Fewer poll workers means fewer votes — here’s why

By Harrison Kreisberg
October 15, 2020

This research was done in partnership with the AFL-CIO.

Across the largest counties in battleground states, a major deficit of poll workers could lead to polling closures costing a potential 40 votes per location or 7–10 per poll worker, with voters of color being most affected.

Americans looking to cast their ballots this year face unprecedented challenges. With the COVID-19 pandemic potentially complicating in-person voting, the United States is facing a record shortage of poll workers. Poll workers are ordinary citizens who take a day to work at a polling location and help administer the election. These poll workers make sure elections run smoothly and ensure that everyone’s vote is counted.

Staffing a polling location is hard work, and is often done by older citizens who are at increased risk for complications from COVID-19. A sudden drop in available staff could cause last-minute and chaotic closures. Secretaries of State and Boards of Elections across the country have been consolidating polling places in anticipation of these issues.

Research has repeatedly shown that moving or closing polling locations decreases turnout. In fact, just this year, the sudden closure of over 95% of polling places in Milwaukee was projected to have decreased turnout by almost 10% in the Wisconsin Presidential Primary.

Today we’re excited to share our work that demonstrates just how important it is to keep every polling location open in some of the most consequential counties in this election. BlueLabs analyzed voting trends and staffing requirements to approximate the number of votes lost when polling locations in major counties shut down because there weren’t enough poll workers. These estimates are our best guess at how a poll worker shortage could affect turnout. They point to the critical nature of the work being done by organizations like Power the Polls and More Than a Vote.

In brief, we found that preventing polling place closures by working the polls can potentially boost turnout more than many of the classic GOTV efforts that campaigns engage in:

There will be considerable need for in-person polling locations despite increases in vote by mail. We need to support staffing up to 22,000 locations across the battlegrounds states (AZ, FL, GA, ME, MI, MN, NC, NH, NV, PA and WI) which means we need almost 200,000 poll workers on election day.
If polling locations in major counties close because we lack poll workers, we expect to lose an average of 40 votes per location closed, and 16 of those are likely to come from voters of color.
Each additional location is particularly valuable in the southeast for ensuring access for voters of color. Keeping polls open everywhere is invaluable, but the estimated losses among voters of color is particularly pronounced in Florida, North Carolina, and Georgia.

To develop these estimates, the team at BlueLabs:

Estimated the percent of voters who will vote absentee. We used changes in absentee rates in post-COVID primaries and subtracted this from a national turnout estimate.
Estimated poll workers needed per site and per voter from the Survey of Election Administrators.
Estimated the loss in turnout for voters of color and white voters from locations closing due to lack of poll workers based on the academic research.

This left us with an estimate of how many voters we could lose due to closures affecting turnout of in-person voting. It’s clear from the data how critical the work of a poll worker is to keeping our democracy running smoothly.

In the rest of this article we’ll walk through our methodology for arriving at these estimates. These numbers show that, in the face of large consolidations affecting the majority of voters, ensuring poll worker coverage can be one of the most efficient ways to ensure high turnout, with returns above typical effects seen in canvassing, phone-banking, or texting. That’s why we’d encourage you to sign up to help and make sure that everyone can exercise their franchise.

Estimating Vote by Mail

To first understand the need for sites, we need to estimate how many voters will likely be using them and how many will switch to vote by mail (VBM).

While the loss of poll workers and polling sites will significantly impact in-person voting, this is at least partially offset by large increases in VBM. Many states are expanding access to mail-in voting as a safer alternative to casting ballots in person.

To estimate the number of votes cast by mail, we looked at profiles of who voted by mail in the 2016 primaries and compared that to the five states that had primary elections post-COVID for which we had in-person turnout information. This approach allowed us to see how voting behavior has changed. We focused on summer primaries, after the initial lockdowns, to best estimate the environment in November.

