RESEARCH ARTICLE | SOCIAL SCIENCES

Behavioral and financial coping strategies among

energy-insecure households
Sanya Carleya,1 , Michelle Graffb, David M. Koniskya , and Trevor Memmotta

Edited by Douglas Massey, Princeton University, Princeton, NJ; received March 30, 2022; accepted July 12, 2022

When households struggle to pay their energy bills and avoid being disconnected from
the grid, they may accrue debt, forgo expenses on food, and use space heaters or ovens to Significance
warm their homes. These coping strategies can introduce significant physical and finan-
Millions of Americans are regularly
Downloaded from https://www.pnas.org by "UNIVERSITY OF MASSACHUSETTS BOSTON, HEALEY LIBRARY" on October 8, 2024 from IP address 158.121.54.186.

cial risks. In this study, we analyze an original survey with a representative sample of
low-income households during the first year of the COVID-19 pandemic, from June unable to pay their energy bills.
2020 to May 2021. We evaluate the prevalence of a wide range of coping strategies and For these Americans, avoiding
empirically estimate the determinants of these strategies. We find that more than half of being shut off from their electricity
all low-income households engage in at least one coping strategy, and many use multiple service is a daily challenge and
strategies. Households with vulnerable members, including young children or those who one that requires them to take
rely on electronic medical devices, and households that live in deficient housing condi- financial and behavioral risks,
tions, are more likely to use a range of coping strategies, and many at once. Our findings such as acquiring utility debt or
have direct implications for public policy improvements, including modifications to the
burning trash to generate heat. In
US Weatherization Assistance Program, the Low-Income Home Energy Assistance
this study, we find these
Program, and state utility disconnection protections.
techniques to be prevalent and
energy insecurity j energy poverty j coping strategies j material hardship j COVID-19 pandemic often used in combination. We
also find that households with
The “Big Freeze” of February 2021 led to rolling power outages across the state of young children and those with
Texas for more than 4.5 million households (1). Within 2 days, the freezing tempera- individuals who rely on electricity
tures and power blackouts resulted in over 150 official deaths, with one media source to power their medical devices are
reporting a death toll above 700 people (2). In the days following the blackout, the more likely to use these coping
news revealed several causes of death, including medical conditions exacerbated by the
techniques, as are households
blizzard, dangerous behaviors that households used to escape the cold (e.g., people
with deficient housing conditions.
sitting in their cars with the heat running, keeping a fire in the fireplace while the fam-
ily slept around it). Even after power had been restored to most Texans, many There are, however, efforts that
remained burdened with exceptionally high utility bills (3). The freezing cold tempera- the government can undertake to
tures and mass blackouts are an extreme case, but the incidence of these severe events is help these especially vulnerable
increasing due to climate change (4, 5), and people engaging in risky behavior to populations.
mitigate exposure to uncomfortable or dangerous temperatures is not a rare phenome-
non. Under “normal” economic and physical conditions, such strategies are practiced
regularly around the country—and world—by individuals and families.
Energy insecurity, or energy poverty, both of which refer to a household’s struggle
to pay energy bills and exposure to inadequate residential energy services, is a wide-
spread problem in the United States as well as across the world (6; see ref. 7 for a com-
prehensive discussion of definitions and metrics). In the United States, an estimated
24.3 million low-income households were unable to pay their energy bills and
Author affiliations: aO’Neill School of Public and
10.3 million were disconnected from their service providers between April 2019 and Environmental Affairs, Indiana University, Bloomington,
April 2020 (8), the year preceding the onset of the COVID-19 pandemic. Further- IN 47405; and bMaxine Goodman Levin College of Urban
Affairs, Cleveland State University, Cleveland, OH 44115
more, the incidence of energy insecurity is not evenly spread across all populations.
Over this time and during the onset of the pandemic, households of color, those with
young children, and those that rely on at-home electronic medical devices were signifi- Author contributions: S.C., M.G., D.M.K., and T.M.
cantly more likely to face utility disconnections (8). These disparities were exacerbated designed research; S.C., M.G., D.M.K., and T.M. performed
research; S.C., M.G., D.M.K., and T.M. contributed
by the pandemic and the resulting economic recession, which left millions out of work reagents/analytic tools; S.C., M.G., D.M.K., and T.M.
analyzed data; and S.C., M.G., D.M.K., and T.M. wrote the
(9) and increased rates of residential energy consumption and costs (10). paper.
Energy-insecure households make difficult decisions daily, navigating keeping their The authors declare no competing interest.
power on and maintaining safe indoor temperatures, while still meeting their other This article is a PNAS Direct Submission.
essential needs, such as food and health care. To pay their bills, households need to Copyright © 2022 the Author(s). Published by PNAS.
weigh several possibilities, including asking for a loan, seeking government assistance, This article is distributed under Creative Commons
Attribution-NonCommercial-NoDerivatives License 4.0
or engaging in potentially risky behavior to keep their bodies warm. (CC BY-NC-ND).
The extant literature provides some insights on how families cope when they struggle 1
To whom correspondence may be addressed. Email:
[email protected].
to pay their monthly bills. Material hardship scholars often study the role of various
This article contains supporting information online at
assistance programs in reducing hardship (see, for example, ref. 11), and they addition- http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.
ally consider the prevalence of different forms of hardship (see, for example, ref. 12). 2205356119/-/DCSupplemental.

