SPOTLITE Data

SPOTLITE Data

Download SPOTLITE Data

 

SPOTLITE Image 2

SPOTLITE data aims to capture all incidents involving police uses of lethal force occurring anywhere in the United States between 2014 and 2021. We define a police use of lethal force as any discharge of a firearm by law enforcement personnel as well as any other use of force by law enforcement personnel that produces a lethal outcome.

To be included in the database, SPOTLITE incidents must fall into one or more of the following categories:

  • Incidents where law enforcement personnel discharge a firearm, including both lethal and non-lethal outcomes
  • Incidents where law enforcement personnel use any other type of force that results in a lethal outcome
  • Incidents where law enforcement personnel engage in a chase–whether in vehicles or on foot–during which either an individual being chased or an uninvolved third party dies as a result

We also require that at least one credible document verify the details of an incident for it to be included in the SPOTLITE dataset. In most cases, these documents are news reports but in some cases they are government records or press statements from law enforcement agencies. We collect these documents primarily from two sources: the Gun Violence Archive (GVA) and Fatal Encounters (FE). We also draw on administrative data produced by law enforcement agencies when available, and use associated contextual data to determine when multiple sources are describing the same incident. Each document is evaluated by trained analysts to assess whether or not it describes an incident that would meet the SPOTLITE definition of a police use of lethal force. If there is at least one credible document that includes enough information to determine that the described event meets the SPOTLITE inclusion criteria, then the incident is added to the dataset. Assessments made by analysts are reviewed by supervisors and multiple rounds of quality checks are conducted on incident details included in the dataset.

 

The preferred citation for the SPOTLITE United States datasets is:

Jennings, Jay; Singh, Ajay; Althaus, Scott; Martin, Michael; Bajjalieh, Joseph; and Robbennolt, Jennifer (2023) Cline Center SPOTLITE United States Dataset. University of Illinois at Urbana-Champaign.

 

The preferred citation for the SPOTLITE Illinois datasets is:

Jennings, Jay; Singh, Ajay; Althaus, Scott; Martin, Michael; Bajjalieh, Joseph; Robbennolt, Jennifer and Schlosser, Michael (2022) Cline Center SPOTLITE Illinois Dataset. University of Illinois at Urbana-Champaign.

 

The SPOTLITE Incident Counts Dataset is a count of use of force incidents from 2014-2021 in the United States broken down by state and county. The data are available in the SPOTLITE dashboard and to download as a CSV file. There are additional fields available in the data download that are not visible in the dashboard (see below). In the data we link to the Gun Violence Archive website to provide documentation for incidents tracked by that organization. The links in the dataset will take you to incident specific pages with documents describing the incidents. For Fatal Encounters sourced incidents, users can find additional information about the incident by using the FE_id (Unique ID)  included in the downloadable data on the Fatal Encounters data page

Data Fields  for SPOTLITE Incident Counts Datasets 

incident_id - unique identifier for Illinois SPOTLITE incident. ID is based on date plus a three digit code at the end to uniquely identify incidents on the same day in this format: YYYYMMDD###

date  - Date of the incident

county_name - County where the incident took place

county_fips* - Fips code of county where incident took place

state - State where the incident took place 

year* - Year of the incident

source - Organization the source document is from: Fatal Encounter (FE), Gun Violence Archive (GVA), or an administrative record from a law enforcement agency (e.g., City of Chicago’s Civilian Office of Police Accountability)

GVA_url - Link to the relevant Gun Violence Archive incident page

ADMIN_url - Link to the relevant administrative record from a law enforcement agency 

FE_id - unique identifier within the Fatal Encounters dataset for involved individual

FE_id2 - unique identifier within the Fatal Encounters dataset for a second individual

FE_id3 - unique identifier within the Fatal Encounters dataset for a third  individual

GVA_id*  - Unique identifier within the Gun Violence Archive dataset

ADMIN_id* - Unique identifier with relevant administrative record from a law enforcement agency

*Items not visible in the datatable view of the dashboard and only included in the data download.

For each SPOTLITE incident in Illinois, research team members attempt to identify the names and locate images of individuals involved in lethal force incidents when they are available in credible source documents. When a source document does not contain an image of the individuals involved, analysts look for related imagery from other news articles published about the case, obituaries of the deceased, law enforcement intake photos, and aggregator sites that compile inmate intake photographs. For involved individuals that can be identified both by name and with an image, four trained analysts are assigned to assess the person’s ascribed racial and ethnic characteristics based on the person’s image and name alone. SPOTLITE provides ascribed racial and ethnic characteristics generated from the perceptions of third-party observers using only name and image information because this is likely to be the kind of information available to law enforcement personnel encountering a person on the street. Reliably documenting these externally-ascribed racial or ethnic characteristics is therefore key for understanding how and to what extent the perceived characteristics of civilians involved in these incidents might factor into lethal force decisions by law enforcement personnel. 

