Much has been made about last Tuesday’s special election in Ohio’s sixth congressional district, where Republican state senator and local grocery chain owner Michael Rulli defeated Democrat Michael Kripchak, who had no prior political experience, by nine percent. Left without congressional representation when former Rep. Bill Johnson resigned to become president of Youngstown State University, the district is deeply conservative, where Donald Trump received approximately 64 percent of the vote in the 2020 presidential election.
The roughly 20-point overperformance by Kripchak follows a recent trend of Democrats running ahead of Biden’s 2020 performance in special elections. Dating back to mid-March, Democrats have overperformed in seven straight special elections, a handful of which have occurred in the critical battleground states of Michigan and Pennsylvania. However, this latest special election came as a shock to many due to the nature of the district: while Democrats have recently benefited from high-propensity voters, predominantly white, college-educated, and suburban, showing up to low turnout elections, Ohio’s 6th congressional district is a predominantly rural district in the Rust Belt, with just 22 percent of the population over 25 years old having earned a bachelor’s degree or higher.
While many believe this to be a positive sign for Democrats and Biden’s odds in the 2024 election, it is also the case that turnout in this election was exceptionally low—only 11 percent of registered voters cast a ballot in this election—such that ten thousand more Republicans cast a ballot for incoming Rep. Rulli in the Republican primary, held concurrently with the presidential and senatorial primaries on March 19, than in the special. With such low turnout, some argued that the Democratic overperformance, rather than signifying a new development in the election cycle, was merely a continuation of already-observed political trends. Ultimately, any discussion while or shortly after the first election results came in was unfruitful due to the lack of granular data to draw conclusions from.
However, as precinct results from the election have started to become available, we’re now able to better isolate college education as a factor in both turnout and support for Kripchak, the Democratic candidate: since the Census Bureau’s American Community Survey provides estimates for educational attainment by county subdivision, aggregating election results by township* allows for direct relationships to be made between election results and college education.
It is fairly apparent that there does seem to be a positive correlation between college attainment and turnout; while this would ordinarily suggest that Kripchak’s relative success in the special reflects the trend of Democratic candidates cleaning up among high-propensity voters, who disproportionately sway election results in low turnout environments, it is also the case that educational attainment appears to be somewhat negatively correlated with Kripchak’s vote share, suggesting that while college-educated voters did turn out to vote at a higher rate, they did not necessarily support the Democratic candidate. However, the data is also fairly noisy and appears somewhat random: running linear regressions weighted by the number of registered voters in each jurisdiction reveals small coefficients (the slopes of the best-fit lines are fairly flat) with weak R-squared values (college attainment is not a particularly strong predictor for turnout nor support for the Democratic candidate, in either direction), which suggests that there may be other, more influential variables affecting the dynamics of the race. Rulli may have experienced a significant boost in support in the relatively highly-educated Mahoning County, which he represented in Ohio’s state senate and was home to locations of his namesake grocery chain Rulli Bros Market. In contrast, Tuscarawas County, where Kripchak enjoyed the greatest leftward shift from Biden’s performance in 2020, is in a separate media market compared to the bulk of the district.
Perhaps a more definitive method to determine a relationship between education and turnout is with an ecological inference**, which can isolate the college-educated and non-college-education voter cohorts from the aggregated election data. Again weighing by the number of registered voters in each jurisdiction suggests that roughly one in twelve non-college-educated registered voters voted in this election, compared to over one in five college-educated registered voters, suggesting a fairly stark turnout differential—voters with a college degree were more than twice as likely to vote in this election than their counterparts without a degree. While it is difficult to compare this result to other special elections without having conducted this analysis for each election, this difference contrasts fairly significantly from general election turnout: roughly 60 percent of non-college-educated eligible Americans cast a ballot in the 2020 general election compared to roughly 78 percent of college-educated eligible Americans; while differential turnout remains, it is at a much more modest rate.
Ultimately, whether or not isolated special elections reveal much about the general election environment is irrespective of the fact that overperforming in these elections is significant in their own right: Democrats certainly would prefer to be competitive in these races and have the opportunity to flip otherwise-impractical seats. The results from individual special elections, of course, must be taken with a grain of salt, and constitute just one of the many ‘fundamentals’ in trying to predict the outcome of the general election; the several extraneous factors influencing this race only work to minimize the predictive power of this race. However, it does seem that there is some, albeit very limited, evidence that differential turnout between college- and non-college- educated voters does not fully explain Kripchak’s overperformance in this race.
*as of the writing of this piece, Columbiana County has not released precinct-level results for the election. Some counties do not release precinct-level results until a canvass has been conducted and results can be made official. Results from Columbiana were taken as-is instead
**this implementation of ecological inference is the PyEI python library created by Knudson et al.