Material hardship research, however, typically excludes energy insecurity, even though Published August 29, 2022.

PNAS 2022 Vol. 119 No. 36 e2205356119 https://doi.org/10.1073/pnas.2205356119 1 of 10

 it has been found to be among the most common forms of the home, such as small children or medically compromised fam-
material hardship for US families (12). The analyses that ily members, they may be more likely to pursue certain coping
include measures of energy insecurity focus specifically on strategies over others. The exceptions are studies by Harrington
financial coping strategies, finding that those who face energy et al. (27) and Gibbon and Singler (24), which find that house-
insecurity usually rely on assistance from friends or family, bill holds with small children or members with disabilities, respec-
balancing or alternating payments between bills, or seeking tively, more frequently accrue utility debt presumably because
assistance from government programs (11–17). these populations are less able to withstand extreme temperature
By contrast, scholarship on energy insecurity and related or food deprivation. No study, to the authors’ knowledge, has
topics focuses more heavily on behavioral coping strategies. A gathered all coping strategies into a single analysis, assessed the fre-
study of Austrian households, for example, found that the most quency and overlap among them, and evaluated which factors
common strategies for keeping a comfortable temperature in (i.e., sociodemographic and housing conditions) lead to the greater
the home were wearing extra layers of clothing, heating a single use of any given strategy.
room in the house, and “slipping under the covers” (18). Chard We address these gaps in our analysis. Here, we analyze the fre-
Downloaded from https://www.pnas.org by "UNIVERSITY OF MASSACHUSETTS BOSTON, HEALEY LIBRARY" on October 8, 2024 from IP address 158.121.54.186.

