Chris J. Vargo

Boulder County Domestic Violence Enforcement Audit

Findings

The strongest verified finding is currently a limitation: the final disparity hypotheses are unresolved in the public release because the required model outputs and denominators are absent.

hypothesis id exact preregistered wording primary estimand primary model expected direction observed estimate confidence or credible interval multiplicity rule smallest effect size of substantive interest sample size denominator agency and period robustness threshold support status analysis commit data hash preregistration hash verification timestamp
H0 Recorded suspect sex is not associated with an enforcement decision among incidents with comparable pre-decision evidence and circumstances. P(adverse enforcement | comparable event-party pre-decision facts) not run for confirmatory inference two-sided unless preregistered otherwise pending pending all qualifying domestic-related event-parties, including no-arrest and no-report-only events not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.502942+00:00
H1 The share of male suspects changes materially at one or more funnel transitions. change in recorded male share across suspect, arrest, referral, filing, conviction, and sentence stages not run for confirmatory inference two-sided unless preregistered otherwise pending pending stage-specific linked counts from call/event-party through court disposition not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503135+00:00
H2 Male suspects have a different adjusted probability of arrest or sole-aggressor designation after conditioning only on facts available before the police decision. adjusted risk difference in arrest or sole-aggressor designation by recorded suspect sex not run for confirmatory inference two-sided unless preregistered otherwise pending pending all qualifying event-parties with pre-decision facts not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503138+00:00
H3 In opposing-complaint or bidirectional incidents, the within-incident probability of being the sole arrested party differs by recorded sex. within-incident probability of sole arrest among opposing-complaint/bidirectional incidents not run for confirmatory inference two-sided unless preregistered otherwise pending pending incidents with opposing complaints or independently codable bidirectional conduct not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503140+00:00
H4 Investigation-completeness items differ by suspect or alleged-victim sex after accounting for feasibility and relevance. item-level and index-level investigation-completeness differences by recorded sex/dyad not run for confirmatory inference two-sided unless preregistered otherwise pending pending all relevant and feasible investigative-action opportunities in sampled records not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503141+00:00
H5 Independently coded civil-only incidents involving male potential suspects have a different probability of criminal classification, DV designation, arrest, or referral. P(criminal/DV escalation | independently rated civil-only event) by potential suspect sex not run for confirmatory inference two-sided unless preregistered otherwise pending pending independently coded civil-only event-parties not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503143+00:00
H6 Independently coded probable-cause incidents involving female potential suspects have a different probability of report-only or civil disposition. P(report-only or civil disposition | independently rated probable-cause event) by potential suspect sex not run for confirmatory inference two-sided unless preregistered otherwise pending pending independently coded probable-cause event-parties not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503144+00:00
H7 DA filing, count retention, dismissal, plea, conviction, and sentence differ by recorded defendant sex after stage-appropriate adjustment. stage-specific DA/court outcome differences by recorded defendant sex not run for confirmatory inference two-sided unless preregistered otherwise pending pending all referred/filed/resolved DV-related counts with sex/gender and stage outcomes not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503145+00:00
H8 Agency-specific adjusted disparities differ across Boulder County agencies. heterogeneity of adjusted disparities across Boulder County agencies not run for confirmatory inference two-sided unless preregistered otherwise pending pending agency-specific comparable event-party cohorts not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503147+00:00
H9 Boulder County estimates differ from matched Colorado agencies after case-mix and data-completeness adjustment. Boulder-vs-matched-Colorado-agency difference in adjusted estimates not run for confirmatory inference two-sided unless preregistered otherwise pending pending matched Colorado agency incident-level cohorts with comparable fields not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503148+00:00
H10 Reporting party, caller, scene-presence, and police-contact selection differ across sex and dyad categories, making arrest-only comparisons materially biased. sex/dyad differences in reporting, caller, and scene-presence selection not run for confirmatory inference two-sided unless preregistered otherwise pending pending calls/events with caller role/sex, reporting role/sex, scene presence, and outcomes not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503150+00:00
H11 Verified policy or training changes alter investigation completeness or enforcement patterns. policy/training change effect on investigation completeness or enforcement patterns not run for confirmatory inference two-sided unless preregistered otherwise pending pending pre/post cohorts around verified implementation dates with stable measurement not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503151+00:00
H12 Officer, supervisor, prosecutor, agency, and year effects account for meaningful heterogeneity. variance and clustering by officer, supervisor, prosecutor, agency, and year not run for confirmatory inference two-sided unless preregistered otherwise pending pending case-level cohorts with pseudonymous actor IDs and outcomes not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503153+00:00
H13 Results differ across race or ethnicity, age, disability or mental-health indicators, housing status, language access, and LGBTQ status where fields are reliable, lawful, and sufficiently powered. heterogeneity by race/ethnicity, age, disability/mental-health indicators, housing, language, and LGBTQ status not run for confirmatory inference two-sided unless preregistered otherwise pending pending legally and ethically releasable group fields with adequate cell sizes not established for confirmatory model pending unresolved_required_data_unavailable not available in non-git research workspace 31bccdf8841d4c11b9c1491d2f5eca49471b065bca18d25bb3e56858c1ed2927 4cd0ab806332360a3fde285c63a1493540a862398c5bff0a9bc7c6332b6f47c5 2026-06-24T16:49:49.503154+00:00