Statistics
AuraScore 77/100

Municipal Social Benefits Allocation Disparity and Statistical Fairness Audit Checklist

Evaluate algorithmic eligibility scoring models and municipal aid allocation systems for disparate impact and statistical bias.

Use this checklist when auditing automated or statistical scoring mechanisms used by local governments and social services to distribute benefits. It enables data protection officers and public sector statisticians to evaluate fairness metrics, false negative disparities, and demographic parity prior to policy deployment.

Template

Role: Lead Quantitative Policy Analyst & Algorithmic Equity Auditor

Context

  • Benefit Program: {{program_entitlement_name}}
  • Protected Attributes: {{protected_demographic_classes}}
  • Cutoff Threshold: {{decision_rule_threshold}}
  • Historical Baseline: {{historical_acceptance_rates}}
  • Equity Formalism: {{fairness_metric_criteria}}
  • Bias Risks: {{data_censoring_risks}}

Task

Construct a comprehensive, rigorous statistical fairness checklist to evaluate the eligibility scoring algorithm for {{program_entitlement_name}}, ensuring compliance with equity requirements across {{protected_demographic_classes}}.

Method

  1. Disaggregate applicant distribution and scoring outcomes across {{protected_demographic_classes}}.
  2. Quantify disparate impact ratios against baseline rates defined in {{historical_acceptance_rates}} (e.g., four-fifths rule).
  3. Test the chosen {{fairness_metric_criteria}} (e.g., false negative rate parity across sub-groups).
  4. Evaluate sensitivity of the classification cutoff at {{decision_rule_threshold}} across demographic strata.
  5. Audit feature importance to identify unapproved proxy variables correlated with protected attributes.
  6. Investigate survivor bias, unobserved selection effects, and {{data_censoring_risks}} in the training data.
  7. Compute intersectional disparity metrics combining multiple protected characteristics simultaneously.
  8. Establish continuous post-deployment statistical drift monitoring protocols.

Constraints

  • Checklist MUST enforce concrete statistical disparity tolerances (e.g., adverse impact ratio limits).
  • MUST NOT permit deployment if false negative disparities for vulnerable populations exceed stated safety limits.
  • Must provide explicit guidance for trade-offs between predictive accuracy and demographic fairness.
  • Use clear tabular format requirements for diagnostic documentation.

Output format

  • Structured markdown quality checklist with five functional categories:
    • Phase 1: Input Data & Proxy Variable Screening (4-5 checklist items)
    • Phase 2: Statistical Parity & Error Rate Disparity Checks (5-6 checklist items)
    • Phase 3: Threshold Sensitivity & Frontier Analysis (3-4 checklist items)
    • Phase 4: Long-term Monitoring & Drift Detection (3-4 checklist items)
  • Concluding section: Algorithmic Equity Determination Matrix with quantitative thresholds and remediation actions.

Self-review

  1. Are all protected groups in {{protected_demographic_classes}} specifically evaluated in the error-rate checks?
  2. Does the checklist explicitly address the specific metrics defined in {{fairness_metric_criteria}}?
  3. Are proxy leakage and {{data_censoring_risks}} covered in the pre-deployment tests?
AuraScore breakdown
77/100Provisional
Instruction clarity15/15 · Strong

Explicit role, a named task, and discrete steps the model can follow.

Context architecture12/12 · Strong

Background, inputs and variables the model needs before it starts.

Constraint engineering8/12 · Adequate

Hard boundaries — what the model must and must not do.

Output specification6/14 · Thin

A named, field-level shape for the response.

Reasoning structure10/10 · Strong

Ordered work items that force analysis before an answer.

Model compatibility10/10 · Strong

Length and structure that travel across frontier models.

Token efficiency5/10 · Thin

Signal density — instruction weight without padding.

Reusability7/7 · Strong

Documented variables so the scaffold adapts to new inputs.

Robustness3/5 · Adequate

Quality bar, assumptions and behaviour when inputs are thin.

Observed performance1/5 · Thin

How much real usage the template has behind it.

data-analytics
data-statistics
public-sector-nonprofit
algorithmic fairness
disparate impact
public sector