How Did Utility Shutoff Procedures and Household Shutoff Outcomes Change During the Covid-19 Pandemic? Assessing Sociodemographic and Geographic Variation Across the United States

Abstract

Utility shutoffs due to an inability to pay can have severe consequences for the health and well-being of those who experience them, especially during a public health crisis. Starting in March 2020, many states and utilities across the United States instituted utility shutoff moratoria for the first time in response to the COVID-19 pandemic. This new policy approach was often based on the expressly stated logic that these services were essential for public health. Recent research indicates that these moratoria were associated with significant reductions in COVID-19 infections and deaths, though with some local variation. Yet nationally, we still know relatively little about the scale of these protections at the household level, including which households received protection from utility shutoffs during and following the pandemic and how these patterns reflect relatively well-documented sociodemographic disparities in utility debt and shutoffs before the pandemic. To help address these gaps, we used nationally representative, Census Bureau-managed household survey data from 2017 and 2021 to assess sociodemographic and geographic variation in utility shutoff notification and actualization before and after the COVID-19 pandemic. We found that both shutoff notices and realized shutoffs decreased from 2017 to 2021, as did the ratio between those who received a shutoff notice and subsequently experienced a shutoff, suggesting a likely change in utility-level policy. Results from multivariate analysis indicate that household income and monthly utility amounts were significant predictors of experiencing an actual utility shutoff in 2017 but not in 2021. However, the relative risk of experiencing an actual utility shutoff increased for Black (as compared to white) households between 2017 and 2021. Overall, these findings indicate some promising changes in household shutoff prevalence and procedures reflecting follow-on effects of the moratoria but also highlight the persistence of uneven experiences and outcomes, which merit additional policy attention to achieve environmental justice.

BACKGROUND

Lower-income and racially marginalized households tend to have higher utility service debt burdens and an increased risk of experiencing utility shutoffs or disconnections due to nonpayment.^1,2,3,4 This is a serious problem given that utility service shutoffs due to an inability to pay the utility bill can have severe consequences for the health and well-being of affected households. For example, a lack of adequate summer cooling and winter heating puts residents at an increased risk of experiencing a range of illnesses or death. Children, older adults, and persons with chronic health conditions are particularly vulnerable to the negative implications of exposure to extreme temperatures.^5,6,7 Similarly, insufficient access to potable water is associated with dehydration and exposure to contaminants.^8,9,10 These consequences are exacerbated during a pandemic, when water and energy for adequate household cleaning and personal hygiene are especially crucial for limiting the spread of contagion.^8,9,11 Even the unrealized threat of a utility shutoff can place economic, broader opportunity, and stress burdens on residents, as they may be forced to make trade-offs among basic household needs such as utility costs, rent and mortgage payments, transportation, and food expenses.^12,13 Further, late fees levied because an inability to pay utility bills on time can deepen financial distress for already vulnerable households.²

At the height of the COVID-19 pandemic, public recognition of these implications was recognized in written policy for the first time at scale in the United States. The continuation of utility services was explicitly characterized as essential for public health in the language of many state and local-level utility shutoff moratoria. Relatedly, in March 2020, the federal government used its public health emergency authority to institute a national moratorium on evictions in the UnitedStates as part of the Coronavirus Aid, Relief, and Economic Security Act. However, specific policy action to stop utility shutoffs during the pandemic was left to the discretion of state and local governments, resulting in a patchwork of utility shutoff moratoria, varied by utility service and duration of protection, across the country.¹⁴

Although still sparse, there has been some research on both the design and general impacts of variation in local and state decision-making about utility service shutoff policies and protections.¹⁵ In both the energy and water sectors, state and local policies vary widely depending on the political-economic context and by utility institutional type.^16,17 The COVID-19 pandemic exacerbated these differential trends.^11,18,19 Recent research indicates that the utility shutoff moratoria instituted in the UnitedStates during the pandemic reduced illness and death, though also with significant county-level variation. Further, metropolitan areas with large proportions of lower-income energy burdens tended to have lower protections against utility shutoffs during the pandemic.¹³ States with mandatory moratoria on utility shutoffs witnessed significant drops in utility disconnections, though Black and Hispanic households still faced an increased risk of experiencing a shutoff.²⁰ Yet, beyond a few narrow cases, we still know relatively little about which U.S. households received protection from utility shutoffs before and after the height of pandemic protections, as well as the relationship between shutoffs and other system and political-economic characteristics. Our study thus focuses on household-level impacts across the United States

To help address this research gap, we used nationally representative survey data to analyze sociodemographic and geographic variation in utility shutoffs in the UnitedStates in 2017 (pre-COVID) and 2021 (“post”-COVID). Our study was guided by two research questions: (1) How did utility service shutoff procedures and household outcomes change during the COVID-19 Pandemic? (2) Were changes in utility shutoff procedures and household outcomes experienced evenly across sociodemographic groups and geographic regions?

