Rate Design Equity Impact Score (RDEIS)

GMI-EU-044Energy & UtilitiesInfrastructure-Dependent

Definition

Measures whether AI-optimized rate designs disproportionately burden specific customer segments (low-income, renters, specific geographies) relative to their share of system costs.

The Measurement Gap

Traditional rate case evaluation applies cost-of-service analysis to determine whether each rate class pays its allocated share of system costs, and revenue sufficiency analysis to confirm the utility's revenue requirement is met. Both tests are necessary for rate case approval. Neither tests whether the resulting bill impacts fall proportionally across income levels.

This gap has existed in rate design for decades, but AI optimization makes it consequential in a new way. When a human rate design team proposes a new TOU structure, the distributional implications are typically discussed qualitatively in the stakeholder process. When an AI optimizer generates 500 candidate rate structures and selects the one that maximizes a defined objective, the distributional implications of every candidate are invisible unless RDEIS is computed as a filter.

The operational decision RDEIS enables is the equity gate: no rate design produced by the AI optimizer proceeds to filing without a RDEIS check. If RDEIS exceeds 2.0 (low-income customers bear twice the proportional burden of high-income customers), the rate design must be revised: either by adding an equity constraint to the optimizer or by adding a targeted low-income credit structure to the proposed rate.

What goes wrong without it: the utility files a rate designed by an optimizer, consumer advocates compute the income-distribution impact using publicly available data, the commission holds a hearing focused on equity, and the rate is remanded for redesign. The RDEIS check, had it been performed before filing, would have cost an analyst two weeks. The rate case delay cost the utility months of carrying costs and a damaged regulatory relationship. This metric requires Census income geocoding, hence Infrastructure-Dependent, but that data is publicly available and the geocoding work, done once, serves every future rate case.

Formula

RDEIS = Bill_impact_pct_Q1 / Bill_impact_pct_Q5

Variables:

  • Bill_impact_pct_Q1: Average annual bill change (proposed rate minus current rate) as a percentage of estimated household income for the lowest-income quintile of customers (%). Source: CIS billing simulation, Census ACS income data geocoded to service address
  • Bill_impact_pct_Q5: Average annual bill change as a percentage of estimated household income for the highest-income quintile (%). Source: Same
  • Income quintile boundaries: Defined by Census ACS median household income at the Census tract level, matched to customer service addresses in GIS. Source: GIS, CIS, Census ACS

Calculation Frequency: Per rate design iteration (pre-filing); not a continuous operational metric
Output: Ratio, dimensionless. Values above 1.0 indicate the rate change is regressive (low-income customers bear proportionally more burden). Values below 1.0 indicate the rate is progressive. Values at 1.0 indicate proportional burden. RDEIS above 2.0 represents a significant regulatory exposure in most U.S. jurisdictions: the lowest-income customers are paying twice the proportional burden of the highest-income customers. Track RDEIS by geography as well as income; some rate designs are regressive by location even when aggregate income analysis appears acceptable.

Proposed formula, pending validation pilots.

At a Glance

Category
Customer AI Fairness and Revenue Integrity

Decision Supported

Will this rate design create an equity problem that becomes a regulatory or political liability, and which customer segments bear disproportionate burden?

← Dispute Resolution Error Propagation Rate (DREPR)Rate Recommendation Multi-Objective Alignment Score (RRMOAS) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment