Articles › Variable groups and schema changes

Variable groups and schema changes

This is a detailed reference for the columns in the RIPA stop dataset and how the schema evolved across data years. For a general introduction to the data — what RIPA is, how the data is collected, and its limitations — see About the RIPA Stop Data.

Variable groups

The dataset contains 235 columns organized into the groups described below. Binary flag columns use 0/1 coding. Columns marked “2024 only” are NULL for earlier years. Columns marked “2018–2023 only” are NULL for 2024 data.

In some cases, columns were combined or renamed in order to facilitate cross-era analyses. All such transformations are recorded in this schema document, which drives the cleaning step in this project’s GitHub repository. The cross-era harmonization used for the summary statistics on this site (e.g. combining ADS_* with NFA_*/OFA_* action columns) is defined in the materialized views in database/src/schema.sql.

Identifiers and stop context

Column Description
DOJ_RECORD_ID Unique stop identifier
PERSON_NUMBER Person within the stop (1, 2, …)
AGENCY_ORI Agency ORI code (CJIS identifier)
AGENCY_NAME Agency name
DATE_OF_STOP Date of the stop
TIME_OF_STOP Time of the stop (HH:MM)
STOP_DURATION Duration in minutes
LOC_CLOSEST_CITY Closest city to the stop location

The statewide data does not include fine-grained location (latitude/longitude or address). Some jurisdictions publish more detailed location data through their own open data portals.

Race and ethnicity (RAE_*)

RAE_FULL is the primary race code. Individual race flags allow for multiracial identification.

Code Label
1 Asian
2 Black/African American
3 Hispanic/Latino(a)
4 Middle Eastern/South Asian
5 Native American
6 Pacific Islander
7 White
8 Multiracial

The RIPA Board has documented consistent racial disparities in stops across all seven years of data. Black individuals are stopped at roughly 2–2.5 times their share of California’s population. In the most recent report (2026, covering 2024 data), the Board found that Black individuals comprised 12.09% of stops but only about 5.4% of the state’s residential population—stopped approximately 128% more frequently than expected.

Gender (G_*)

G_FULL is the primary gender code.

Code Label
1 Cisgender Man
2 Cisgender Woman
3 Transgender Man
4 Transgender Woman
5 Nonbinary Person
6 Multigender

Gender column names changed between schema eras. In the raw 2018–2023 data, the columns were G_MALE, G_FEMALE, and G_GENDER_NONCONFORMING. In 2024, they became G_CISGENDER_MAN, G_CISGENDER_WOMAN, and G_NONBINARY_PERSON. The cleaned dataset uses the 2024 canonical names for all years.

The RIPA Board’s 2022 report (covering 2020 data) included the first detailed analysis of stops of transgender individuals. They found:

  • transgeneder men/boys were searched in over 40% of stops, the highest rate among gender groups
  • transgender women/girls were handcuffed (42%) and subjected to detention curbside or in a patrol car (35%) at higher rates than other groups
  • transgender people who were stopped were subjected to force at higher rates than cisgender people

Sexual orientation (SOR_*)

Column Description
SOR_LGB Perceived as lesbian, gay, or bisexual (0/1)
SOR_STRAIGHT Perceived as straight (0/1)

In 2018–2023, only a single LGBT flag was collected (renamed to SOR_LGB in the cleaned data). SOR_STRAIGHT is derived as 1 - SOR_LGB for those years. Starting in 2024, both fields are collected natively.

Age (AGE, AGE_GROUP)

AGE is the officer’s estimate of the person’s age in years. AGE_GROUP provides standardized bins derived from AGE:

Code Ages
1 1–9
2 10–14
3 15–17
4 18–24
5 25–34
6 35–44
7 45–54
8 55–64
9 65+

The raw age group bins changed across years (2023 used different breakpoints), so AGE_GROUP is re-derived from AGE for consistency. The original year-specific value is preserved in age_group_orig.