For each group (ex. White Democratic men under 30 who voted in 2016) we scaled their likelihood to vote by mail from the 2016 rate by the group’s increase in between the 2016 and 2020 primaries. While all groups voted by mail at much higher rates in 2020, certain groups disproportionately opted to cast an absentee ballot this year. For example:

Older voters who traditionally use VBM the most were even more likely to vote by mail this year.
By race, voters of color switched their voting method at lower rates than White voters.

Based on these assumptions we came up with an electorate with vote by mail rates in the 55–60% range. That rate is much higher with older voters and lower with voters of color. These are estimates, and may be slightly high given that most primaries were conducted when COVID-19 was still a new issue and concerns about the Postal Service weren’t as prevalent, however, we wanted to be conservative in our analysis of the in-person challenges.

Estimating In-Person Infrastructure Needed

Next, we calculated the vote-to-poll worker ratio for each county from the 2018 Election Administration and Voting Survey (EAVS). For states that did not have available data on poll workers and locations, we imputed.

The numbers vary substantially by state and county so we used county-level results to refine our estimates. The box-plots show the average per state (line in the middle) as well as the interval between which most locations fell (the boxes) as well as the full range (the lines — these exclude outliers which you can still see in tiny dots).

With our assumed rate of mail voting, this analysis suggests we need between 19,000 and 22,000 voting sites to support the in-person electorate. That translates to between 165,000 and 190,000 poll workers across the battleground states.

Estimating The Effect of Consolidation

Previous research shows a wide range of negative impacts on turnout from closing or moving polling locations. Effects tend to be more severe in low turnout elections (where more voters are on the fence about voting to begin with) or when closures are more widespread. Most research found the decrease in turnout came from individuals having to vote at different sites or having to travel further to vote. This makes a lot of sense because we know from other campaign tactics that voting is habitual. In this case, when that habit is changed, voters who now have to go to different locations or go further out of their way are slightly less likely to cast a ballot.

The graph below has a dot for each study we analyzed:

The effect size is plotted on the vertical axis;
The year of the study is plotted on the horizontal axis;
The mechanism analyzed (e.g. distance to poll, consolidation, etc.) is indicated by color.

Most studies are in the 1–3% range, but a few show some very significant effects.

Two of the studies we analyzed seemed particularly instructive for our case:

Brady & McNulty looked at poll consolidation in LA County for the 2002 gubernatorial recall and found a 2% net decrease in turnout of voters affected by moving a polling location.
Miller & Morris looked at the 2020 Wisconsin primary where turnout dropped by nearly 9 percentage points when Milwaukee consolidated 95% of its polling locations.

Many of these studies have also found that Black and Latinx voters are often unduly disenfranchised by polling place closures and consolidation. This coupled with the fact that these communities have borne the brunt of the COVID-19 pandemic has the potential to have a compounding impact on their turnout in 2020.

Based on all of this information, we calculated turnout decreases for two possible scenarios:

Consolidation on the order of 65% of polling locations and a subsequent decrease in turnout by 3% for that proportion of remaining (non-VBM) voters. Given the outside effects on voters of color, we assumed their turnout dropped 3.3% (this is scaled down from some academic findings given the higher turnout nature of the Presidential election).
Consolidation similar to Wisconsin, with the same effects found by Miller & Morris (8.6pp overall and 10.2pp for voters of color).

If in-person turnout drops 3% due to ~65% decreases in locations, we expect that in population dense counties across the battleground states we’d lose an average of 42 votes per location, and 9–11 votes per poll worker. On average we’d lose 16 voters of color, or 3–5 voters of color per poll worker.

We found effects to be largest for voters of color in particular in key counties in North Carolina, Florida, and Arizona (as well as Clark in NV and Fulton in GA).

If in-person turnout drops 8.6% due to ~95% decreases in locations, we expect that in major counties across the battleground states we’d lose an average of 122 votes per location, and nearly 50 votes from voters of color (an additional 9 per poll worker).

With each individual poll worker being responsible for potentially dozens of votes, and the disproportionate impact on voters of color, we’d encourage you to sign up. Volunteering to staff a polling location can help ensure that this election runs smoothly for everyone.