Tier 1 OCC Deployment Case Study: Washington, DC FOIA Request Processing (Aug 2020–Aug 2025)

Tier 1 OCC Deployment Case Study: Washington, DC FOIA Request Processing (Aug 2020–Aug 2025) DOI 10.5281/zenodo.18073749 Abstract This record publishes a complete Tier 1 deployment of the Obligation Closure Constraint (OCC) on a single contestable administrative boundary: Washington, DC FOIA request processing. Using an immutable raw dataset included in the archive and locked boundary definitions, the deployment computes the Tier 1 accounting series—arrivals, attempted closures, durable closures via stick rates at 30/90/180-day horizons, return-work via a defined reopen channel, unresolved stock, tail age metrics, and a challenge-intensity proxy based on appeals. The stock–flow identity is validated across the observation window, and standard diagnostic plots and machine-readable summaries are produced. The resulting regime diagnosis is “Busy but stuck”: closures are durable (high stick rate), but durable capacity lags demand, driving sustained backlog growth. This record is intended as a reproducible case example of OCC instrumentation rather than a causal explanation of staffing, policy, or institutional intent. 1. What this record is This record hosts a complete Tier 1 OCC deployment package for a real operational boundary. It is designed to be auditable and reproducible. All measurement definitions are locked, the raw input is included with integrity tooling, and all reported series and figures are derived deterministically from those inputs. No external data or hidden assumptions are required to reproduce the results. 2. Boundary and measurement object All boundary definitions are locked in the file occ_manifest. These definitions constitute the measurement object. Boundary (B): DC government FOIA request processing Case unit (U): one FOIA request per REFERENCE_NO Standard (F): a request is treated as resolved when the dataset indicates closure (closure date / closed status) Primary horizon (T): 90 days Horizon set: 30, 90, 180 days Reopen channel (RT): status transition to “Requestor Appeal Received” after prior closure Challenge proxy (Z): appeals per attempted closure, defined using the reopen channel Changing any of these definitions changes the boundary being measured and invalidates direct comparison. 3. Data and integrity The raw DC FOIA Requests dataset is included under raw/ as FOIA_Requests. Dataset integrity is supported by the included checksum artifacts (CHECKSUMS.sh and its companion checksum file), binding all derived outputs to a specific immutable input. The dataset schema provides the minimum required fields for Tier 1 OCC instrumentation: a stable case identifier (REFERENCE_NO), an arrival timestamp (CREATE_DATE), a closure timestamp (CLOSE_DATE) and/or closure status, status history sufficient to detect the defined reopen event. 4. Tier 1 instrumentation method The deployment uses monthly aggregation from August 2020 through August 2025 and computes the standard Tier 1 OCC series: Arrivals \lambda_{\text{exo}}(t): monthly arrivals by CREATE_DATE Attempted closures \mu_a(t): monthly closures by CLOSE_DATE Stick rate q(t;T) for T \in \{30,90,180\}: fraction of attempted closures that do not return via the defined reopen channel within horizon T Durable closures \mu_d(t;T) = \mu_a(t)\,q(t;T) Return-work \rho(t;T): reopen events per attempted closure Unresolved stock L(t): end-of-month unresolved inventory under the defined standard Tail age metrics: median and upper percentiles (including p90 and p95 where computed) of unresolved request age Challenge-intensity proxy Z(t): appeals per attempted closure Tier 1 prioritizes accounting consistency and observability over causal interpretation. 5. Accounting identity check The stock–flow identity is validated across the observation window: unresolved stock evolves as prior stock plus arrivals minus closures, within the limits imposed by monthly aggregation and documented data irregularities. This identity functions as a measurement sanity check; failure would indicate a boundary or instrumentation error. 6. Outputs and artifact map All outputs included in the archive are derived directly from the raw dataset and locked definitions. Definitions and integrity occ_manifest — locked boundary definitions and dataset reference raw/FOIA_Requests — raw input dataset CHECKSUMS.sh and checksum file — integrity tooling Derived data derived/timeseries — monthly Tier 1 time series derived/diagnostics.json — machine-readable summary metrics and checks Figures (figures/) arrivals_vs_closures dcr_rolling stick_rate_by_horizon stock_and_tails challenge_intensity Documentation README — narrative summary, assumptions, limitations, and headline metrics 7. Headline observations As summarized in README: Average monthly arrivals: ≈152 requests Average monthly attempted closures: ≈224 90-day stick rate: effectively ~100% (appeals are extremely rare under the defined reopen channel) Durable clearance ratio (90-day rolling): average ≈0.88 Unresolved stock grows from ~1 case at the start of the window to >3,360 by August 2025 90th-percentile unresolved age reaches ~114 days Regime diagnosis: “Busy but stuck” — closures are durable, but durable capacity lags arrivals, producing sustained backlog growth These are descriptive measurements under the locked definitions, not causal claims. 8. Limitations Documented limitations include: Monthly aggregation suppresses within-month dynamics. Appeal timing is approximated to the extent supported by available status history. Duplicate records and dataset hygiene issues are handled as described in README. Reopen detection is limited to the defined reopen channel (“Requestor Appeal Received”); other dispute mechanisms not represented in the schema are out of scope for Tier 1. 9. How to cite Espeleta, K. (2025). Tier 1 OCC Deployment Case Study: Washington, DC FOIA Request Processing (Aug 2020–Aug 2025). Zenodo. https://doi.org/10.5281/zenodo.18073749