The Data Around How Many Poll Workers We Need in 2020
Charles Stewart, founder of MIT’s Election Data and Science Lab, and Democracy Works CEO, Seth Flaxman, have published a simple simulation aimed at estimating the number of election workers who must be recruited to staff in-person polling places on Election Day for the November 2020 election. Many refinements can be made to it, and you are invited to try your hand at it. This model is offered as the beginning of a discussion about the urgent national need to recruit new election workers for the 2020 election.
The main worksheet is Workers needed, which reports the number of poll workers and number of in-person polling places in each election jurisdiction in the U.S. using 2018 data from Democracy Works and the U.S. Election Assistance Commission in the 2016 Election Administration and Voting Survey (EAVS). To help overcome reporting and statistical variability, the average number of poll workers was calculated for each state, and it is used to estimate the needed number of poll workers by multiplying this statewide average by the number of in-person polling places in each jurisdiction.
The worksheet Vote sites per locality reports the number of Election Day and early voting sites reported by each jurisdiction in 2018. Although this spreadsheet only estimates the number of needed Election-Day poll workers, data in this sheet could be used to help add information about early in-person voting.
Avg poll worker per site is, as the label suggests, the average number of poll workers in each polling site for each state, calculated using EAVS data. The spreadsheet notes a few states where the data provided are implausible (e.g., less than 1 poll worker per site) or missing. In those cases, we have substituted the nationwide average.
Mail ballot pct in 2016 provides estimates of the percentage of ballots in each state cast by mail in 2016, using data from the U.S. Census Bureau Current Population Survey, Voting and Registration Supplement. (This dataset is used, rather than the EAVS, because of the missing data and obvious reporting errors that plague the EAVS for many states on these items.) The purpose of this sheet is to estimate how many voters (measured as a percentage of turnout) will shift from in-person voting to mail ballots in 2020, compared to 2016. The idea is that a state such as Arizona, which saw 80% of voters cast ballots by mail in 2016, is already staffed on Election Day as if it is a "vote-at-home" state. Therefore, shifts to mail balloting in Arizona will have little impact on the need for poll workers. A state like North Carolina, on the other hand, which only saw 5% vote by mail in 2016, is likely to see a much larger relative flow of voters from in-person to mail balloting in 2020. This is important for modeling, because we suspect that states may reconsolidate and close polling places in response to voters moving to mail balloting. This response will be of a greater magnitude in North Carolina than in Arizona. The calculations in this spreadsheet are used to help calculate the "allowance for closures" column in the Workers needed spreadsheet.
The Absenteeism spreadsheet records the percentage of existing poll workers who are anticipated not to work in 2020. For simplicity, the calculation is uniform for the country, although once could apply the logic to each state individually. The sheet notes that according to the EAVS data, 35% of 2016 poll workers were 61-70 years of age and 25% were over 70. If half of the 61-70 year olds decline to work and 90% of the 70+ cohort (and all other poll workers stay on the job), this works out to a conclusion that 40% of poll workers will need to be replaced for 2020. Attention to state-by-state variation in the age distribution of poll workers, along with informed estimates of potential absenteeism, could further refine these estimates.