Equity Impact Assessment¶
Assessment — instantiates Structural Harm Mapping
An ex-ante test of a proposed policy, program, or design that estimates how its benefits and harms would fall across groups, measured against the alternative of not doing it.
An Equity Impact Assessment is a forward-looking test applied to a proposed change — a policy, program, budget, or design — that estimates how its benefits and burdens would land across different groups before it is adopted, judged against an explicit counterfactual: the status quo, or an alternative design. Its defining feature is that it is ex ante, comparative, and analytic. It is a structured disparate-impact check a proposal must pass, producing a finding — "this change helps group A and quietly costs group B" — that decision-makers weigh before they commit. It evaluates a hypothetical, not a route people already walk or a system already running.
Example¶
A platform plans to expand an automated system that suspends accounts flagged for policy violations. An equity impact assessment runs before launch. It estimates suspension rates across language communities and compares them against a counterfactual — human-only review, and the current suspension rate. It finds the classifier's training data under-represents a regional dialect, so accounts posting in that language would be suspended at a markedly higher rate and would clear appeals less often, because the appeal flow is English-first. Illustratively, the modeled disparity is several-fold and concentrated among users with the least recourse. The assessment does not build the fix; it surfaces the disparate impact — as a difference against the alternative, not an absolute number — so the launch can be gated on it or the classifier retrained first.
How it works¶
What distinguishes it is that it assesses a proposal against an alternative, before the fact:
- Assess a proposal, not a running system. The object is a change under consideration, evaluated while it can still be altered cheaply.
- Model effects by group. Estimate how benefits and burdens fall across the relevant protected or vulnerable cuts.
- Anchor on a counterfactual. Define the alternative (do nothing, or a different design) so that "impact" means impact relative to that alternative, not a raw level.
- Report a distributional finding. Deliver a go / adjust / block input keyed to the size and concentration of the disparity.
Tuning parameters¶
- Group taxonomy — which protected or vulnerable cuts you assess. A group you omit is a disparity you cannot see.
- Counterfactual choice — status quo, do-nothing, or the best alternative design. The counterfactual you pick largely decides the verdict.
- Effect horizon — first-order effects only, or downstream and behavioral ones. Longer horizons catch second-round harms but add uncertainty.
- Evidence standard — a modeled estimate or a piloted measurement. Piloting is more credible and slower.
- Materiality threshold — how large a disparity trips a flag, set before the run so the bar isn't moved to fit the answer.
When it helps, and when it misleads¶
Its strength is catching uneven harm while the design is still on the drawing board — cheap to change and not yet defended — which is the logic behind formal disparate-impact review.[1] Its failure mode is the assessment as a compliance checkbox, filed after the decision is effectively made, so the finding can no longer alter anything. The classic misuse is choosing a flattering counterfactual — comparing to a strawman rather than the real alternative — so that any change looks equitable. The discipline is to fix the counterfactual and the materiality threshold before running the assessment and to give a red finding real teeth: the authority to block or alter the launch, not merely to note a concern.
How it implements the components¶
burden_distribution_check— its core output: who gains and who bears the cost of the proposed change, resolved by group.counterfactual_comparison— it measures impact against an explicit alternative, so the finding is a difference attributable to the change, not a standing level.
It tests a proposal in advance; it does not track whether harm actually fell once a remedy is live — harm_reduction_metric is the Harm Reduction Dashboard's — nor does it convene affected people to speak for themselves: affected_voice_and_evidence_check belongs to the Remedy Co-Design Workshop and the standing Affected-Party Review Panel.
Related¶
- Instantiates: Structural Harm Mapping — it supplies the ex-ante disparate-impact check that gates a proposed change on its distributional effects.
- Sibling mechanisms: Harm Reduction Dashboard · Access Pathway Map · Institutional Barrier Review · Policy Pathway Analysis · Structural Audit · Systems Harm Analysis · Social Determinants Mapping · Remedy Co-Design Workshop · Affected-Party Review Panel
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Equity Impact Assessment operates as a bounded evaluation of existing evidence or work that produces a finding or disposition because it an ex-ante test of a proposed policy, program, or design that estimates how its benefits and harms would fall across groups, measured against the alternative of not doing it.
Independent corroboration: The frozen evidence defines Equity Impact Assessment as 'An ex-ante test of a proposed policy, program, or design that estimates how its benefits and harms would fall across groups, measured against the alternative of not doing it', so its operative form is Assessment, Review & Assurance.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Public Administration & Policy
Origin pattern: Single lineage
Present-day reach: Multi-domain
Rationale: Equity-centered policy analysis cohered ex-ante assessment of how a proposed policy's benefits and burdens compare across groups and alternatives.
Related originating lineages:
- Law & Governance — Disparate-impact and equality review supply protected-group and counterfactual tests.
- Sociology & Anthropology — Structural inequality analysis supplies causal pathways by which formally neutral changes distribute harm unevenly.
Review resolution: The current reviewers agree that public_administration_policy is primary. For the reported differences (alternate_origin_disagreement, origin_mode_disagreement), the evidence supports single_lineage, multi_domain, and law_governance, sociology_anthropology; these choices preserve materially formative origins without conflating later domain reach.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Supreme Court of the United States. Griggs v. Duke Power Co., 401 U.S. 424. United States Reports 401 U.S. 424 (1971). Establishes that a facially neutral practice may be unlawful when it produces discriminatory effects. registry ↩