The RIPA Board’s 2025 and 2026 reports focused heavily on youth, finding that the youngest age groups (particularly ages 12–14) experienced the highest rates of searches, handcuffing, and curbside detention. Black youth ages 12–14 had limited force used against them at a rate of 42.3% (2023 data).

Disability (PD_*)

PD_FULL is the primary disability code. PD_MULTI indicates whether multiple disabilities were perceived (0 = none, 1 = one, 2 = multiple).

Code Label
1 Deafness/Difficulty Hearing
2 Speech Impairment
3 Blind/Visual Impairment
4 Mental Health Condition
5 Intellectual/Developmental Disability
6 Hyperactivity Disorder
7 Other Disability
8 None

The RIPA Board’s 2022 report found that officers used force against people perceived to have a mental health condition at 5.2 times the rate of people with no perceived disability, and searched them at 4.8 times the rate.

Other demographics

Column Description
LIMITED_ENGLISH_FLUENCY Perceived limited English fluency (0/1)
PERSON_UNHOUSED Perceived as unhoused (0/1; 2024 only)

The 2026 report found that 64.8% of stops of unhoused individuals were initiated for reasonable suspicion, higher than for any other group. People perceived as unhoused were also subjected to search, force, and arrest at higher rates than other group.s

Stop type and circumstances (2024 only)

Column Description
TOS_VEHICULAR Vehicular stop
TOS_BICYCLE Bicycle stop
TOS_PEDESTRIAN Pedestrian stop
PASSENGER_IN_VEHICLE Person was a passenger
INSIDE_RESIDENCE Stop occurred inside a residence
CALL_FOR_SERVICE Stop initiated by a call for service
WELFARE_WELLNESS_CHECK Welfare/wellness check

Important limitation: Prior to 2024, the data does not distinguish among vehicular, bicycle, and pedestrian stops. The TOS_* fields are NULL for 2018–2023. This means that analyses of pedestrian or bicycle stop patterns are only possible using 2024 data.

Similarly, the CALL_FOR_SERVICE field is only available for 2024, though the 2020 report noted that approximately 95% of stops were officer-initiated (not calls for service) in the early data.

Reason for stop (RFS_*)

REASON_FOR_STOP codes the primary reason the officer initiated the stop:

Code Label Notes
1 Traffic violation See RFS_TRAFFIC_VIOLATION_TYPE
2 Reasonable suspicion See RFS_RS_* subcategories
3 Known to be on parole/probation/PRCS/mandatory supervision
4 Knowledge of outstanding arrest warrant/wanted person
5 Investigation of whether student violated school policy
6 Consensual encounter resulting in a search
7 Possible conduct warranting discipline under Education Code
8 Determine whether to issue truancy-related document
9 Probable cause to arrest 2024 only
10 Welfare & Institutions Code 5150 2024 only

Traffic violation subcategories

RFS_TRAFFIC_VIOLATION_TYPE further specifies the type of traffic violation when REASON_FOR_STOP is 1:

Code Label
1 Moving violation
2 Equipment violation
3 Non-moving violation (including registration)

Equipment violations (code 2) are a key indicator of pretextual stops — stops where the stated reason is minor but the officer’s actual intent is to investigate something else. The RIPA Board has documented significant racial disparities in equipment violation stops. In the 2022 report (covering 2020 data), non-moving and equipment violation accounted for 31.3% of stops of people perceived as Black, compared to 19.9% for people perceived as White.

The Board has called for eliminating pretextual stops since 2022 and has documented the effectiveness of pretextual stop bans:

  • LAPD (Policy 240.06, March 2022): The 2026 report noted reductions in non-moving violations, decreases in searches, and increases in contraband discovery rates following the policy’s implementation.
  • SFPD (Policy 9.07.04(a), adopted 2023): The 2026 report (covering 2024 data) noted that the policy was still very recent, and the effects were “difficult to ascertain.” A more recent analysis including data through September 2025 found reductions in stops of Black drivers following the policy’s going into effect.

