Monitoring democratic institutions through public records
Structural Dimensions
What this shows#
This heatmap displays structural anomaly scores across all 14 monitored categories over time. Structural anomaly detection is fully deterministic and uses only document metadata — no text analysis. It compares the current week's document patterns against historical baselines across six dimensions to identify statistical departures from normal publishing activity.
Structural anomalies are descriptive, not evaluative. A spike in executive orders could reflect either an emergency response or a power grab — the structural layer identifies that something changed, not whether the change is concerning. Concern status is driven separately by AI document review.
How to read the heatmap#
Each cell represents one category in one week. Colors are based on z-scores, which measure how far a value deviates from the baseline mean in units of standard deviation. A z-score of 0 means the value matches the baseline exactly. A z-score of +2 means the value is 2 standard deviations above the baseline — an unusual departure.
- Blue cells (negative z-scores) — below-baseline activity
- Neutral cells (z-score near 0) — typical activity matching the baseline
- Red cells (positive z-scores) — above-baseline activity
- Striped cells — no data available for that category-week
When anomalies cluster across multiple categories in the same week (a vertical red stripe), it suggests an external event affecting government-wide publishing rather than a category-specific pattern.
Dimensions#
Use the pill selector above the heatmap to view individual dimensions or the weighted composite. Each dimension measures a different facet of document publishing patterns:
| Dimension | Weight | What it measures |
|---|---|---|
| Composite | — | Weighted average of all available dimension z-scores. This is the default view. |
| Volume | 22% | Document count relative to baseline mean. High z-scores indicate unusually high or low publishing volume. |
| Type | 18% | Distribution of document types (rules, notices, executive orders, etc.) compared to baseline, measured by Jensen-Shannon divergence. |
| Functional | 22% | Distribution of institutional functions (rulemaking, enforcement, personnel actions, etc.) compared to baseline. |
| Agency | 13% | Distribution of publishing agencies compared to baseline. Detects shifts in which agencies are active. |
| Tempo | 12% | Variance in daily publication counts within the week. High values suggest bursty publishing patterns. |
| Convergence | 13% | Ratio of government-origin sources to news/rhetoric sources. Shifts may indicate changing information dynamics. |
Weights are defined in scoring-config.ts and sum to 1.0. Categories with fewer than 10 documents in a week receive dampened scores to reduce noise.
Baseline#
Z-scores are computed using cycle-year matching: Year 1 of the current administration is compared against Year 1 of the Biden administration, Year 2 against Year 2, and so on. This accounts for predictable seasonal patterns — first-year administrations systematically differ from second-year administrations (higher executive order volume, more personnel changes), so comparing like-to-like avoids false positives.
The system maintains four historical baselines: Biden 2021 (Year 1), Biden 2022 (Year 2), Trump 2017 (Year 1), and Trump 2018 (Year 2). The Biden baselines are the primary reference; Trump baselines are available for cross-administration comparison in the CSV export.
A composite z-score above 2.5 (or a category-specific override for thin categories) is flagged as structurally anomalous. This threshold is shown in cell tooltips. For full methodology details, see the methodology page.
Loading structural data…