METHODS

We used data from the 2017 and 2021 waves of the American Housing Survey (AHS) to capture the distribution of utility shutoffs among U.S. households in each period. The AHS is a nationally representative biennial survey sponsored by the U.S. Department of Housing and Urban Development and conducted by the U.S. Census Bureau. It is designed to collect data on a wide range of factors associated with the nation’s housing stock and resident households. The AHS sample has been redrawn twice (in 1985 and 2015) since the first AHS survey effort in 1973. During both redraws, significant changes were also made to the survey instrument and publicly available measures. However, the sample and available measures in the 2017 and 2021 waves are identical, providing a unique data source for analysis of changes in utility shutoff notification and actualization before and during the height of the pandemic.

The AHS national public use files (PUFs) include probability weights to ensure that the data are representative of the U.S. housing stock and a set of 160 replicate weights to account for variance in single survey estimates. We used the svyset command in Stata 17 to identify the survey design before analyzing the data, following documentation provided by the U.S. Census Bureau to ensure that the weights were applied correctly.²¹

The three-category outcome variable in our study captures utility service shutoff notification and actualization for unique survey respondent households. Regarding timeframe, AHS participants were asked about utility shutoff notification and actualization within the last three months of the survey. Participants had the following response options: (1) received “no notice,” (2) “received a notice, but did not have utilities not shut off,” or (3) “received a notice, and had utilities shut off.” Because our outcome variable has three nominal categories, we relied on multinomial logistic regression for our multivariate analysis. Multinomial logistic regression extends standard binary logistic regression—which is limited to two outcome categories—to allow for three or more outcome categories. The following equation illustrates the multinomial logistic regression model: $\ln (\frac{π_{j}}{π_{1}}) = β_{j 0} + β_{j 1} x_{1} + \dots + β_{j p} x_{p} .$ (1)

The ratio on the left-hand side of the equation represents the probability of being in category j (i.e., “received a notice but did not have utilities shut off” or “received a notice, and had utilities shut off”) versus the probability of being in the baseline category (i.e., “no notice”), while $β_{j 0}$ represents the intercept and $β_{j 1} x_{1} + \dots + β_{j p} x_{p}$ represents the covariates included in the model.

We estimated two separate models for 2017 and 2021. For both study periods, we used the “no notice” category as the base outcome.22 We report exponentiated coefficients, which provide relative risk ratios (RRRs) compared with the base outcome.23 We also show the balanced repeated replicate standard errors for each estimate.

In the United States., it is nearly (if not actually) a universal requirement that utilities send a written notice to customers overdue on their bill payments and give them additional time to pay before shutting them off from service. As just one example, in 2018, Senate Bill 998 in California standardized water utility shutoff policy due to nonpayment for both public- and investor-owned water utilities.²⁴ Under this law, customers of water utilities cannot receive a notice of disconnection earlier than 60 days following bill payment due date, and an additional notice at least seven days from the disconnection date. However, minimum standards for the interval between shutoff notification and actualization again vary by state and utility type, and enforcement is spotty.²⁵

The explanatory (independent) variables in our multivariate models included a range of measures encompassing the social, economic, and housing characteristics identified in previous research as important for understanding the justice and equity dimensions of utility shutoffs, and available in the AHS data. The social characteristics included the householder’s age, degree of formal education (as bachelor’s degree or higher versus less than a bachelor’s degree), and sex (as female versus male). Regarding race/ethnicity, we included discrete categories for non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Asian, and non-Hispanic Hawaiian/Pacific Islander householders. We also included an additional discrete category for multiracial householders who did not identify as Hispanic.

The primary economic characteristic in our analysis was average monthly household income, which we created by dividing the annual household income measure provided in the AHS data by 12 (we used average monthly household income in $1,000 units in the multivariate analysis to facilitate a more parsimonious interpretation of the results). In addition, to capture non-monetary public assistance and potential access to broader social protections that may have assisted households in paying their utility service bills, we included a variable indicating whether households received Supplemental Nutrition Assistance Program (SNAP) benefits within the last 12 months. We also included two measures of monthly household costs: the first capturing monthly mortgage, rent, and lot rent costs, and the second capturing the total monthly amount spent on utilities (both in $100 units in the multivariate models).