and Walker (19) found additional behaviors that UK families quency of various energy-coping strategies and what factors lead
with elderly members routinely adopted as commonsense solu- households to engage in one approach or another. We draw from
tions, including going to sleep earlier and only heating specific an original survey of a representative sample of approximately
rooms in their homes in the winter months. Evaluations in the 2,000 low-income households (with a resulting sample after multi-
United States and United Kingdom also found that households ple survey waves and accounting for the nonresponse of 5,187
facing energy insecurity tended to use less fuel or electricity, respondents)—defined as those within 200% of the federal pov-
allowing their home to reach uncomfortable temperatures erty level (FPL)—that we sampled at multiple points in time over
before they turned on heat or air conditioning (20–22). To the course of the COVID-19 pandemic, from roughly June 2020
afford monthly energy bills, other studies have found families to May 2021.
often cut expenditures, such as food or medical care (23, 24), The contributions of our analysis are twofold. First, we analyze
accrue considerable debt by taking out high-interest payday an original survey dataset that allows us to measure a wide range of
loans (20, 24, 25), or signing up for utility payment plans (22). coping strategies, including both behavioral and financial responses
In a recent study, Hernandez and Laird (26) examined how to energy insecurity, at the household level. The survey is longitudi-
households cope with energy insecurity. Analyzing household nal and measures outcomes across time periods that span the course
data from the 2015 Residential Energy Consumption Survey, of a year, which allows us to capture seasonal variation (e.g., tem-
they consider whether households facing the threat of discon- perature) and control for conditions faced by households in previ-
nections or actual disconnections respond by forgoing other ous time periods. Second, this analysis pulls together disparate
necessities, adjusting their home’s temperature to an unsafe or strands of literature to present and assess a wide range of coping
unhealthy level, or seeking energy assistance. They find that strategies, which we use to test which conditions lead certain house-
households, particularly those who report having had their
holds to adopt one approach or another, or combinations thereof.
energy services disconnected, use each of these coping strategies Our analysis is also set in the dynamic context of the
to some extent and often in combination.
COVID-19 pandemic, which restricted social engagements and
Collectively, the literature reveals that energy-insecure house-
set stay-at-home orders, unsettled labor markets, and shifted more
holds use a range of coping strategies and often pursue many at
energy use to residential settings. During this time, millions of
once (see, especially, refs. 11, 13, 20, 26), but the existing literature
households experienced material hardships and were unable to pay
has a few important gaps. First, few studies include a wide range
their monthly bills, including their energy bills. Unexpected eco-
of potential coping strategies in a single analysis. The material
hardship literature more often focuses on financial behavior, and nomic shocks require low-income households to make sudden
the energy literature more often considers temperature-seeking and difficult decisions (13). Therefore, the timing of the data col-
behavior. Studies that have considered both in the same analysis lection provides unique insight into how low-income households
have data limitations, including a limited set of household-level cope both financially and physically during times of economic
coping strategies. No previous study, to the authors’ knowledge, and social stress.
has evaluated a comprehensive range of coping strategies for this
distinct, yet prevalent, form of material hardship. Results
Second, while the literature emphasizes the challenges associated Coping Strategies. We combined insights from the extant liter-
with material hardship, it is less common for studies to explore ature to generate an extensive list of coping strategies, including
the risk associated with using coping strategies to mitigate such both financial and behavioral, which we then included in our
hardship. For example, households must make decisions about survey analysis. The survey results reveal that energy-insecure
whether to be disconnected from their service provider(s), risk a households tend to engage in four general coping strategy cate-
house fire by running a space heater, or forgoing meals to pay an gories. First, they seek the right temperature through behavioral
energy bill. In fact, with some exceptions (e.g., seeking bill assis-
techniques, some of which are riskier than others. We define
tance from the government or a friend or family member), many
risky temperature behavior as the use of space heaters, the fire-
coping strategies carry considerable risks, either financially or phys-
place, the oven, the dryer vent, or burning trash to generate
ically. Evaluating a wide range of coping strategies, both financial
heat.* Second, to pay energy bills, households tend to forgo
and behavioral, puts such risks in perspective and highlights how
dire the tradeoffs can be for energy-insecure households.
*In this analysis, we do not include the behavior of keeping one’s home at an uncomfort-
Third, less often addressed in the literature, and central to the ably high or low temperature because we do not have such a measure in our survey
present analysis, is the question of why energy-insecure households instrument. We assume that these riskier behavioral techniques, however, are to com-
pensate for an inability or unwillingness to turn on the heat or air conditioning to a safe
engage in one strategy or another and how certain conditions or comfortable temperature. Future studies may seek to disentangle these two behaviors
faced by a household may influence such behavior. For example, and may also consider including a more extensive set of both warmth- and cold-seeking
behaviors that individuals pursue with the expressed purpose of coping with energy
if a household has one or several vulnerable members residing in insecurity.

2 of 10 https://doi.org/10.1073/pnas.2205356119 pnas.org
 paying for other essential needs, such as medical bills or food. dynamics. In the early winter months, we observe that most types
Third, households seek assistance from both formal (e.g., gov- of coping strategies rise, except for forgoing expenses. Compara-
ernment, banks) and informal (e.g., friends, family, churches) tively, temperature-based behaviors rise significantly in the winter
networks to pay their bills. Fourth, households engage in a vari- months, which is expected since all the temperature behaviors
ety of bill strategies, such as carrying debt across utility bills, included in this analysis facilitate warmth. All four strategies
skipping payments on select bills each month until the conse- declined between the winter and spring months as temperatures
quences become severe (e.g., notice of disconnection, report to became more temperate and additional relief (i.e., COVID-relief
the credit authority), or practicing bill balancing, which is pay- checks and child tax credits) were released by the federal govern-
ing down a portion of one or more bills to have enough to pay ment during this time.
for another.
We display these practices in Fig. 1, along with the per- Determinants of Coping Strategies. Select regression analysis
centage of respondents who reported engaging in each activ- results are presented graphically in Fig. 3, with a table of full
ity at least once during the year. General coping strategy model results available in SI Appendix, Table S1. Our primary
Downloaded from https://www.pnas.org by "UNIVERSITY OF MASSACHUSETTS BOSTON, HEALEY LIBRARY" on October 8, 2024 from IP address 158.121.54.186.

