presidential election will probably be very high as well. Compared to previous off-year elections with the same types of races on the ballot, Ohio, New Jersey and Pennsylvania all had their highest turnout since at least 2011, and Kentucky and Virginia came very close to matching their turnout acmes from 2019.
The state with the highest 2023 turnout was Ohio; its statewide turnout rate of 43 percent almost matched its 47 percent turnout rate in the 2022 midterms, when the ballot included a competitive open Senate race between Republican J.D. Vance and Democrat Tim Ryan (as well as a less competitive gubernatorial race). This was likely due to high voter engagement on Issue 1, a (successful) ballot measure to codify the right to abortion in the state constitution.
Contentious ballot questions have motivated similar off-year turnout levels in Ohio before: Turnout hit 42 percent in 2011 as voters chose to repeal an unpopular GOP-backed collective bargaining restriction. But Ohio's high turnout this year seems to be a strong argument in favor of the theory that abortion is still driving voters (especially liberal voters) to the polls, even more than a year removed from the Supreme Court's decision in Dobbs v. Jackson Women's Health Organization. Ballot questions addressing abortion access will likely be on the ballot again in several states in 2024, and while their effects on turnout may be less pronounced in a presidential year, they could still make a difference — especially when it comes to turning out young voters.
More consistent off-year elections in other states yield clearer trendlines. This year in Pennsylvania, where off years are generally headlined by statewide judicial races, turnout trended up for a fifth consecutive cycle. Thirty-one percent of the citizen voting-age population turned out, compared with 28 percent in 2021. Both years featured competitive, open state Supreme Court contests.
In Kentucky, turnout didn't quite match that of the 2019 gubernatorial election, when 43 percent of the citizen voting-age population went to the polls, and Democrat Andy Beshear upset unpopular incumbent Gov. Matt Bevin. This year, in Beshear's successful reelection campaign, statewide turnout was 39 percent. In line with national trends, this still represented a notable increase over previous off-year turnout, which was 29 percent in 2015 and 26 percent in 2011.
Finally, in Virginia and New Jersey, all seats in both state legislative chambers were up for grabs. Virginia's statewide turnout rate of 40 percent was 1 point short of 2019, but it was 14-15 points higher than the sleepier 2011 and 2013 elections. Meanwhile, New Jersey's 27 percent turnout rate was 4 points higher than in 2019, though still lower than the other four states we examined. This is not unusual for New Jersey; with the exception of the 2018 midterms, the state hasn't seen the same turnout increases as much of the country since 2016.
To get a better sense of what the electorate looked like in 2023, we also looked at turnout rates by county and cross-referenced them with county-level demographic data. In some cases, it appears that competitive or hot-button local races were the biggest drivers of turnout, but one consistent theme across states was that suburban counties saw significantly higher turnout than adjacent cities.
For example, in Ohio, turnout ranged from 36 to 46 percent in the four counties housing the state's largest city centers. But turnout was particularly high in populous exurbs, like Delaware County north of Columbus and Geauga County east of Cleveland, which saw turnout rates of 62 and 55 percent, respectively. The results suggest that Democratic mobilization around abortion was one factor driving turnout here: For example, Issue 1 was approved with 57 percent of the vote statewide, while President Joe Biden got just 45 percent in 2020 — a 12-percentage-point difference. But in Geauga County, Issue 1 got 55 percent of the vote, and Biden got 38 percent — a 17-point improvement.
Virginia's unique administrative divisions, which include 38 independent cities that are not part of larger counties, give us a particularly clear picture of the turnout disparities between cities and non-cities. The aggregate turnout rate in these cities in the 2023 election was 33 percent, while the aggregate turnout rate in the state's 95 counties was 43 percent. By comparison, in the last presidential election, the city turnout rate (63 percent) lagged behind the county turnout rate (74 percent) by around 10 points as well. Unsurprisingly, turnout was consistently high in the Northern Virginia exurbs, which also hosted a number of highly competitive legislative races.
One potential explanation for why turnout was generally lowest in urban areas is race. People of color tend to turn out at lower rates than whites, and cities tend to have larger nonwhite populations than suburban or rural areas. For example, Virginia's largest turnout decrease this year was in Danville City, a 50 percent Black locality that went from 33 percent turnout in 2019 to 14 percent turnout in 2023. Surrounding Pittsylvania County, which is 20 percent Black, dropped only two points, from 36 percent to 34 percent.