Furthermore, we included four dichotomous variables capturing whether each household pays directly for specific types of utilities (i.e., electricity; gas, oil, and other fuels; trash service; and water). This is important because each service is typically handled by a different utility provider, including standards for shutoffs due to nonpayment, and for some utility services, direct utility bill payment is circumvented by tenure status (i.e., renters often do not directly pay for all of their utilities).²⁶ Relatedly, to account for housing characteristics that also influence utility payment arrangements, we included tenure (i.e., housing owned, rented, or occupied without payment of rent) and building type (i.e., one-family house, detached; one-family house, attached; mobile home or trailer; or apartment).

Although limited, our analysis includes all available geographic location measures available in the post-2015 AHS data. Unfortunately, we were unable to include state fixed effects to account for variation in state-level policy and culture because a state-level identifier is not available in the AHS PUFs. However, we were able to include a nine-category Census Division variable in our analysis to help account for geographic variation (See Supplemental Figure 1 for map of Census Divisions). Meanwhile, the original Census Division and the Management and Budget defined core-based statistical area code) variable has 17 categories: one for each of the 15 largest metropolitan areas, one for all other metropolitan areas, and one for non-metropolitan areas. For parsimony, we recoded this variable as a dichotomous measure capturing location within or outside of a metropolitan area.

We included all households available in the AHS except for those living in boat/RV/van, etc. We excluded these unconventional housing types because they likely lack access to stable utility services and thus are not subject to shutoff in the same way (this reduced the number of weighted cases by less than 0.1% in both study periods). We also assessed cases coded as “not reported” for any of the included variables as missing data. Because the overall deletion rate due to variable missingness was modest for 2017 and 2021 (5.5% and 4.5%, respectively), we used listwise deletion to handle the missing data in our multivariate regression models.²⁷

RESULTS

Descriptive statistics and bivariate results

Descriptive statistics for all the variables used in the study are shown in Table 1. As expected, given the landmark, albeit patchwork, utility shutoff moratoria, utility shutoff notifications and actualization among the U.S. households both decreased between 2017 and 2021. In both periods, unactualized shutoff notices are over ten times more prevalent than realized shutoffs and were almost equally common among lower- and higher-income households.

Table 1.

Weighted Descriptive Summary Statistics for All Study Variables (Weighted Counts, Percentage Distribution, and Means)

		2017		2021
Variable	Variable categories	Weighted count	%	Weighted count	%
Utility Shutoff Notification and Actualization	1: No notice	98,809,912	84.4	109,728,220	88.0
	2: Received notice, utilities not shut off	17,124,642	14.6	14,258,161	11.4
	3: Received notice, utilities shut off	1,200,030	1.0	731,292	0.6
Age of Householder (years)	Continuous (mean shown)	52.1		52.5
Formal Education of Householder	0: Less than bachelor’s degree	79,514,291	65.5	80,903,612	63.0
Formal Education of Householder	1: Bachelor’s degree or higher	41,970,628	34.5	47,538,336	37.0
Sex of Householder	0: Male	62,833,924	51.7	66,819,234	52.0
Sex of Householder	1: Female	58,650,996	48.3	61,622,714	48.0
Race/Ethnicity of Householder	1: White	80,500,159	66.3	82,935,799	64.6
	2: Black	15,793,765	13.0	17,089,250	13.3
	3: Hispanic	16,484,961	13.6	18,425,770	14.3
	4: American Indian, Alaska Native	1,040,098	0.9	1,070,517	0.8
	5: Asian	5,781,415	4.8	6,802,756	5.3
	6: Hawaiian, Pacific Islander	342,469	0.3	395,137	0.3
	7: Two or more races	1,542,052	1.3	1,722,719	1.3
Avg. Monthly Household Income ($)	Continuous (mean shown)	6760.85		7378.60
SNAP Benefits (Last 12 Mo.)	0: Did not receive SNAP benefits (ref)	105,933,280	90.4	112,555,783	90.2
	1: Received SNAP benefits	11,253,781	9.6	12,282,334	9.8
Monthly Rent/Mortgage Amount ($)	Continuous (mean shown)	827.06		961.21
Monthly Utility Amount ($)	Continuous (mean shown)	212.69		227.76
Utilities Directly Paid for by Household	Electric (0: No, 1: Yes)	115,948,929	95.4	122,420,003	95.3
	Gas, oil, and other fuels (0: No, 1: Yes)	80,297,164	66.1	83,944,944	65.4
	Trash (0: No, 1: Yes)	56,769,022	46.7	65,298,279	50.8
	Water (0: No, 1: Yes)	55,371,286	45.6	62,314,398	48.5
Tenure	1: Owned	77,494,934	63.8	82,475,988	64.2
	2: Rented	42,028,854	34.6	44,037,514	34.3
	3: Occupied without payment of rent	1,961,132	1.6	1,928,445	1.5
Building Type	1: One-family house, detached	76,832,535	63.2	81,744,238	63.6
	2: One-family house, attached	8,957,749	7.4	8,154,709	6.4
	3: Mobile home or trailer	6,727,424	5.5	6,730,665	5.2
	4: Apartment	28,967,212	23.8	31,812,336	24.8
Census Division	1: New England	5,809,266	4.8	6,105,762	4.8
	2: Middle Atlantic	16,012,348	13.2	16,633,229	12.9
	3: East North Central	18,516,170	15.2	19,217,215	15.0
	4: West North Central	8,526,758	7.0	8,821,503	6.9
	5: South Atlantic	23,961,528	19.7	25,625,855	20.0
	6: East South Central	7,416,733	6.1	7,855,895	6.1
	7: West South Central	14,058,223	11.6	15,324,566	11.9
	8: Mountain	8,905,686	7.3	9,695,128	7.6
	9: Pacific	18,278,207	15.0	19,162,794	14.9
Nonmetro vs. Metro	0: In a metropolitan area	102,884,406	84.7	109,536,501	85.3
Nonmetro vs. Metro	1: Not in a metropolitan area	18,600,513	15.3	18,905,446	14.7