categories are colored blue and are the dependent variables in analysis estimates the correlation between various socioeconomic
the primary analyses that follow. More specific strategies are household attributes and the four general coping strategies. We
colored green, which serve as the dependent variables in the estimate each model twice, first without and then with two meas-
secondary analyses. The figure illustrates that ∼55% of the ures of self-reported energy insecurity. The first measure is a
sample engaged in at least one coping strategy. Given that single-wave lagged measure of difficulty paying a household energy
the survey is a representative sample of low-income house- bill and the second is a self-reported utility disconnection at any
holds in the United States, we merge the data with estimates point between roughly May 2019, 1 year before survey adminis-
of those who live at or below 200% of the FPL from the tration, and the time of survey administration. Including these
2018 American Community Survey (ACS) to extrapolate that lagged variables enables us to account for both recent energy-
just under 10.5 million domestic households, or 53.4 million insecurity conditions that may lead one to engage in a coping
individuals, engaged in at least one coping strategy, with many strategy (e.g., if a person struggled to pay their bill last month,
using more than one during this time (28). The most common they may curtail energy usage and engage in other behaviors this
techniques, at 32% of the sample, are financial strategies, such as month), as well as previous, more extreme experiences that may
bill balancing or acquiring utility debt. The second most affect how one behaves in the present time period (e.g., if a person
common, at 26%, is behavioral—engaging in risky temperature- has been disconnected previously, they may be more likely to use
related techniques, such as using a space heater. A relatively lim- coping strategies to avoid being disconnected again). We ran
ited number of respondents, at 11%, seek government assistance robustness checks on the construction of these measures in SI
to cope with energy insecurity, despite this being one of the least Appendix, as discussed in the Determinants of Multiple Cop-
risky strategies of the full set. ing Strategies.
In Fig. 2, we present the proportion of the sample that engaged Results reveal that the two energy-insecurity measures are
in the general coping strategy categories, or the dependent varia- statistically significant determinants across all four general cop-
bles in the primary set of results, over the time period of analysis. ing strategies. If a household struggles to pay its bill in the pre-
The summer wave represents June 2020 through August 2020, vious wave or if it was disconnected at any time in the previous
the fall and winter wave represents September 2020 through Janu- year, then it is more likely to engage in all four general coping
ary 2021, and the winter and spring wave represents February strategy categories, particularly bill balancing. Fig. 3 also shows
2021 through May 2021. This figure reveals important seasonal that sociodemographic characteristics are associated with certain

Coping Strategies
(55%)

Forgo other
Risky Temperature Expenditures Payment Assistance Bill Strategies
Behavior (26%) (medical, food) (23%) (32%)
(17%)

Formal Assistance Informal Assistance

Networks (16%) Networks (10%)

Government Personal Networks

Banks and Payday Ulity Assistance
Assistance (Friends, Family, Bill Balancing Ulity Debt
or Other and Payment Plans
(WAP, LIHEAP, Neighbors, Religious (18%) (27%)
Lenders (2%) (6%)
Local Government or Other Group)
Assistance) (10%)
(11%)

Fig. 1. Coping strategies, with percentage of respondents who engaged in each activity between June 2020 and May 2021.

PNAS 2022 Vol. 119 No. 36 e2205356119 https://doi.org/10.1073/pnas.2205356119 3 of 10

Downloaded from https://www.pnas.org by "UNIVERSITY OF MASSACHUSETTS BOSTON, HEALEY LIBRARY" on October 8, 2024 from IP address 158.121.54.186.