But at least in Virginia, population density, not race, seemed to be the most important variable. Aggregate turnout in Virginia's 12 independent cities that are at least 40 percent Black was 31 percent, well under the statewide turnout rate of 40 percent. But turnout in the five rural counties with a 40+ percent Black population was actually a bit higher than statewide: 42 percent. Precinct results suggest that Black turnout was high in suburban Virginia too; for example, in the new 84th House District outside Norfolk, precincts with large Black populations exceeded turnout expectations.
Next door, in Kentucky, suburbs once again led the turnout race. The four counties with the highest turnout rates were all suburban or exurban counties in the Louisville and Lexington metropolitan areas: Woodford, Oldham, Anderson and Franklin, which all saw turnout of more than 50 percent.
Unlike in some other states, though, cities weren't too far behind. Turnout was 45 percent in Louisville's Jefferson County and 44 percent in Lexington's Fayette County. The state's lowest turnout rates were largely in rural counties, particularly in the southern and eastern parts of the state. But lest we forget all politics is local, turnout was through the roof in one rural county on the state's southern border: Monroe County, a formerly dry county that voted to legalize alcohol sales in a 2,162-1,820 vote. Monroe was one of only two counties in the Bluegrass State that saw turnout increase over 2019.
In Pennsylvania, high turnout in Pittsburgh (Allegheny County) and the vote-rich Philadelphia suburbs (Bucks, Chester, Montgomery and Delaware counties) was key to Democrat Dan McCaffery's win in the state Supreme Court race. Democrats seemed to benefit from local races driving turnout in these blue-leaning areas, like competitive countywide elections for district attorney and county executive in Allegheny and contentious school board contests in Bucks and Montgomery.
On the other hand, Philadelphia County had one of the lowest turnout rates in the state (25 percent) for the second straight year. (In 2022, low turnout in this deep-blue county raised alarm bells for Democrats.) A big part of that appears to be low turnout among people of color. While Philadelphia's overall turnout rate in 2023 was higher than it was in 2019 (the last time a mayoral race was on the ballot), precinct-level results show that turnout dropped in majority Black and Hispanic neighborhoods compared to four years ago. These findings will do little to assuage Democrats' concerns about low minority turnout in swing states like Pennsylvania.
Although it is not expected to be competitive in 2024, New Jersey is another illustration of this challenge for Democrats: There, the lowest 2023 turnout rates belonged to Hudson and Essex counties (both 72 percent nonwhite).
Going into Election Day, a New Jersey-specific turnout trend that politicos were watching was whether Republicans would maintain or build on their enthusiasm advantage from 2021. That year, significant turnout gains in conservative areas — including more than 7 percent in Monmouth and Ocean counties, the state's most populous Republican-leaning counties — fueled some surprising Republican upsets, and there was even talk of Republicans flipping one or both of the state's legislative chambers in 2023.
But it was not to be. New Jersey Democrats maintained their 10-seat majority in the state Senate and actually gained six seats in the state Assembly. And when you look at changes in turnout by county, it's clear why. In 2021, the redder a county was, the more its turnout increased relative to four years prior. But in 2023, there was no such correlation.
Another interesting takeaway from this chart is the divergence of urban and mostly or somewhat urban localities. While this group of localities has always voted at higher rates than their major metropolitan counterparts, turnout in both clusters followed similar trajectories from 2010 through 2016. That trend dissipated during Trump's presidency, though, especially in 2019, when urban turnout dropped (as pre-Trump data would have expected it to) but mostly or somewhat urban turnout rose. The same thing happened in 2023, suggesting that whatever changed voting patterns here during Trump's presidency has not changed back.
Pennsylvania's graph is spikey before 2016 too: Rural turnout was consistently below statewide turnout in even years and above it in odd years. That hasn't been the case since, though. Every year since 2016, relative turnout levels in Pennsylvania's most rural counties have varied little, and always been lower than in the state overall, in part due to increased turnout rates in more urban areas.*****
From 2010 through 2018, turnout in Pennsylvania's three urban counties (Allegheny, Delaware and Philadelphia) was consistently higher than statewide turnout in even years and equal to or lower than statewide in odd years. But in 2019, these counties broke the trend and had their highest relative turnout since 2012. Then, in 2020, their relative turnout uncharacteristically declined — and it plummeted in 2021, while turnout rates in mostly urban counties (a cluster that includes the other three Philadelphia collar counties and smaller cities like Allentown, Harrisburg and Scranton) skyrocketed. Turnout in these areas has always outpaced the rest of the state, but it tends to crest in odd years.