SNAP, Supplemental Nutrition Assistance Program.

Reported utility shutoff notices dropped from 15.6% in 2017 to 12.0% of households in 2021, a 23.1% decrease. At the same time, realized shutoffs were less frequent in 2021, dropping from 1% to 0.6% of households, a 40.0% decrease. Furthermore, and perhaps most interestingly, the share of households that received a notice and subsequently had their utilities shut off (realized shutoffs) decreased from 6.5% to 4.9%, a 24.6% drop from 2017 to 2021.

We also examined the variation in utility shutoff notification and actualization between Census Division geographies across the two periods (Table 2). Based on a survey-adjusted test of independence, regional variation for both outcomes was statistically significant (p < 0.001). Most Census Divisions saw modest drops in utility shutoff notices between 2017 and 2021, with the largest change occurring in the East North Central Division (from 15.0% in 2017 to 10.1% in 2021, a 32.7% decrease). The exception was the New England division, where shutoff notices increased from 15.0% in 2017 to 16.5% in 2021 (a 10.0% change). Meanwhile, all Census Divisions experienced drops in realized shutoffs from 2017 to 2021. In contrast to the results for shutoff notices, the most drastic decreases in actual shutoffs were seen in the New England (90.0%) and Middle Atlantic (72.7%) Divisions, while the East South Central (10.0%) and Mountain (12.5%) Divisions saw the most modest reductions.

Table 2.

Weighted Percentages for Utility Shutoff Notification and Actualization by Census Division.

	2017			2021
	No notice	Received notice, utilities not shut off	Received notice, utilities shut off	No notice	Received notice, utilities not shut off	Received notice, utilities shut off
New England	84.0	15.0	1.0	83.4	16.5	0.1
Middle Atlantic	85.1	13.9	1.1	90.9	8.7	0.3
East North Central	84.0	15.0	0.9	89.2	10.1	0.7
West North Central	87.8	11.5	0.7	90.0	9.6	0.5
South Atlantic	82.5	16.5	1.0	87.2	12.1	0.8
East South Central	83.4	15.6	1.0	85.1	14.0	0.9
West South Central	82.7	15.7	1.7	85.9	13.3	0.8
Mountain	83.6	15.7	0.8	86.7	12.6	0.7
Pacific	87.0	12.1	0.9	89.3	10.4	0.3

2017 p value < 0.001, 2021 p value < 0.001.

Multivariate results

Next, we explored the variation in unrealized shutoff notices and actualized shutoffs within a multivariate regression framework. The regression results shown in Table 3 provide more nuanced insights, though they are broadly consistent with our descriptive findings. In 2017, higher-income households were both less likely to receive an unactualized shutoff notice and were less likely to experience an actual shutoff. In 2021, higher-income households were still less likely to receive an unactualized shutoff notice but had an equal likelihood of experiencing an actual shutoff.

Table 3.