Fig. 2. Proportion of respondents (with 95% CIs) who engaged in each coping strategy category, by wave.

coping strategies. Broadly speaking, households with vulnerable (1) households that have at least one member who relies on an
members are more likely to engage in at least one of the general electronic medical device, (2) households with young children,
categories of coping strategies. and (3) those who live in deficient housing conditions. Consistent
As presented, our results reveal that households with children with Harrington et al. (27) and Gibbon and Singler (24), house-
younger than age 5 years are more likely to engage in all of the holds with small children and with members who are medically
strategies: use temperature strategies, forgo expenses, seek pay- compromised, either disabled or rely on an electronic medical
ment assistance, and use bill strategies. Households with device, are more likely to accrue debt; yet we also find that several
health-compromised members are more likely to seek a broader other vulnerable populations are more likely to take on utility
range of coping strategies as well. Specifically, if a household debt as a coping strategy, including those who have experienced
has a member with a medical disability, it more typically seeks energy insecurity in the past, Black and Hispanic households,
payment assistance, and households with members who rely on those who live in deficient housing conditions, and those at the
an electronic medical device, a demographic that is particularly lowest income level.
vulnerable to a lack of power, engage in all of the strategies, We additionally find that households that have a member
including using risky temperature behaviors, which, of course, with a disability as well as Black households are both more
could particularly compromise the health of the household. We likely to seek out government assistance, while Hispanic house-
also find racial disparities, with Black and Hispanic households holds and respondents with less than a high school education
more likely than White households to use bill strategies, such are more likely to engage in bill balancing. Those in the lowest
as bill balancing or debt accrual, and Hispanic households income strata and those who are 100 to 150% of the FPL are
more likely than White households to forgo expenses, but associated with seeking government assistance and tapping their
neither is more or less likely to undertake temperature strate- informal networks, while those under 100% FPL are also more
gies. Lastly, when we stratify the sample by income, the results likely to balance their bills.
reveal that lower income levels correlate with seeking payment
assistance, forgoing expenses, and using bill strategies. Determinants of Multiple Coping Strategies. The results pre-
Beyond the sociodemographic characteristics, Fig. 3 reveals that sented thus far suggest that households often engage in more than
those living in deficient housing conditions—mold in the home, one coping strategy, and often at the same time, as reaffirmed by
poor insulation, drafty air, holes in the wall, bad plumbing, exposed other scholars (11, 13, 20, 26). A natural extension is an examina-
electric sockets, broken air conditioning, or nonworking stove or tion of which sociodemographic or household factors correlate with
refrigerator—are more likely to engage in all four general categories the use of a greater number of strategies. To do so, we ran a Pois-
of coping strategies. Finally, there is significant seasonal variation, son model in which the outcome measure is a count variable that
likely due to weather conditions. SI Appendix, Table S1 reveals that ranges from zero (no coping strategies) to four (all four main coping
in the colder months, households are more likely to engage in all of strategy types, as presented in blue in Fig. 1). Fig. 5 presents the
the coping strategies, whereas in the warmer months, respondents marginal effect size of key variables, and the full set of results are
used less warmth-seeking behavior and more bill balancing. presented in SI Appendix, Table S3. Here, we see that previous inci-
We also tested the specific coping techniques under the pay- dence of energy insecurity, sociodemographic variables, and defi-
ment assistance and bill strategies categories (i.e., those in green in cient housing conditions are correlated with a greater number of
Fig. 1). Select results are presented in Fig. 4, with a model specifi- strategies used in the home. Previous inability to pay one’s energy
cation that matches the models presented in Fig. 3; full model bill has the largest effect size, followed by a previous disconnection
results are found in SI Appendix, Table S2. Here, we find three experience. On average, those with small children in the household
groups that are more likely to engage in all six coping strategies: or with medical conditions use more coping strategies. Similarly, we

4 of 10 https://doi.org/10.1073/pnas.2205356119 pnas.org
 We test the robustness of these results through several tech-
niques. First, we modify the construction of the energy-
insecurity variables by making them both single time period
lags and a measure of occurrence anytime in the past year,
respectively. Second, we consider the possibility that the unbal-
anced nature of our panel introduced bias if observations are
missing not at random. Third, we expand the count variable in
the final regression to include all coping strategy measures, not
just the higher order categories. In this set of regressions, we
include behavior strategies and forgoing expenses, and then sep-
arate out each individual measure of payment assistance and
bill strategies, for a total of eight measures. Finally, we replace
the state fixed effects with regional fixed effects because regional
Downloaded from https://www.pnas.org by "UNIVERSITY OF MASSACHUSETTS BOSTON, HEALEY LIBRARY" on October 8, 2024 from IP address 158.121.54.186.

variables may help account for climate and weather patterns.

