Unveiling hidden energy poverty using the energy equity gap

The study addresses a key blind spot in energy poverty measurement: households that deliberately limit energy consumption to reduce financial strain and thus may not be captured by traditional income-based metrics such as energy burden. In developed regions with near-universal access to electricity, energy poverty manifests as difficulty maintaining comfortable indoor temperatures, high risk of shutoffs, excessive energy burdens, and barriers to adopting efficient technologies. The authors hypothesize that low-income households delay using cooling, enduring higher outdoor temperatures before turning on air conditioning to save money, which places them at risk of heat-related illness and adverse indoor environmental conditions. The purpose is to introduce a behavior-based metric—the energy equity gap (EEG)—that quantifies disparities in temperature-triggered cooling behavior between income groups, thereby unveiling hidden energy poverty and complementing existing financial measures. The study focuses on a high-heat region (Arizona) where cooling-related health risks are significant.

The paper categorizes energy poverty metrics along two axes: primary vs. secondary (household-level data vs. aggregated/derived) and absolute vs. relative (threshold-based vs. comparative). Examples include: primary-absolute (energy burden thresholds such as 10%), primary-relative (self-reported surveys on thermal comfort and deprivation), secondary-absolute (composite indices combining income, demographics, housing), and secondary-relative (national access-consumption matrices tracking fuel transitions). Existing income-based measures are widely used but have limitations: sensitivity to energy prices, lack of differentiation between gross and disposable income, omission of indoor comfort, and insufficient consideration of local cost of living. Survey-based metrics elicit perceptions but are time-intensive and hard to compare across households. Most approaches focus on equality (uniform thresholds) rather than equity (context-sensitive ability to achieve comfort). There is a gap in metrics that quantify the degree of energy a household forgoes (energy-limiting behavior) in developed contexts, motivating a data-driven, behavior-sensitive measure.

Data: Hourly electricity consumption for approximately 6000 Arizona households from May 2015 to April 2019 provided by Salt River Project, plus billing plan information and a 2017 residential equipment and technology survey with sociodemographics and dwelling characteristics. Hourly data were aggregated to daily consumption and paired with daily average outdoor temperatures (WeatherForYou.com). Electricity pricing plan rules were incorporated as a uniform weighted average daily price by plan. Study years are reported as annual windows (e.g., 2015–2016).

Inflection temperature estimation: For each household-year, daily electricity consumption is modeled as a quadratic function of daily average outdoor temperature with controls for electricity price (by billing plan and season), holiday indicator, day-of-week fixed effects, and month fixed effects. The convex quadratic relationship (median R^2 ≈ 0.8) captures minimal electricity use at a temperature between heating and cooling seasons. The inflection temperature is defined as the outdoor temperature at the minimum of the fitted quadratic (i.e., the onset of cooling use in a warm-dry climate context). Outliers with inflection temperatures outside 30–120 °F were removed (0.5% to 1.6% per year). Not all households have complete data across all years due to service changes.

Energy Equity Gap (EEG): After estimating household-level inflection temperatures, households are grouped by income (eight brackets). For each year, the EEG is the difference between the maximum and minimum median inflection temperatures across income groups, quantifying relative disparities in cooling onset linked to budget constraints. Statistical significance of between-group differences was tested using Mood’s Median tests for income, ethnicity, and age strata.

Tiered risk classification: Using the median inflection temperature of the highest income group as the ideal benchmark, households are categorized by risk: low-risk if within one to two EEGs above the ideal; energy insecure if between two EEGs and 78 °F; energy poor if above 78 °F. The 78 °F threshold is based on recommended indoor setpoints for government buildings and utility guidance, used as a conservative health- and comfort-relevant benchmark for outdoor conditions implying needed cooling.

Traditional energy burden metric: Energy burden is computed as annual electricity expenditure over midpoint of reported income bracket, using daily consumption and average price for the household’s plan. Threshold for insecurity is set at ≥10% of income (EB10). The study compares EEG-based tiers with EB10 and eligibility for LIHEAP and WAP (based on program income thresholds).

Additional analyses: EEG and inflection temperature distributions were examined across ethnicity and head-of-household age groups. Trends in EEG were related to changes in cooling degree days and average residential retail electricity prices year-to-year. Consideration of residence type, age, and size was limited due to multicollinearity with income.