Multinomial Logistic Regression Predicting Utility Shutoff Notification and Actualization among U.S. Households in 2017 and 2021 with ‘No Notice’ as the Base Outcome

Covariates	Received notice, utilities not shut off			Received notice, utilities shut off			Received notice, utilities not shut off			Received notice, utilities shut off
Covariates	RRR	(BRR SE)		RRR	(BRR SE)		RRR	(BRR SE)		RRR	(BRR SE)
Age of householder	0.998	(0.001)	^*	0.987	(0.003)	^***	0.998	(0.001)	^†	0.984	(0.006)	^**
Formal Education of Householder
Less than bachelor’s degree (ref)	—			—			—			—
Bachelor’s degree or higher	0.946	(0.033)		0.663	(0.104)	^**	0.911	(0.038)	^*	0.437	(0.087)	^***
Sex of Householder
Male (ref)	—			—			—			—
Female	1.007	(0.032)		1.029	(0.110)		1.007	(0.036)		0.892	(0.140)
Race/Ethnicity of Householder
White (ref)	—			—			—			—
Black	0.997	(0.062)		2.157	(0.319)	^***	1.055	(0.063)		2.671	(0.556)	^***
Hispanic	1.014	(0.043)		1.635	(0.232)	^***	0.991	(0.061)		1.497	(0.311)	†
American Indian, Alaska Native	1.382	(0.250)	^†	4.452	(1.573)	^***	1.004	(0.188)		0.334	(0.273)
Asian	0.841	(0.067)	^*	0.669	(0.226)		0.796	(0.075)	^*	0.580	(0.297)
Hawaiian, Pacific Islander	1.007	(0.296)		2.177	(1.461)		0.565	(0.163)	^*	2.304	(2.252)
Two or more races	1.231	(0.149)	^†	2.162	(0.692)	^*	1.206	(0.190)		2.184	(1.106)
Avg. Monthly Household Income ($1,000 seconds)	0.989	(0.003)	^***	0.933	(0.015)	^***	0.989	(0.003)	^***	0.990	(0.021)
SNAP Benefits (Last 12 Mo.)
Did not receive SNAP benefits (ref)	—			—			—			—
Received SNAP benefits	1.419	(0.064)	^***	2.930	(0.369)	^***	1.193	(0.063)	^***	1.798	(0.331)	^**
Monthly Rent/Mortgage Amount ($100 seconds)	0.994	(0.005)		1.001	(0.005)		0.999	(0.002)		0.995	(0.008)
Monthly Utility Amount ($100 seconds)	1.035	(0.018)	^†	1.139	(0.054)	^**	1.000	(0.015)		0.930	(0.075)
Utilities Directly Paid for by Household
Electric	1.613	(0.130)	^***	5.561	(3.553)	^**	1.948	(0.202)	^***	8.882	(8.219)	^*
Gas, oil, and other fuels	0.996	(0.043)		1.054	(0.138)		1.020	(0.052)		1.089	(0.194)
Trash	1.065	(0.045)		1.127	(0.127)		1.093	(0.045)	^*	1.209	(0.192)
Water	0.968	(0.034)		1.107	(0.113)		0.973	(0.037)		1.048	(0.198)
Tenure
Owned (ref)	—			—			—			—
Rented	1.009	(0.050)		2.278	(0.358)	^***	0.982	(0.055)		2.261	(0.482)	^***
Occupied without payment of rent	1.134	(0.132)		1.662	(0.694)		1.125	(0.190)		2.266	(1.129)
Building Type
One-family house, detached (ref)	—			—			—			—
One-family house, attached	1.051	(0.065)		0.985	(0.190)		0.943	(0.067)		1.023	(0.290)
Mobile home or trailer	1.131	(0.068)	^*	1.404	(0.265)	^†	1.152	(0.084)	^†	1.861	(0.378)	^**
Apartment	0.874	(0.053)	^*	0.653	(0.107)	^**	0.900	(0.050)	^†	0.578	(0.139)	^*
Census Division
New England (ref)	—			—			—			—
Middle Atlantic	0.910	(0.076)		0.836	(0.230)		0.477	(0.052)	^***	1.613	(0.736)
East North Central	0.924	(0.082)		0.630	(0.195)		0.516	(0.053)	^***	2.767	(1.098)	^*
West North Central	0.694	(0.077)	^***	0.512	(0.176)	^†	0.495	(0.060)	^***	1.954	(0.844)
South Atlantic	1.050	(0.093)		0.625	(0.172)	^†	0.645	(0.064)	^***	2.635	(1.065)	^*
East South Central	0.939	(0.122)		0.542	(0.212)		0.724	(0.091)	^*	2.688	(1.325)	^*
West South Central	0.954	(0.086)		0.958	(0.280)		0.690	(0.071)	^***	2.658	(1.131)	^*
Mountain	0.955	(0.099)		0.482	(0.152)	^*	0.674	(0.090)	^**	2.816	(1.195)	^*
Pacific	0.776	(0.060)	^***	0.613	(0.187)		0.577	(0.065)	^*	1.182	(0.547)
Nonmetro vs. Metro
In a metropolitan area (ref)	—			—			—			—
Not in a metropolitan area	0.864	(0.065)	^†	0.982	(0.153)		0.985	(0.081)		1.041	(0.172)
Constant	0.139	(0.019)	^***	0.003	(0.002)	^***	0.136	(0.021)	^***	0.001	(0.001)	^***

Model statistics, 2017: n = 54,666; weighted count = 114,814,425; Probability > F = 0.000; Replications = 160.