We estimate these models with both standard regional classifi-
cations (e.g., West, South, Northeast, and Midwest) and
climate regions using the classification from Karl and Koss
(29), which are also used by the National Oceanic and Atmo-
spheric Administration.
All robustness checks are available in SI Appendix, Tables
S4–S7. Model results are remarkably stable with the variation
in energy insecurity measures (SI Appendix, Table S4) and with
the alternative count variable (SI Appendix, Table S6). The bal-
anced panel results (SI Appendix, Table S5) are also quite simi-
lar to the main results in terms of statistical significance,
although with a few differences in which a variable falls out of
the conventional statistical significance thresholds. Specifically,
those with household members who rely on medical devices are
not associated with risky temperature behavior or payment
assistance. While these minor differences between the main
results and the balanced panel results are important to weigh
when assessing the overall validity of the findings, none of these
suggest significant model specification problems, nor do they
undermine the main findings of the analysis. The final set of
models, in SI Appendix, Table S7, which controls for region, is
entirely consistent with the main results.
Fig. 3. Linear probability model regression results, with 95% CIs for
selected variables with dependent variables: risky temperature behavior, Discussion
forgo expenses, seek payment assistance, and engage in a bill strategy
(n = 5,187). The omitted race category is White. The omitted income In this analysis, we address two tiered questions. First, descrip-
category is household within 150 to 200% of the FPL. Additional control
variables include members in the house are older than 65 years, other
tively, what are the primary ways in which households cope with
race, respondent has a high school education or less, household is within energy insecurity and how prevalent is each strategy? Second, who
100 to 150% of the FPL, respondent is employed, home ownership/renter engages in these strategies among the US low-income population?
type, type of home, state fixed effects, and wave fixed effects. To answer these questions, we analyze data from an original,
nationally representative, multiwave survey of low-income
find that households of color and lower-income households also households that was administered during the first year of the
typically use more strategies. These behaviors are more pronounced COVID-19 pandemic. We find that households do not solely
in the winter months than in the summer months. engage in financial coping techniques, such as debt accrual, paying
Fig. 5 provides insights as to who is more likely to engage in their bills at strategic intervals, or applying for government assis-
a greater number of coping strategies, but it does not reveal tance, as is typical for other forms of material hardship. In fact,
how households may pair specific strategies together. In Table 1, households pursue additional, arguably quite risky, energy-specific
we extrapolate the number of low-income households that behaviors. For example, to pay their energy bills, households will
engage in any given pairing of the four general coping strategies forgo purchasing groceries, delay going to the doctor, or compen-
over the course of the year. Once again, we calculate these val- sate for colder indoor temperatures through risky warmth-seeking
ues by multiplying the 2018 ACS estimate of all US households behaviors. Our results reveal that all of these practices are not only
within 200% of the FPL (28) by the percentage of households common among low-income American households but also many
in the sample that reported using these coping strategies over households engage in several of these approaches simultaneously.
the course of the year. The most common pairings, with more We also find variation across sociodemographic groups.
than 2 million households each, are between forgoing expenses Most important, we find that households with vulnerable resi-
and bill strategies as well as risky temperature behavior and dents, such as young children or medically compromised indi-
bill strategies. The least common is the pairing of payment viduals, cope through several concurrent techniques. Both types
assistance and bill strategies; however, more than 190,000 of households use both behavioral strategies, such as engaging
households still engaged in these two financial strategies simul- in potentially risky efforts to remain warm, and financial strate-
taneously during the course of the year to pay energy bills. gies, such as taking on utility debt, strategically shifting their

PNAS 2022 Vol. 119 No. 36 e2205356119 https://doi.org/10.1073/pnas.2205356119 5 of 10

Downloaded from https://www.pnas.org by "UNIVERSITY OF MASSACHUSETTS BOSTON, HEALEY LIBRARY" on October 8, 2024 from IP address 158.121.54.186.