Reasonable suspicion subcategories (RFS_RS_*)

When the reason for stop is reasonable suspicion (code 2), the officer records which factors contributed: offense witnessed, matched suspect description, witness identification, carrying a suspicious object, actions indicative of crime, suspect appearance/demeanor, suspected drug transaction, violent crime suspect, or other. A CJIS code (RFS_RS_CODE) may also be recorded.

Probable cause subcategories (RFS_PC_*, 2024 only)

Added in 2024 for the new “probable cause to arrest” stop reason (code 9). Parallels the reasonable suspicion subcategories.

Reason given to person stopped (RFS_RG_*, 2024 only)

A major addition in 2024: the reason the officer communicated to the person for the stop, which may differ from the officer’s actual recorded reason. This includes 22 subcategories covering traffic violations (moving, equipment, non-moving), investigative reasons, parole/warrant status, and notably RFS_RG_NOT_COMMUNICATED — indicating that the officer did not tell the person why they were being stopped.

Actions taken during the stop

The data records what the officer did during the stop. The structure differs significantly between schema eras.

2018–2023: Combined actions (ADS_*)

In the earlier era, force and non-force actions are recorded in a single set of ADS_* (Actions During Stop) flags. Key columns:

Non-force actions:

  • ADS_ASKED_SEARCH_PER/PROP — asked to search person or property
  • ADS_SEARCH_PERSON/PROPERTY — conducted search
  • ADS_SEARCH_PERS_CONSEN/PROP_CONSEN — search with consent
  • ADS_CURB_DETENT — curb detention
  • ADS_PATCAR_DETENT — placed in patrol car
  • ADS_SOBRIETY_TEST — field sobriety test
  • ADS_PHOTO — photographed
  • ADS_WRITTEN_STATEMENT — written statement taken
  • ADS_VEHICLE_IMPOUND — vehicle impounded
  • ADS_PROP_SEIZE — property seized
  • ADS_NO_ACTIONS — no actions taken

Force actions:

  • ADS_HANDCUFFED — handcuffed
  • ADS_REMOVED_VEHICLE_ORDER/PHYCONTACT — removed from vehicle (by order / physical contact)
  • ADS_FIREARM_POINT/DISCHARGE — firearm pointed or discharged
  • ADS_ELECT_DEVICE — conducted energy device (combined)
  • ADS_IMPACT_DISCHARGE — impact projectile
  • ADS_CANINE_SEARCH/BITE — canine used
  • ADS_BATON — baton used
  • ADS_CHEM_SPRAY — chemical spray
  • ADS_OTHER_CONTACT — other physical contact

2024: Split into non-force (NFA_*) and force (OFA_*)

The 2024 schema separates actions into non-force actions (NFA) and officer force actions (OFA), with more granular subcategories.

Non-force actions (NFA_*) add:

  • NFA_TERRY_FRISK — Terry frisk (pat-down for weapons)
  • NFA_ASKED_ID_PASSENGER — asked passenger for identification
  • NFA_ASKED_PAROLE — asked about parole status
  • NFA_RAN_NAME_PASSENGER — ran name check on passenger
  • NFA_SEARCH_PERS_CONSENT/PROP_CONSENT — consent search of person or property

Force actions (OFA_*) split previously combined categories:

  • Conducted energy device split into OFA_ELECT_DEVICE_POINT, OFA_ELECT_DEVICE_STUN, and OFA_ELECT_DEVICE_DART
  • Baton split into OFA_BATON_DRAWN and OFA_BATON_USED
  • Impact projectile split into OFA_IMPACT_PROJECTILE_POINT and OFA_IMPACT_PROJECTILE_DISCHARGE
  • New categories: OFA_FIREARM_UNHOLSTERED, OFA_PHYSICAL_COMPLIANCE, OFA_USE_VEHICLE, OFA_CANINE_COMPLIANCE

Cross-era analysis note: To compare actions across all years, you must harmonize the ADS_* columns (2018–2023) with the NFA_* and OFA_* columns (2024). For example, “was searched” for 2018–2023 is GREATEST(ADS_SEARCH_PERSON, ADS_SEARCH_PROPERTY), while for 2024 it is GREATEST(NFA_SEARCH_PERSON, NFA_SEARCH_PROPERTY, NFA_TERRY_FRISK). The harmonization used by this site is implemented in the materialized views in database/src/schema.sql.