Model statistics, 2021: n = 53,706; weighted count = 122,684,695; Probability > F = 0.000; Replications = 160.

†

p value ≤ 0.1.

p value ≤ 0.05.

p value ≤ 0.01.

***

p value ≤ 0.001.

RRR, Relative risk ratio; SNAP, Supplemental Nutrition Assistance Program.

Interestingly (and contrary to our expectations), we also found that households’ monthly utility costs had a null or negative relationship with utility shutoffs, even after controlling for other factors. In 2017, higher monthly utility costs were associated with a marginally reduced risk of receiving an unactualized shutoff notice and experiencing an actual shutoff. However, in 2021, monthly utility costs did not have a significant effect on either outcome category.

On the other hand, directly paying for electricity (usually the largest among utility service bills) in 2017 was associated with an increased risk for both outcome categories. In 2021, the effect of directly paying for electricity was similar, though the significance level decreased slightly for actual shutoffs (from p < 0.001 to p ≤ 0.01). In 2021, households directly paying for trash service had an increased risk of receiving an unactualized shutoff notice.

Older householders in 2017 were less likely to receive an unactualized shutoff notice or experience an actual shutoff. In 2021, age had a substantively similar effect for both outcome categories, but at a reduced level of significance. Regarding formal education levels, householders with a bachelor’s degree or higher also had a substantially reduced risk of actual shutoffs in 2017 (RRR = 0.663, p < 0.01). In 2021, these householders had a reduced risk of receiving an unactualized shutoff notice, while their decreased risk of experiencing an actual shutoff became even more pronounced (RRR = 0.437, p < 0.001).

Regarding the race/ethnicity of the householder, Black householders had a substantially increased risk of experiencing an actual shutoff in 2017 compared with White householders (RRR = 2.157, p < 0.001). The substantially increased risk of experiencing an actual shutoff remained for Black householders in 2021 and, in fact, increased in magnitude (RRR = 2.671, p < 0.001). In 2017, Hispanic householders also had an increased risk of experiencing an actual shutoff, but this effect was reduced to marginal significance in 2021. In 2017, American Indian/Alaska Native householders had an increased risk of receiving an unactualized shutoff notice at marginal significance and a substantially greater risk of experiencing an actual shutoff (RRR = 4.452, p < 0.01). The effect for both unactualized shutoff notices and actualized shutoffs became nonsignificant for American Indian/Alaska Native householders in 2021. Asian householders had a decreased risk of receiving an unactualized shutoff notice in both study periods. Finally, householders who identified with two or more racial identities had an increased risk of receiving an unactualized shutoff notice at marginal significance in 2017 and a substantially increased risk of experiencing an actual shutoff (RRR = 2.162, p < 0.05). However, this effect became nonsignificant for both outcome categories in 2021.

As compared to homeowners, renters had a substantially increased risk of experiencing an actual shutoff in both 2017 (RRR = 2.278, p < 0.001) and 2021 (RRR = 2.261, p < 0.001). Moreover, there were notable differences by housing type. In 2017, residents of mobile homes or trailers had an increased risk of receiving an unactualized shutoff notice and an increased risk of experiencing an actual shutoff at marginal significance. In 2021, the increased risk of receiving an unactualized shutoff notice associated with living in a mobile home or trailer was reduced to marginal significance, while the risk of experiencing an actual shutoff increased.

Consistent with our bivariate findings, the regression results also indicate that geographic variation in utility shutoff notification and actualization was more pronounced in 2021 than in 2017. Compared with the New England division, where realized utility shutoffs dropped from 1.0 percent of residents in 2017 to 0.1 percent of residents in 2021, residents in the East North Central, South Atlantic, East South Central, West South Central, and Mountain divisions had a significantly higher risk of experiencing a utility shutoff in 2021. Non-metropolitan households had an increased risk of receiving an unrealized shutoff notice in 2017 at marginal significance, but this effect was not significant in 2021.

DISCUSSION

Our analysis indicates that, overall, utility shutoffs declined after the onset of the COVID-19 pandemic, and there were shifts in how shutoff procedures were implemented. This shift suggests that the state and local moratoria, along with associated economic support instituted near the onset of the pandemic, likely reduced utility shutoff actions. The national proportion of households that received an unactualized shutoff notice decreased from 2017 to 2021, as did realized shutoffs. Further and most interestingly, there was a marked decline in the proportion of households who received a notice and subsequently had their utilities shut off during the same period.