Fig. 4. Linear probability model regression results, with 95% CIs for selected variables with dependent variables government assistance, informal network
assistance, energy provider assistance, loan, bill balancing, and utility debt (n = 5,187). The omitted race category is White. The omitted income category is
household within 150 to 200% of the FPL. Additional control variables include members in the house are older than 65 years, “other” race, respondent has a
high school education or less, household is within 100 to 150% of the FPL, respondent is employed, home ownership/renter type, type of home, state fixed
effects, and wave fixed effects.

bill payments, or seeking payment assistance from both formal condition. Our empirical analysis further suggests that when a
and informal networks. This suggests that these households household has dilapidated or inefficient dwelling conditions, such
need to do everything within their abilities, including the most as a broken heating, ventilation, and air conditioning system or
risky strategies, to avoid a life-threatening utility disconnection. holes in the wall, they more frequently use all possible coping strat-
The results of this analysis suggest several opportunities for pub- egies. Helping these households repair equipment, upgrade appli-
lic policy action. First, one of the leading predictors of engaging in ances, and update the physical structures in which they live would
any, or a combination of, coping strategies is the condition of improve energy efficiency and lower home energy bills, and
one’s home. Approximately 17.8% of survey respondents, or an thereby alleviate the need to engage in risky coping behaviors.
estimated 3.4 million low-income households (28), reported Thus, the government can expand programs such as the Weatheri-
that their dwelling suffered from at least one deficient housing zation Assistance Program (WAP)—a federally funded program

Fig. 5. Marginal effects from Poisson regression with dependent variable as a count of coping strategies.

6 of 10 https://doi.org/10.1073/pnas.2205356119 pnas.org
 Table 1. Estimates of households that engaged in coping could specifically and swiftly encourage utilities to provide dis-
strategy pairings connection protections for households with medically compro-
mised individuals as well as households with young children.
Risky Forgo While only five states provide protections for households with
temperature expenses Payment young children, nearly all states currently have a disconnection
behavior (%) (%) assistance (%) policy to protect those with medical conditions. However, the
Forgo expenses 1,429,578 stringency of the medical protections varies across states and
(7.5) often requires one or more notes from a physician (33), which
Payment assistance 1,848,921 1,753,615 may be a hurdle for some applicants to obtain, especially if
(9.7) (9.2) they are unable to afford timely medical care. Therefore, states
Bill strategy 2,153,897 2,287,325 190,610 could reconsider the breadth of these protections for their vul-
(11.3) (12.0) (1.0) nerable constituencies as well as consider easing the burdens
households with particularly vulnerable members must over-
Downloaded from https://www.pnas.org by "UNIVERSITY OF MASSACHUSETTS BOSTON, HEALEY LIBRARY" on October 8, 2024 from IP address 158.121.54.186.