• The Energy Equity Gap (EEG) reveals systematic energy-limiting behavior: In the Arizona metropolitan region, EEG ranges 4.7–7.5 °F (2.6–4.2 °C) across years, indicating low-income groups endure higher outdoor temperatures before initiating cooling compared to high-income groups.
• Yearly EEG values and samples: 2015–2016 EEG = 5.9 °F (N=4577); 2016–2017 EEG = 4.7 °F (N=4522); 2017–2018 EEG = 5.2 °F (N=3852); 2018–2019 EEG = 7.5 °F (N=2650). Differences in median inflection temperatures across income groups are statistically significant (Mood’s Median tests).
• Tiered counts (2015–2016): 86 energy-poor (2nd tier, >78 °F) and 214 energy-insecure (1st tier) households. Across years, EEG-based classifications identify larger numbers of energy-poor households than the income-based EB10 metric.
• Comparison with energy burden: EB10 identified 141 (2015–2016), 135 (2016–2017), 111 (2017–2018), and 88 (2018–2019) households as energy-insecure, with very limited overlap (≤20) with EEG tiers; only three households overlapped in 2015–2016 between EEG and EB10 categories cited in the abstract. EB10 misses over 95% of high inflection temperature households at greater heat exposure risk. Example from Fig. 7: in 2015–2016, 2.7% of households exceeded the 10% energy burden threshold.
• Program eligibility gaps: Among EEG 2nd-tier households, 48–72 per year are not eligible for LIHEAP; 29–53 per year are not eligible for WAP (despite high inflection temperatures), indicating assistance gaps.
• Temporal dynamics: EEG narrowed by 20.3% from year 1 to 2, then widened by 10.6% (year 2 to 3) and 44.2% (year 3 to 4). Changes in cooling degree days (+3.6%, +2.5%, −5.2%) and residential electricity prices (+2.4%, +2.7%, −2.7%) correlate with subsequent EEG changes, suggesting delayed price and climate sensitivity among low-income households.
• Demographic disparities: Black households have the highest overall median inflection temperatures and large EEGs, indicating disproportionate adverse impacts and high inequity; Asian households show lower median inflection temperatures but wide EEGs (high within-group income disparities). Younger age groups (18–24, 25–34) exhibit sharp increases in EEG in later years (e.g., >14 °F increase for 18–24 from 2018 to 2019), while older groups show relatively stable EEG. Income is strongly associated with inflection temperatures across all years.
• Housing context: Low-income households are more likely to live in older residences, potentially increasing cooling needs and financial strain.
• Modeling performance: The quadratic temperature response achieves median R^2 ≈ 0.8; outliers (<30 °F or >120 °F inflection) are rare (≤1.6% per year).

Findings confirm the hypothesis that low-income households delay cooling initiation to alleviate financial burden, resulting in higher inflection temperatures and increased risk of heat-related illness and adverse indoor environmental conditions. The Energy Equity Gap captures a behavioral dimension of energy poverty that income-based metrics overlook, especially for households that keep energy spending low by reducing consumption rather than by meeting needs. EEG trends respond to electricity price and climate variations, suggesting that low-income households reduce cooling use in response to higher prices and warmer conditions, exacerbating inequity over time. Demographic analyses show that minority and younger populations can experience elevated EEGs, pointing to compounding vulnerabilities. The implications are that energy poverty assessments should combine EEG with energy burden to capture both financial strain in meeting energy needs and behavioral under-consumption that increases health risk. The proposed tier system enables prioritizing households with high inflection temperatures (e.g., above 78 °F benchmark) for targeted assistance and weatherization, including sliding-scale interventions. Policymakers can track EEG over time to assess equity progress and tailor interventions by income, ethnicity, and age group. Incorporating EEG into program eligibility could address gaps where households are not classified as low-income yet experience hazardous cooling limitations.

This work introduces the Energy Equity Gap, a primary metric that quantifies disparities in cooling onset behavior across income groups, revealing hidden energy poverty in a high-heat U.S. region. By defining household inflection temperatures and comparing medians across income brackets, the EEG complements traditional energy burden metrics and uncovers households at risk due to energy-limiting behavior. A tiered classification anchored to the highest-income group’s median inflection temperature and a 78 °F benchmark enables practical targeting of energy-insecure and energy-poor households. The study demonstrates that price and climate changes can widen EEG, and that demographic subgroups (e.g., Black and younger households) may experience pronounced inequities. Future research should integrate heating fuels (gas, oil) for cold climates, incorporate richer housing efficiency data (e.g., insulation, windows, orientation), and, where feasible, measure indoor temperatures to refine risk assessments. Expanding EEG applications across regions and coupling with income-based thresholds could support equity-centered policies, sliding-scale weatherization, and adaptive strategies for heat resilience.

• Fuel coverage: Heating energy from gas or oil was not included; approximately 40% of households likely use non-electric heating. Although the region is cooling-dominated, excluding heating may understate winter-side dynamics.
• Building characteristics: Detailed housing efficiency parameters (e.g., insulation, window count, wall construction, orientation) were unavailable, limiting control for structural drivers of cooling demand.
• Indoor conditions: No thermostat or indoor temperature data were available; true comfort preferences and indoor heat exposure could not be directly measured. Occupants’ thermostat settings reflect economic constraints and conservation habits, not necessarily preferred comfort.
• Urban heat island: Low-income households may experience higher outdoor temperatures due to less vegetation and shade, potentially widening actual disparities relative to estimates based on regional averages.
• Multicollinearity: Residence age/type/size correlate with income; excluding them reduces overfitting but may omit important determinants.
• Generalizability: Results are from a single, high-heat metropolitan region; adaptation to colder climates requires integrating heating fuels and season-specific modeling.