# Quality report

## Release

- Dataset: Washington Monthly Motor-Vehicle-Theft Offenses, 2018-2025
- Research release: 0.1
- Expected monthly rows: 96
- Expected range: 2018-01 through 2025-12
- Primary source: FBI Crime Data Explorer summarized state endpoint

## Automated checks

The build fails unless:

1. There are exactly 96 monthly observations.
2. Every month from January 2018 through December 2025 appears exactly once.
3. Offense, clearance, population, rate, and population-coverage values are present.
4. Counts are nonnegative integers.
5. Rates and coverage values are nonnegative.
6. Population coverage does not exceed 100%.
7. Annual summaries reconcile exactly to monthly values.
8. Every generated artifact receives a SHA-256 digest.

## Definitions

### Reported offenses

`wa_reported_offenses` is the FBI CDE value labeled "Washington Offenses" for the motor-vehicle-theft summarized endpoint. It is an offense count, not a count of unique vehicles, victims, suspects, arrests, or pursuits.

### Clearances

`wa_clearances` is the FBI CDE value labeled "Washington Clearances." Clearances are not synonymous with arrests, and a clearance recorded in one month need not correspond to an offense first reported in that month. The ratio of monthly clearances to monthly offenses must not be described as a cohort resolution rate.

### Rates

Rate fields reproduce the values returned by the FBI. FactHub does not recalculate or replace them. The FBI labels these values as offenses or clearances per 100,000 people.

### Population coverage

Coverage fields reproduce the CDE tooltip values labeled "Percent of Population Coverage." They estimate the share of the relevant population represented by reporting agencies. They are not percentages of records known to be complete.

## Coverage

Washington's CDE population coverage is above 98.7% in every month from 2018 through June 2025. It drops into approximately the 95.7%-96.1% range from July through December 2025.

Consequences:

- 2025 counts should not be compared mechanically with earlier years without considering coverage.
- Rate fields may be preferable for descriptive comparisons, but they do not eliminate all reporting bias.
- Any model should include a coverage control or run a lower-coverage sensitivity analysis.

## Reconciliation and source conflict

The FBI CDE series does not reconcile to the publicly reported WSP/WACIC stolen-vehicle totals:

| Year | FBI CDE reported offenses | Public WACIC/WATPA figure | Difference |
|---|---:|---:|---:|
| 2020 | 28,036 | 26,520 vehicles stolen | 1,516 |
| 2021 | 35,662 | 31,032 vehicles stolen | 4,630 |
| 2024 | 34,650 | 28,300 auto thefts | 6,350 |
| 2025 | 20,199 | 17,499 auto thefts | 2,700 |

WACIC references:

- The 2020 WATPA annual report directly states 26,520 cars stolen.
- WASPC's April 2022 release directly states 31,032 vehicles stolen in 2021.
- WATPA's 2025 annual report directly states 17,499 auto thefts in 2025, 10,801 fewer than 2024, implying 28,300 in 2024.

This release does not choose one system as universally correct. It treats the values as different measures until the agencies provide a definition and reconciliation. Articles must identify which series they use.

## Known limitations

1. This is a statewide aggregate. It cannot identify which agencies or counties drove a change.
2. It does not contain vehicle make or model and therefore cannot directly control for the Kia/Hyundai theft wave.
3. It does not measure attempted theft separately.
4. It does not contain arrests, suspect age, vehicle recoveries, or unique victims.
5. It does not measure failures to yield or incidents in which officers did not initiate a pursuit.
6. It does not contain police staffing, deployment, prosecution, detention, or task-force activity.
7. FBI source values may be revised after this release.
8. The national series is not a causal counterfactual for Washington.
9. Policy changes occurred mid-month.
10. The OFM/UW pursuit study covers a non-census sample of agencies, uses heterogeneous records, excludes non-pursued "fleds," and covers 2019-2024 rather than the full theft period.

## Policy-period descriptives

`data/policy-period-descriptives.csv` excludes the partial effective month for each legal change and reports simple averages for the remaining full months. These figures are descriptive, unadjusted, and highly confounded. They are included to make the first comparison reproducible, not to estimate a causal effect.

The period averages must not be described as the amount by which a law increased or decreased theft.

## Manual review still required

Before this dataset supports a published causal article:

- Obtain a definition crosswalk from FBI UCR, WASPC, and WSP/WACIC.
- Request the complete WACIC agency-by-month extract.
- Add agency reporting status and revision history.
- Select comparison states using prespecified criteria.
- Add vehicle-make composition where available.
- Write a pre-analysis plan before fitting alternative models.
- Reproduce the analysis independently.