What these new trends mean for turnout next year isn't immediately clear. In contrast to the Obama years, odd- and even-year turnout patterns haven't been totally predictable since 2016, so it's hard to say whether turnout in urban areas in 2024 will be as high as Democrats hope. But regardless of whether the new voting patterns in cities and rural areas are here to stay, it looks like suburban and mostly urban counties now reliably have the highest turnout — no matter what the calendar says.
Over the past several years, turnout slumps in major cities have been a cause for concern for Democrats. But at the same time, surging turnout in the suburbs — which have become bluer as they've grown more racially diverse and as college graduates have moved toward Democrats — seems to have helped offset this. Though suburbs are certainly not monolithic in their party preferences, even small leftward shifts in these areas have benefited Democrats — particularly as they've been accompanied by, or perhaps helped drive, high turnout. With this trend continuing in 2023, it seems safe to say suburbs will remain a major political battleground for years to come.
*Specifically, we looked at Pennsylvania and Virginia, because they both hold either statewide or state-legislative elections every year and had sizable and relatively even shares of their population in counties that can be categorized as urban, suburban or rural.
**We also clustered counties by other demographic groupings, like race and education, but we found that their relative turnout has been fairly consistent across off-year elections in the past decade. And when we looked at differences across all election years, we found that population density, not race or education, was most closely correlated with changes in turnout patterns.
***We calculated relative turnout among each cluster of counties by dividing their aggregate turnout rate by the statewide turnout rate. For example, if urban counties had a turnout rate of 60 percent in a year when the statewide turnout rate was 50 percent, we would calculate the relative turnout in those counties as 60 percent divided by 50 percent, or 1.2 — indicating 20 percent higher turnout than the state as a whole. This relative metric allowed us to directly compare trends in how county clusters voted across both odd- and even-year elections.
****Localities are classified based on the share of their population living in urban areas according to the 2020 decennial census. For Virginia, localities are considered urban if they are 97.5-100 percent urban; mostly or somewhat urban if they are 70-97.5 percent urban; and rural if they are 0-70 percent urban.
*****For Pennsylvania, counties are considered urban if they are 97.5-100 percent urban; mostly urban if they are 80-97.5 percent urban; somewhat urban if they are 65-80 percent urban; and rural if they are 0-65 percent urban.
This analysis is based on historical voter turnout in Kentucky, New Jersey, Ohio, Pennsylvania and Virginia. Data was obtained from each state's election website. In cases where states did not publish county-level turnout data, the vote in each county was estimated as the total votes cast in that county for the top-of-the-ticket ballot item. In cases where the top-of-the-ticket ballot item consisted of more than one seat (e.g., the three Supreme Court seats at the top of Pennsylvania's ballot in 2015), this number was divided by the number of seats available. Odd-year elections are analyzed only when the top-of-the-ticket item was either a statewide election or a statewide vote for state legislature.
Turnout in a given geography was calculated as the total votes cast in that geography divided by its citizen voting-age population as estimated in the five-year American Community Survey. We used CVAP estimates from either the 2018-22, 2013-17 or 2008-12 ACS — whichever included the year we were analyzing. For 2023, we used the 2018-22 ACS, as it is the most recently available dataset. We selected non-overlapping five-year estimates because five-year estimates provide more complete geographic coverage than one-year estimates and because only the five-year estimates include CVAP data.
For county-level demographics other than urban population, we used five-year ACS estimates (the year determined in the same method as above). For urban population, we used the decennial U.S. Census's estimated share of the county's population living in urban areas. To most closely align with the geographies used in the ACS, we used 2020 census data for turnout years 2020-23 and 2010 Census data for years 2010-19.
When creating clusters of localities by urban population, we considered the population distribution in each state to ensure that each cluster represented a similar proportion of the population in each state and accurately represented the state's demography. In Pennsylvania, each cluster of counties represents between 20 and 30 percent of the votes cast in the last four elections. In Virginia, each cluster of localities represents between 30 and 40 percent of the votes cast in the last four elections.