This last finding suggests a change in utility procedure, even among utilities that were authorized to send notices and conduct shutoffs. In other words, there is potential evidence of changes in utility-level shutoff procedure beyond adherence to top–down moratoria policies. At the same time, our finding that, in 2021, monthly utility costs did not have a significant effect on shutoff realization may reflect unobserved increases in household receipt of utility debt assistance and broader economic support from various sources (beyond the role of receiving public assistance, which we were able to measure and had a null effect). Thus, the relative role of household economic support versus utility policy in lowering residential utility shutoff rates bears further investigation.²⁸

Beyond overall lowered shutoff rates, shutoff moratoria and other associated policy changes were associated with a reduction in the risk of experiencing a utility shutoff among lower-income households between 2017 and 2021, thereby closing the income disparity in the risk of experiencing a shutoff. Meanwhile, our analysis also indicates that tribal shutoff moratoria were relatively effective, as the substantial discrepancy in the risk of experiencing utility shutoffs between American Indian/Alaska Native householders and White householders in 2017 was not present in 2021. However, given the magnitude of the discrepancy in 2017 and the well-documented disparities in public health support for utility infrastructure among American Indian/Alaska Native populations, tracking utility shutoff notification and actualization among American Indian/Alaska Native households in future waves of the AHS is an important objective for future research.^29,30

Despite the significant reductions in utility shutoffs during the COVID-19 pandemic discussed earlier, the relative risk of experiencing a utility shutoff increased for Black householders and residents of mobile homes or trailers from 2017 to 2021, while renters had a substantially higher risk of shutoffs in both periods than homeowners. These findings echo recent research indicating that utility shutoff moratoria were part of a broader, continued pattern of socially reproduced and unequal protections during the pandemic.³¹ In turn, these unequal protections reflect longstanding trends of racially and socioeconomically marginalized households and communities experiencing increased utility debt and insecurity.^{1,2,3,4,11,13}

Furthermore, we found a sizable degree of regional variation in utility shutoffs in 2021, even after accounting for the effects of the other covariates included in our multivariate analysis. These state and local disparities likely reflect the uneven design and implementation of shutoff protection policies by decision-makers and utility staff.^11,20 Empirically connecting and analyzing the impact of differences in de jure and de facto policymaking (explored in previous studies) alongside household socioeconomic and housing characteristics (explored here) on utility shutoff outcomes remains a gap in the literature.

Most major utilities maintained some form of shutoff protection policies pre-pandemic; this is likely evidenced in our finding that pre-pandemic, older householders tended to have lower utility shutoff rates than younger householders. Our findings also have nuanced findings for forward-looking policymaking and advocacy. Our study provides empirical evidence for the general sentiment in the sector that shutoff protection policies changed radically during the onset of the pandemic and that they have changed permanently and progressively, although permanent, publicly declared moratoria only remain in a few U.S. cities. Our results underscore the general effectiveness of sending shutoff notices as a method for utilities to collect payment from residential customers, which helps explain why many revenue-constrained localities and states have re-instituted some form of shutoff procedures.

At the same time, our study clearly demonstrates that inequities persisted in shutoff policy, even after landmark changes, among renters, mobile home residents, and Black households. This is especially urgent as there has been an additional rollback in shutoff protections at a national scale since 2021. What can be done to address ongoing shutoffs and the inequities in their implementation? First, it is important to remember that shutoff protections are an end-of-the-line—or crisis—intervention to support customer affordability. In other words, shutoff and debt relief policies serve as a backstop to rate design, conservation, and recurring bill discount interventions. These first-order policies should be instituted by utilities to holistically support service affordability for customers and lessen the need for shutoff procedures and associated protections.³²

Even if these upstream interventions are offered, however, the need for both monetary and procedural shutoff protections for marginalized households will remain. One potentially impactful option is to institute a permanent moratorium on shutoffs only for customers enrolled in recurring bill assistance programs, as the City of Los Angeles has done.³³ Another more modest option is to robust, rather than one-off, bill debt forgiveness programs with short-term shutoff protections, which are more commonly employed in other sectors but have been instituted post-moratoria by some energy utilities.³⁴ Procedurally, utilities must analyze and reform their own internal shutoff procedures to ensure that convenience, especially practices transacting shutoffs in bulk in neighborhoods where bill non-payment is more common, does not lead to further racial bias.³⁵ Utilities, and their regulators must also do more to ensure that legal tenant safeguards, such as protection from shutoff due to landlord non-payment of bills, are enforced.

