Understanding the Election Administrator Job Market

Much about the landscape of the American elections workforce is poorly understood, especially those roles beyond the chief officials in the over 10,000 local election jurisdictions across the country. Information about chief local election officials (LEOs) is regularly collected as part of the Elections and Voting Information Center (EVIC) LEO survey, but there are few comprehensive data sources about the non-chief, rank-and-file election staff who likely constitute the majority of the election administrator workforce.

A new dataset compiled by BPC sheds light on the job market for election administrators, including non-chief election staff. Our dataset compiles the job postings for election administrators that have been posted publicly on the Electionline Weekly site each week since 2011. Electionline is a popular newsletter and hub for the elections community, aggregating election-related headlines, jobs, and resources on a daily and weekly basis.

Access the dataset directly:

View in Google Sheets > 

Download the CSV from Github >

For the benefit of any researchers interested in analyzing this dataset, we use OpenAI’s GPT-3.5 to process the dataset, extracting information on compensation and determining which jobs belong to chief election officials. (As of this writing, the dataset is updated automatically each week, the day after an edition of Electionline Weekly is issued.)

This project, along with the work underway by BPC’s Election Workforce Advisory Council, reflects BPC’s commitment to researching, enhancing, and innovating election administration recruitment, retention, and training.

Election staff salaries have been flat for years

As an example of how the dataset can be used, we analyzed the salaries included in these job descriptions. We find that the mean of the posted salary ranges for election staff has increased only by about 10% in the last 10 years, from about $65,000 to about $71,000.

The numbers above each point indicate how many job descriptions are included in that point. There are too few posted job descriptions for chief officials to enable us to draw strong conclusions, but EVIC data also suggests that salaries for chief officials have been level since at least 2019.

Staff in larger jurisdictions command higher salaries

EVIC data also suggests that the key driver of chief official salaries is the size of the jurisdiction. Larger jurisdictions have more funding and their chief officials manage more staff, so it’s unsurprising that those chief officials would be better compensated. With the Electionline data, we are able to show that the same trend appears to hold for their staff. According to these postings, a staff member in a jurisdiction with 2.5 million voters is likely to make about $100,000, while a staff member in a jurisdiction with 100,000 voters is likely to make about $66,000.

Caveats

There are some caveats that should be considered when analyzing and interpreting this dataset. First, there is no reason to believe that our dataset is a representative sample of all jobs in the workforce; only 40 states have posted jobs on Electionline, and some of those states post much more frequently than others. The data also consists only of positions that the employer deems worthy of a national online search, and it is possible these positions are more highly compensated than others that an employer wouldn’t post online. Second, the dataset only includes the (typically broad) salary ranges advertised on Electionline, which may not be an accurate reflection of how much officials are actually paid once they interview and negotiate their compensation. Third, while GPT-3.5 seems to be relatively good at extracting basic information from job descriptions such as salary range and employer, it is not perfect. Moreover, its assessment of whether a job description describes a chief election official or a non-chief election staff member appears to be considerably less accurate; we consider this assessment experimental for now.

Looking ahead

With these caveats in mind, we look forward to seeing what researchers are able to glean from this dataset. In particular, we invite researchers adept at natural language processing to analyze the text in these job descriptions. Are there any trends in which responsibilities are commonly listed over the years? Are certain responsibilities associated with higher pay? Are some responsibilities more common to smaller jurisdictions?

The decentralization of election administration in the U.S. makes it challenging to fully understand the election workforce. We know little about how election officials are recruited, what qualities they are recruited for, and how compensation varies across the field. We hope that this dataset is a useful resource for researchers aiming to better understand these critical questions.

Methods 

The code used to create this dataset includes both traditional web scraping techniques and modern AI-powered tools. It is publicly available on a GitHub repository. We summarize that process here, but encourage readers to check the repository for the most current summary.

The following steps are executed automatically each Friday, the day after an edition of Electionline Weekly is typically issued:

• Data acquisition: Download the latest issue of Electionline Weekly.
• Job posting extraction: Use the Beautiful Soup web scraping package to extract job descriptions, organizing them into a pandas table.
• Filtering: Exclude job descriptions posted in previous weeks. Exclude job postings from the most popular private employers, to maintain focus on public sector election roles.
• Feature extraction: Use the scrapeghost package (which in turn uses GPT-3.5) to extract features from the job descriptions, including job title, employer, state name, salary range, and pay basis (e.g., yearly, monthly, hourly).
• Salary normalization: Use pay basis information to adjust pay fields for comparison to a standard yearly salary (e.g., by multiplying monthly salaries by 12).
• Job classification (experimental): Use GPT-3.5 via the OpenAI Python API to determine whether a job description is for a position as a chief election official, a non-chief election staff member, or for a non-election-official position. This classification appears to be only moderately accurate, so should be considered experimental.
• Append new jobs to the previous version of the dataset.
• Upload dataset to GitHub and Google Sheets.