As is common in law enforcement data, SPOTLITE uses a single measure of race and ethnicity with the following categories: Native American, Asian or Asian Pacific Islander, Black or African-American, White, and Hispanic or Latino. Only one classification is made for each involved individual. When all four analysts agreed on an ascribed racial or ethnic category for an involved individual, we report that classification. 

SPOTLITE also uses name-based algorithmic classification software to assign ascribed racial or ethnic categories in two instances: 

  1. When there is disagreement among analysts, and the classification program agrees with at least two of the four analysts.
  2. When we have no image of the person and the name classification program is confident of its selection with a probability of greater than 0.8. 

SPOTLITE assigns a level of confidence in our racial or ethnic classification (low, medium, high) based on patterns of agreement among analysts and probability scores from the name classification algorithm. We tested the reliability of this classification process using images and names of individuals taken from administrative records that included race or ethnicity classification and found that this process matched the administrative records in 88% of test cases when using a benchmark set of names and images that included equal numbers of cases across six racial/ethnic categories. When applied to the Illinois incident data, this process yielded an ascribed racial or ethnic characteristic for 96% of cases where we had a name and an image for an involved civilian. Within that set of Illinois cases, our approach yielded results with a high level of confidence 87% of the time, with confidence levels derived from the reliability of this approach when comparing its results against the benchmark cases used in the testing process. A forthcoming white paper will present the full results of these reliability tests for the methods used by SPOTLITE. 

Data fields included with the Illinois subject race/ethnic dataset

incident_id - unique identifier for SPOTLITE incident. ID is based on date plus a three digit code at the end to uniquely identify incidents on the same day in this format: YYYYMMDD###

civilian_id - unique identifier for civilians involved in a SPOTLITE incident. This id is the incident_id with a two digit suffix to uniquely identify civilians involved 

date - Date of the incident

year - Year of the incident

county_name - County where the incident took place

county_fips - Fips code of county where incident took place

state - state where incident took place

source - Organization the source document is from: Fatal Encounter (FE), Gun Violence Archive (GVA) or City of Chicago’s Civilian Office of Police Accountability (COPA) 

FE_ID - unique identifier within the Fatal Encounters dataset

GVA_id - Unique identifier within the Gun Violence Archive dataset

ADMIN_id - Unique identifier with the City of Chicago’s Civilian Office of Police Accountability dataset

name_civilian - Names of the subjects of police uses of lethal force (if known)

ascribed_5group_civilian - Race and ethnicity of civilian grouped into 5 racial and ethnic categories

confidence_5group - The level of confidence in the categorization of the race/ethnicity

ascribed_3group_civilian - Race and ethnicity of civilian grouped into 3 racial and ethnic categories

confidence_3group - The level of confidence in the categorization of the race/ethnicity

Gun Violence Archive (GVA) and Fatal Encounters (FE) incidents originate from media reports compiled by each group. GVA data are incident-based, meaning one incident can include multiple participants, while FE data are participant-based, with each record reflective of one participant. FE includes in their data any encounter with law enforcement that results in a death, which is inclusive of deaths resulting from firearms, chases,  and other uses of force. GVA data includes firearm discharges by the police which produce both fatal and non-fatal results. Deaths resulting from other interactions with law enforcement that do not involve a firearm are not included in GVA data.  

Because of the different structures of these data, significant normalization was required. We first determined the overlap of incidents across each dataset. For the FE data with multiple participants from the same incident listed individually, we collapsed these data into one incident, while retaining the participant information. Additionally, each incident required screening before being included in the final SPOTLITE output. Definitionally, FE and GVA include incidents outside the scope of our event definition. For instance, in situations where civilian A kills civilian B, with police responding and killing civilian A, FE includes information on civilian B in their records. Similarly for the GVA data, their incident tag of ‘officer involved shooting’ includes cases where a civilian fires on law enforcement and law enforcement does not fire back. In both instances, these types of cases fall outside the scope of SPOTLITE. 

Any incident listed in the SPOTLITE database has been examined by multiple trained analysts and supervisors to determine if the event meets our definition for a police use of lethal force. Regarding FE and GVA data, we do not rely on the descriptions of the incident as included within the FE and GVA data. Rather, we seek out original source reporting for each incident, locating media reports that describe the incident, and using these reports to establish the basis for inclusion. Because the links to documents included in both FE and GVA sometimes become non-functional, we also expanded our search efforts to locate archived versions of the media reports from the Internet Archive. Only after this extensive normalization and verification process are events then included in SPOTLITE.