Limitations

The present study also has several limitations warranting discussion. Most significantly, we rely on data from the 2021 wave of the AHS, which was collected while some COVID-19 utility shutoff moratoria were still being phased out, whereas others had ended. In addition, some one-off debt assistance funds were beginning to flow, while others had not yet been provided.36 Reassessing our research question using data from future waves of the AHS, when available, is an important objective for future research to assess the long-term trajectory of shutoff protection policy and household experience.

In addition, the measurement of utility shutoff notification and actualization in the AHS data does not distinguish between different types of utilities. However, infrastructural and managerial considerations that affect shutoff procedure vary by utility sector, ownership, and institutional type. The consequences of utility shutoffs for human health and well-being, although potentially serious across the board, also vary by utility type. For example, a lack of sufficient potable water can result in exposure to contaminants, dehydration, and insufficient household cleaning and personal hygiene (which, as previously mentioned, is particularly detrimental during a pandemic).^8,9,10,11 Meanwhile, a lack of access to electricity is associated with social isolation, heat exposure, and difficulty performing basic household tasks.^5,37 Likewise, restricted access to gas, oil, and other household fuels has unique implications for residents’ comfort and health which include food preparation and adequate household heating.^6,7 Information regarding the distribution of the risk of utility shutoffs based on utility type could facilitate a more targeted response from policymakers and utility managers at the national, state, and local levels.

Another significant limitation of our analysis lies with the geographic measures now made available in the AHS data. We were only able to account for variation in utility shutoff notification and actualization at the census division level, which lacks granularity and obscures potentially important state- and metropolitan-level differences. Additionally, regarding the rural-urban gradient, the post-2015 AHS data only allows for a binary comparison of households within and outside metropolitan areas. While we understand the need to protect the anonymity of AHS respondents, providing more granular geographic variables, such as a state-level measure and a more fine-grained measure of the rural-urban gradient in housing location (which was available before the 2015 revisions to the AHS survey instrument), would also help facilitate more thorough assessments of the variation in utility shutoff notification and actualization.

CONCLUSION

Utility service shutoffs inflict health, economic, and broader negative harm to the households who experience them. Meanwhile, lower-income and racially marginalized households and communities are disproportionately burdened by utility shutoffs. Our analysis suggests that utility shutoff moratoria and associated social protections, along with an follow-on policy changes during the COVID-19 pandemic, significantly reduced utility shutoffs and closed the income gap in actual utility shutoffs. In addition, the large disparity in utility shutoffs between American Indian/Alaska Native householders and White householders present in 2017 but not in 2021. It is additionally promising that blanket moratoria on utility service shutoffs due to an ability to pay, which has now expired at the state level, did not appear to be the only driver of changes in shutoff procedure, as this suggests a relatively progressive cultural shift in utility and social protection policy. However, due to limited data availability, we are unable to control for the state and local policy environment, which also evolved during the study period.

The extent to which this policy evolution has been sustained or has regressed post-2021, and the reasons for this in certain states and localities, remains unclear. The depth and breadth of momentum around progressive shutoff prevention or reduction policy in the UnitedStates thus needs to be sustained and further studied. The need for further action is illustrated by, among other things, our findings of sizable disparities in utility shutoffs that remained consistent for some groups between 2017 and 2021, notably Black householders, renters, and residents of mobile homes or trailers. Assessing sociodemographic and geographic variation in utility shutoff notification and actualization, in conjunction with the opportunities presented by evolving public expectations and administrative responses, presents an essential task for future environmental justice research, advocacy, and policy.

Footnotes

AUTHORS’ CONTRIBUTIONS

G.P.: Conceptualization (lead); Investigation (lead); Methodology (supporting); Project Administration (lead); Supervision (lead); Writing—Original Draft Preparation (supporting); Writing—Review and Editing (equal). M.J.B.: Conceptualization (supporting); Data Curation (lead); Formal Analysis (lead); Investigation (supporting); Methodology (lead); Resources (lead); Software (lead); Validation (lead); Writing—Original Draft Preparation (lead); Writing—Review and Editing (equal). S.H.: Conceptualization (supporting); Funding Acquisition (lead); Investigation (supporting); Methodology (supporting); Project Administration (supporting); Supervision (supporting); Writing-Original Draft Preparation (supporting); Writing—Review and Editing (equal).

AUTHOR DISCLOSURE STATEMENT

The authors have no conflicts of interest to declare.

FUNDING INFORMATION

No funding was received for this article.

SUPPLEMENTARY MATERIAL

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.47 MB