Searches and seizures

Officers may ask for consent to search a person or their property. Consent search data is spread across several column groups:

  • Request: ADS_ASKED_SEARCH_PER/PROP (2018–2023) or NFA_ASKED_SEARCH_PER/PROP (2024)
  • Conducted with consent: ADS_SEARCH_PERS_CONSEN/PROP_CONSEN (2018–2023) or NFA_SEARCH_PERS_CONSENT/PROP_CONSENT (2024)
  • Basis: BFS_CONSENT_GIVEN indicates consent as the basis for search
  • Consent type (2024 only): CTP_VERBAL, CTP_WRITTEN, CTP_IMPLIED

The RIPA Board has documented that Black individuals are asked for consent to search at 2–4 times the rate of White individuals, depending on the year and context. Consent-only searches have lower contraband discovery rates than other search types, and the racial gap in discovery rates is even wider for consent searches.

The Board has questioned whether consent is truly voluntary given the power imbalance between officer and civilian. From the 2022 report:

While the data reflect that most people consent to a search when asked by an officer, research discussed in the Report reflects that this “consent” is not necessarily voluntarily because of the inherent power inequality between a law enforcement officer and a member of the public. The research shows that this inherent power inequality is particularly pronounced among vulnerable populations, such as people with mental health disabilities or youth, who may be more likely to succumb to authoritative pressure. Indeed, RIPA data reflects that for both people with mental health disabilities and youth, a larger proportion of their stops that began as consensual encounters resulted in searches, as compared to people without mental health disabilities or adults.

Starting with the 2022 report, the Board recommended severely limiting or ending consent searches. By the 2023 report (covering 2021 data), the recommendation strengthened to prohibiting consent and supervision searches entirely.

Basis for search (BFS_*)

When a search is conducted, the officer records the legal basis:

Column Basis
BFS_CONSENT_GIVEN Consent given
BFS_OFFICER_SAFETY Officer safety
BFS_SEARCH_WARRANT Search warrant
BFS_PAROLE Parole/probation condition
BFS_SUSPECT_WEAPON Suspected weapon
BFS_VISIBLE_CONTRABAND Visible contraband
BFS_ODOR_CONTRABAND Odor of contraband
BFS_CANINE_DETECT Canine detection
BFS_EVIDENCE Evidence of a crime
BFS_INCIDENT Incident to arrest
BFS_EXIGENT_CIRCUM Exigent circumstances
BFS_VEHICLE_INVENT Vehicle inventory
BFS_SCHOOL_POLICY School policy

Multiple bases can be indicated for a single search.

Contraband and evidence discovered (CED_*)

After a search, the officer records what was found:

Column Item
CED_NONE_CONTRABAND Nothing found
CED_FIREARM Firearm
CED_AMMUNITION Ammunition
CED_WEAPON Other weapon
CED_DRUGS Drugs/narcotics
CED_ALCOHOL Alcohol
CED_MONEY Money
CED_DRUG_PARAPHERNALIA Drug paraphernalia
CED_STOLEN_PROP Stolen property
CED_ELECT_DEVICE Electronic device
CED_OTHER_CONTRABAND Other contraband

The discovery rate (also called “hit rate”) — the proportion of searches that find contraband — is a critical metric for assessing whether search decisions are applied equitably across racial groups. From the 2020 report:

Yield rates were lower for all racial groups of color compared to White individuals (1.8 to 5.6 percentage points lower). This shows that officers were less successful at finding contraband or evidence of wrongdoing when searching individuals of color than White individuals.

Property seizure (BPS_*, TPS_*)

When property is seized, the officer records the legal basis for seizure (BPS_*: safekeeping, contraband, evidence, vehicle impound, abandoned property, school policy violation) and the type of property (TPS_*: firearm, ammunition, weapon, drugs, alcohol, money, drug paraphernalia, stolen property, cellphone, vehicle, other contraband).

