Earlier today, I presented data on the share of people who have voted early for each Maryland county. At the end of the fourth day of early voting, 7.1% of registered Marylanders had voted. The share across counties varies much more widely from 2.8% in Allegany to 15.5% in Talbot.
What explains the variation?
St. Mary’s Political Science Professor Todd Eberly suggested on Twitter that he thought there were signs of higher turnout in counties that (1) have competitive executive races, and (2) are in the First Congressional District. Is he right?
Two other hypotheses occurred to me: (3) turnout might be higher in counties with more older residents as age is associated with participation, and (4) race might also be associated with voting. In particular, Latinos and Asian Americans tend to vote at lower rates than whites or African Americans.
I created multivariate models of turnout at the end of the fourth day of early voting in order to test these hypotheses. The models allow one to assess the probable impact of one factor while controlling for others. However, the relatively small number of units in Maryland (24) precludes including too many variables.
So where are more people voting? Increasing the share of the population over age 50 by 2% raises the predicted share of early voters in a county by an estimated 0.59% (p = .02 for the statistically minded who are familiar with OLS models).
Counties in the First CD also seem to be participating at higher rates, though the effect is small. In counties in which 100% of registered voters live in CD 1, turnout is an estimated 0.02% higher than elsewhere in the state (p = .04). Though perceived as a safe Republican seat, voters there seem energized this year.
Counties with hot executive races (defined here as Anne Arundel, Baltimore County, Frederick and Howard) are seeing much higher early voting turnout. The estimated impact of close exec contest on the share of early voters at the end of day 4 is 2.75% (p = .04). Including Montgomery as a hot race causes the effect to vanish, suggesting that our unusual three-way race is not motivating EV turnout in the same way in MoCo.
Contrary to my expectation due to historically lower rates of Latino and Asian turnout, counties with more non-Hispanic whites seem to have fewer early voters. Increasing their share of the population by 2% is estimated to reduce EV turnout at the end of day 4 by 0.10%. However, the effect is marginal as statistical significance is normally assessed (p = .09) and I would interpret with especial caution.
Closer examination of the data reveals that any impact is driven by the presence of Latinos and that changes in the share of blacks have no real impact on EV turnout. Only Montgomery and Prince George’s have relatively high shares of Latinos (and MoCo also has the highest share of Asians), so I suspect that what is being captured by the model is that these two counties have relatively high turnout once you consider that they are not in CD 1, don’t have tight exec races, and the age of their populations.
In case you were counting, Todd Eberly wins the hypothesis prediction race. Both of his panned out while one of my two turned out to be a turkey-and Thanksgiving isn’t quite here yet.