Brian Keegan: Odd election years are temporal gerrymandering
By Brian Keegan, published in the Daily Camera:
“Uninformed.” “Evading scrutiny.” “Power-hungry.” “Carpetbaggers.” “Chutzpah.” This is how proposals to switch the City of Boulder’s elections to an even-year cycle have been described. As a data scientist and assistant professor at the University of Colorado, I was curious who stands to lose from shifting away from odd-year elections given this venomous rhetoric. Using open data from the offices of the county clerk, assessor’s office and the state demography office, I analyzed the makeup and behavior of voters in each of the City of Boulder’s 88 precincts for every election between 2000 and 2021. I found two powerful patterns: the precincts with the highest turnout in odd election years are richer and older than precincts with lower turnout.
The citizens of Boulder should be proud that turnout (total votes divided by registered voters) is much higher than the rest of Colorado and the country: 89.1% in the 2020 election (an “even” year). But in the 2021 election (an “odd” year) only 47.9% of eligible voters cast ballots. One cannot ignore this significant drop-off in turnout between even and odd years and its implications for the legitimacy of decisions being made.
I naively thought that this turnout drop-off between even and odd election years would be relatively consistent across precincts. It is not. The “851” precinct northwest of Tantra Park only has the lowest average drop-off in turnout (18%) from even to odd years: these citizens vote relatively consistently across odd and even years. In contrast, the “834” precinct encompassing CU’s East Campus has the highest average drop-off in turnout (62%): These citizens vote much less in odd years than even years. The average drop-off across precincts was 34.9%, which is similar to the 35% drop-off for precinct “831” encompassing the Hill neighborhood west of CU’s main campus. It’s not just college students ignoring odd-year elections.
Are there patterns behind this variance in even-to-odd year turnout drop-off? Using data from the county assessor’s office, I analyzed the median assessed value for residential properties in each precinct and found a strong negative correlation of -0.43: Precincts with more valuable homes have smaller drop-offs in turnout between even and odd years than precincts with less valuable homes. The county clerk’s public voter file records each registrant’s year of birth. An even stronger negative pattern (r=-0.84) plays out here: Precincts with older voters have smaller drop-offs in turnout between even and odd elections. Wealthy and older voters vote in odd-year elections when few others are voting.
Using only these two variables (age and home value), a simple linear regression model can explain 77.8% of the variance in the turnout drop-off across Boulder’s 88 precincts. Because age has the most unambiguously negative effect among these three variables, I explored Colorado’s Demography Office forecasts for Boulder County (no municipality estimates are available). The over-65 demographic only made up 10% of the population in 2010 but is estimated to overtake the declining 0-17 population in 2023 and could make up 21% of the population in 2030. The rapid greying of Boulder and the power that odd-year elections give to these voters is a relationship that cannot be ignored.
The power of property values and age in explaining why some precincts vote invites us to reflect on the biases baked into the decisions made by odd-year electorates. Boulder citizens are quick to speak out against efforts to limit access to the ballot elsewhere, but the vitriol directed at moving towards even-year elections reveals a beam in our own eyes: Odd-year elections are unmistakable instances of politicians and policies choosing their voters instead of the other way around. In light of these biases, opponents of shifting to even-year elections should be more honest in their arguments: Their issues can only win when fewer people vote. This temporal gerrymandering should be stopped, and democratic decision-making should be returned to the larger and more representative electorates during even-year elections.
The data and code for replicating these analyses and additional findings are available here.
Brian C. Keegan, Ph.D., is an assistant professor of information science at the University of Colorado Boulder.