Security map methodology

Data Collection

Dots on the security map are reflective of two sources: newspaper articles and NCITE data. Newspaper articles were collected via a boolean search by three researchers. Each researcher was assigned a set of operative words (can be found below) and completed the assigned searches for each state.

Data Searches

Searches were divided up based on “subject of threat”. Researchers then used boolean searches such as ("threat" OR "violence" OR "assault" OR "stalking") ("federal worker" OR "federal employee") (“injury” OR “firearm”). Searches were completed in the “news” tab of Google with the date range of January 1st, 2015 - September, 2025. Searches were marked off as they were completed. The spreadsheet of articles was then deduplicated using the urls so that the same story was not reviewed more than once.

1. Action 2 Subject of threat 3. Additional terms to add
threatpublic officialinjury
warngovernment representativeprosecution
swatgovernment workerconvicted
doxelected officialinvestigation
stalkjudgeus marshall
harrasscivil servanttrial
assaultpoll workeracquit
violencemilitaryharm
abusepolicedeath
menaceBoard of educationdied
damagefederal workeronline
congressphysical
legislatorpsychological
senatorfear
law enforcement officerweapon
judicial officerproperty
school boardonline
election workersocial media
politicianarrested
detectivedetained
sheriffgun
lawmakerfirearm
prosecutordata broker
afraid
18m

Data Coding

Dropdowns were created in the google sheets where the newspaper stories were collected to ensure that coding was consistent across coders. Three researchers coded the newspaper stories independently (each was assigned a block of stories to complete). The following information was collected and coded for each story: area impacted (i.e. city or county); state; who threatened; job, agency, and program of threatened person; party affiliation of threatened person; type of threat; threat acted upon?; action taken after threat; who directly impacted; # of people directly impacted”;indirect impact quote; direct impact quote; and source. Stories were coded according to their contents, meaning that if a story mentioned “stalking”, it was coded as “stalking”. Coding strictly according to the story’s contents rather than by interpretations of the text left less room for inconsistencies between coders. Additionally, stories were coded such that multiple types of threats or public servant types were identified. For example, if an elected judge was harassed and stalked, it would be tagged “elected official”, “judge”, “harassment”, and “stalking”. 

Below is a list of the dropdown options for each category that can be sorted by on the map.

Type of Threat Public Servant type
HarassmentElected Official
Threatening StatementPublic Official
StalkingElection Official
DoxxingLaw Enforcement
SwattingJudge
Physical AttackSchool Board Official
Suspicious PackageOther
Vandalism
Death to Threatened Party
Other

Baseline Data

“Local Threats”: Bridging Divides Initiative Data (BDI) The Threats and Harassment Dataset (THD) is a first-of-its-kind public dataset capturing hostility against local officials in the United States. The longitudinal event-based data tracks the rate, frequency, types, and targets of threats and harassment faced by a wide range of local officials around the country, from elected officials at the municipal, county, and township level to appointed officials and election workers. BDI updates the THD each month to provide users with near-real-time data on the evolving threat landscape and support evidence-based decision-making to protect civic spaces.

Total number of Federal, State, and Local government employees from BLS.gov