Articles › What is RIPA? Where does this data come from?

About the RIPA Stop Data

What is RIPA?

The Racial and Identity Profiling Act of 2015 (AB 953) requires California law enforcement agencies to collect and report data on every stop they conduct (California Penal Code 13519.4). Officers record who they stopped, why, what actions they took, and the outcome — all linked to the officer’s perception of the stopped person’s race, gender, age, and other identity characteristics.

The data is submitted to the California Department of Justice, which publishes it through the OpenJustice Data Portal. An independent RIPA Board, housed within the DOJ, analyzes the data and publishes annual reports with findings and policy recommendations.

This dataset covers 2018–2024 and contains approximately 26.3 million person-stop records from 555 law enforcement agencies across California.

This page describes the data itself — where it comes from, how it is collected, and what it does and does not contain. For how this site turns that data into the numbers you see on each agency page — race categories, what counts as a search, force, or arrest, how disparity ratios and population benchmarks are computed — see Methodology: How the Metrics Are Defined.

Data collection

How agencies report

Agencies submit stop data to the DOJ’s Stop Data Collection System (SDCS) through one of three methods: a DOJ-hosted web application, a local database connected via web services or SFTP, or batch file upload. Records must pass logic checks (e.g. age is non-negative) before acceptance.

Phased rollout

Not all agencies reported from the start. RIPA required agencies to begin reporting based on their size:

Wave Agency size Collection start First report due Agencies
1 1,000+ officers July 1, 2018 April 1, 2019 ~8
2 667–999 officers January 1, 2019 April 1, 2020 ~7
3 334–666 officers January 1, 2021 April 1, 2022 ~10
4 1–333 officers January 1, 2022 April 1, 2023 ~500

Instances of data fabrication or misreporting

In 2022, the Los Angeles Office of the Inspector General (OIG) published an investigation into the LA Sheriff’s Department, comparing stop data from CAD logs to data recorded in the Sheriff’s Automated Contact Reporting System (SACRS) that feeds into RIPA. They found the SACR system underreported stops by at least 50,731 stops, and underreported arrests by at least 71,462, between July 2018 and June 2019. They further found:

the practice of not entering data into the SACR system may be pervasive and widespread throughout all of the Sheriff’s Department’s patrol divisions. In addition, the Office of Inspector General found significant differences between CAD system and SACR system totals relating to backseat detentions, consent searches, and reasonable suspicion stops.

The Sacramento Observer reported on the story

In June of 2024, the San Francisco Department of Police Accountability found one of the department’s top ticket writers was systematically misreporting the race of the people he stopped, causing “irreparable harm to the integrity of SFPD’s RIPA reporting”. The story was covered in the SF Standard

Other known data quality issues

Each year’s data release comes with a README documenting “known errors” — records that failed the DOJ’s logic checks but were never corrected by the agency and entered the dataset anyway. The 2024 README notes that roughly 1% of reported records have errors and are not resubmitted by the reporting deadline. Rather than transcribing every year’s list, here are the recurring categories, with the range of affected records per year, and links to each README: 2018, 2019, 2020, 2021, 2022, 2023, 2024.

Contradictory reason/action combinations. The two largest recurring errors are stops coded “consensual encounter resulting in search” with no search recorded (28,148 records in 2018 — likely officers misunderstanding the form — falling to 5,011 in 2019 and between 1 and 34 per year since), and searches justified as “incident to arrest” where no arrest was recorded (5,404 in 2018; 13,649 in 2019; 12,251 in 2020; 11,144 in 2021; this check no longer appears in the 2022–2024 READMEs).

Missing required subfields. Every year includes records where a reason for stop was given but its required subcategory was not: reasonable-suspicion subcategories missing (from 2 records in 2018 up to 1,642 in 2019), traffic violation type or offense code missing (roughly 15–65 records per year).

Search and seizure inconsistencies. Searches with no basis-for-search recorded (2–424 per year), basis-for-search completed with no search indicated, and property-seizure fields that contradict the actions-taken flags (single digits to ~600 records per year).

Missing or impossible values. Small numbers of records each year are missing race (1–3), gender (1–4), actions taken (up to 99 in 2022; 677 non-force and 1,279 force-action records in 2024), or result of stop (300 in 2024); others report impossible ages (0, 120, 445, -120, 1336, 9999) or K-12 students perceived to be older than 22.

Beyond these record-level errors, the READMEs document several systematic, agency-level problems:

  • CHP’s missing transgender records (2020): a transmission bug caused “nearly all individuals perceived to be transgender” to be excluded from CHP’s successfully submitted 2020 data (over 1,000 records). CHP fixed the issue starting with 2021 data.
  • Missing December 31, 2022 records: 4,066 corrected records from the last day of 2022 (0.09% of that year’s records) failed to load into the DOJ’s analysis table and are excluded from the 2022 data file and the 2024 Board Report; the DOJ publishes them as a separate supplement file on OpenJustice.
  • Placeholder citation codes: several agencies — including the LA County Sheriff, LAPD, and CHP in early years, and six agencies in 2023 — submitted only a single generic value (65002) for the citation offense code field, making citation offense detail unusable for those agencies.
  • Community caretaking stops are not separately identifiable: there is no “community caretaking” reason for stop; officers are instructed to select “reasonable suspicion” plus a special offense code (99990), and the READMEs caution that many officers may not have used the code.

Important limitations - what’s not collected

  • No fine-grained location: The statewide data includes only LOC_CLOSEST_CITY — no address, latitude/longitude, or beat/district. Some jurisdictions publish more detailed location data through their own open data portals.

  • No officer identifiers: The statewide data does not include officer badge numbers or other identifiers, preventing analysis of individual officer patterns

  • Complaint data limitations: The RIPA Board collects civilian complaint data alongside stop data, but the complaint system has significant limitations. The sustained rate for profiling complaints has been extremely low — reaching 0.19% in the 2024 data (3 out of 2,282 profiling complaints sustained). Until November 2025, Penal Code section 148.6 imposed criminal sanctions for filing a “false” complaint, which the California Supreme Court found unconstitutional. The Board documented this as a deterrent to filing complaints for years before the ruling.

Unit of observation

Each row in the dataset is a person-stop: one person involved in one stop. A single stop event (identified by DOJ_RECORD_ID) can produce multiple rows if the officer stopped more than one person. The composite key is (DOJ_RECORD_ID, PERSON_NUMBER).

How demographics are recorded

All demographic data reflects the officer’s perception of the stopped person. From the RIPA Board’s 2018 report:

With respect to the person stopped, the officer must report his/her own perception, based upon personal observation only (and not through any other means, such as asking the person or referring to identification), regarding the following:

  1. Perceived race or ethnicity of the person stopped
  2. Perceived age of the person stopped
  3. Perceived gender of the person stopped
  4. Whether the person stopped is perceived to be lesbian, gay, bisexual or transgender
  5. Whether the person stopped is perceived to have limited or no English fluency
  6. Whether the person stopped is perceived or known to have a disability

For a detailed reference covering every variable group in the dataset — and how the schema changed over time — see Variable groups and schema changes.

Additional resources