Result of stop (ROS_*)

The outcome of the stop. Multiple results can apply.

Column Result
ROS_NO_ACTION No action taken
ROS_WARNING Warning (2018–2023; combined verbal and written)
ROS_VERBAL_WARNING Verbal warning (2024 only)
ROS_WRITTEN_WARNING Written warning (2024 only)
ROS_CITATION Citation issued
ROS_IN_FIELD_CITE_RELEASE In-field cite and release
ROS_CUSTODIAL_WARRANT Custodial arrest (warrant)
ROS_CUSTODIAL_WITHOUT_WARRANT Custodial arrest (no warrant)
ROS_FIELD_INTERVIEW_CARD Field interview card completed
ROS_NONCRIMINAL_TRANSPORT Noncriminal transport or caretaking
ROS_CONTACT_LEGAL_GUARDIAN Contacted legal guardian or responsible adult
ROS_PSYCH_HOLD Psychiatric hold (W&I Code 5150)
ROS_US_HOMELAND Turned over to US Homeland Security
ROS_REFERRAL_SCHOOL_ADMIN Referred to school administrator
ROS_REFERRAL_SCHOOL_COUNSELOR Referred to school counselor

Cross-era note: In 2018–2023, verbal and written warnings are combined into ROS_WARNING. In 2024, they are separate (ROS_VERBAL_WARNING, ROS_WRITTEN_WARNING). To compare warning rates across all years, combine the two 2024 columns.

Some results have associated CJIS violation codes (ROS_*_CDS columns).

The RIPA Board found that Black individuals had “no action” taken during their stops at the highest rate of any group — 13.1% in 2020 data versus 5.6% for White individuals — suggesting that many stops of Black individuals lacked enforcement justification. The 2022 report found those who were perceived to be Transgender received field interview cards at 2–3 times the rate of perceived cisgender people. The 2024 report found that Black individuals received field interview cards at 4.4 times the statewide per-capita average and recommended prohibiting field interview cards absent arrest.

Column Description
SCHOOL_CODE School identifier
SCHOOL_NAME School name
STOP_STUDENT Person is a student (0/1)
K12_SCHOOL_GROUNDS Stop on K-12 school grounds (0/1)

These fields capture stops involving students and stops on school property, relevant to the Board’s education-related recommendations.

Schema changes over time

The dataset schema has evolved across two major eras:

2018–2023 era (~142 raw columns)

The original schema used combined action flags (ADS_*), a single warning result (ROS_WARNING), a single LGBT field, and did not include stop type, call-for-service, or many of the demographic detail fields added later.

2024 era (~202 raw columns)

A major overhaul that:

  • Split actions into non-force (NFA_*) and force (OFA_*) with many new subcategories (e.g., Terry frisks, baton drawn vs. used, firearm unholstered)
  • Split warnings into verbal and written
  • Added stop type fields (TOS_VEHICULAR, TOS_BICYCLE, TOS_PEDESTRIAN)
  • Added “reason given” (RFS_RG_*) columns capturing what was communicated to the person stopped
  • Added probable cause subcategories (RFS_PC_*)
  • Added consent type (CTP_VERBAL, CTP_WRITTEN, CTP_IMPLIED)
  • Added CALL_FOR_SERVICE, PERSON_UNHOUSED, PASSENGER_IN_VEHICLE, INSIDE_RESIDENCE, WELFARE_WELLNESS_CHECK
  • Expanded REASON_FOR_STOP from 8 to 10 values (added probable cause to arrest and W&I Code 5150)
  • Renamed demographic columns to more inclusive terminology (e.g., G_MALE to G_CISGENDER_MAN, RAE_HISPANIC_LATINO to RAE_HISPANIC_LATINEX)

The 2018–2023 raw columns are reconciled to a single 235-column union schema during cleaning. The full column-by-column mapping — renames, derived columns, and remapped codes — is specified in schema.yaml and applied by clean.py.