Each cell contains 2 values: (1) an estimate of the No. of US households within 200% of
the FPL that would engage in each coping strategy pairing and (2) the proportions of the
come to benefit from these policies.
survey sample that reported engaging in any given pairing over the course of the year Our results highlight the vulnerability that millions of low-
are in parentheses. income families face, many of whom are putting their financial
and physical well-being at risk to maintain indoor thermal
that helps low-income households update their dwellings each comfort, pay their energy bills, and ultimately avoid utility dis-
year through energy efficiency upgrades and renewable energy connection. We collected these data during a public health cri-
installations—and, with insights from this analysis, target house- sis, during which stay-at-home orders were enacted. The orders
holds with residents who are particularly vulnerable, such as those resulted in people spending more time at home (i.e., consum-
with young children or with individuals who rely on electronic ing more residential energy) than before the pandemic.
medical devices. Even though millions of homes need repairs and Although the timing of the data collection may limit the gener-
efficiency upgrades, WAP currently only helps 35,000 low-income alizability of the findings to future conditions, there is also
households each year (30). To ensure more robust participation in reason to believe that other such factors may lead to worse con-
WAP, the government can also increase annual appropriations and ditions in coming years. For example, utilities will need to
clearly communicate the value of weatherization, evaluate success invest in low-carbon capacity and infrastructure, which will
based on established energy poverty metrics (6), and collaborate likely increase energy costs and related household bills (34). In
with local community groups and members to ensure the informa- addition, climate change will continue to produce more erratic
tion is shared with energy-insecure households (31). and extreme temperatures, leading to more energy demand—
Second, the federal government can also allocate more funds heat in the winter and air conditioning in the summer—as well
to the Low-Income Home Energy Assistance Program as more weather-related emergencies (35). These factors are
(LIHEAP), a federally funded energy bill assistance program. likely to converge, which will expand the energy-insecure popu-
We found that applying for government payment assistance is lation and the need for already vulnerable individuals and fami-
significantly less common than several other far riskier coping lies to use at least one potentially risky coping strategy to pay
techniques, such as temperature-seeking behavior, forgoing monthly energy bills. As such, it is imperative that we recognize
expenses on food and health care, and using bill strategies. This the prevalence of the domestic energy insecurity problem, the
is not surprising because currently, due to limited Congressio- risks associated with household coping behavior, and the need
nal appropriations, LIHEAP serves 20 to 25% of the eligible for targeted policy interventions to help alleviate this especially
population, is only offered once per year to each eligible appli- dire form of material hardship.
cant, and has a limited program year (32). We also found that
households engage in more coping strategies at the end of the
year, or in the colder winter months, than other times of year,
Table 2. Distribution of respondents by region and
even though nearly 50% of LIHEAP funds are expended on
climate region
helping households heat their homes. Based on this analysis,
demand for bill assistance outstrips supply and outreach efforts
Sample, %
to energy-insecure households are likely insufficient. Therefore,
Congress should consider appropriating more money toward Regional distribution
LIHEAP. In the absence of increased funding, LIHEAP admin- West 21.3
istrators should continue to target outreach efforts to the popu- Northeast 15.8
lations that are the most likely to engage in risky financial and Midwest 21.0
physical behaviors identified in this article, including house- South 44.8
holds with young children, those with medically compromised Climate region distribution
members, and those who live in deficient housing conditions. Northeast 17.7
Third, state institutions can protect their constituents from util- Upper Midwest 6.5
Ohio Valley 16.9
ity disconnection by applying temperature-based, date-based, or
Southeast 21.3
targeted protections for vulnerable populations (33).† Many states
Rockies and Plains 1.7
offer such protections, but there remain opportunities for expand- Southeast 14.9
ing the scope, scale, and duration of coverage, especially in the Southwest 5.2
particularly hot and cold months, when households seek danger- Northwest 3.5
ous coping strategies. Based on the results in this analysis, states West 11.6
Alaska 0.1
†
Note that we included these protections in the present analysis insofar as they are cap- Hawaii 0.5
tured in the state fixed effects term.

PNAS 2022 Vol. 119 No. 36 e2205356119 https://doi.org/10.1073/pnas.2205356119 7 of 10

 Table 3. Variable names, definitions, and descriptive statistics

Type of Linearized
Variable name Operational definition variable Mean SE 95% CI

Dependent variables
Risky temperature behavior Respondent reports engaging in any of the following: using the stove for Binary 0.2597 0.0101 0.2398–0.2796
space heat, using a space heater, burning trash in the home for heat,
using the fireplace for heat, or using the dryer vent for heat
Forgoes expenses Respondent reports the need to forgo expenses on food or medical care Binary 0.1697 0.0092 0.1516–0.1877
to pay for energy
Seeks payment assistance Respondent reports seeking payment assistance to help pay energy bill Binary 0.2270 0.0109 0.2057–0.2484
Engages in bill strategy Respondent reports engaging in a bill strategy, either balancing Binary 0.3170 0.0126 0.2922–0.3417
payments across bills or acquiring utility debt
Seeks government assistance Respondent reports seeking government assistance through WAP, Binary 0.1146 0.0077 0.0995–0.1298
LIHEAP, or another local government program
Seeks payment assistance Respondent reports seeking payment assistance from friends, family, Binary 0.1036 0.0075 0.0889–0.1181
Downloaded from https://www.pnas.org by "UNIVERSITY OF MASSACHUSETTS BOSTON, HEALEY LIBRARY" on October 8, 2024 from IP address 158.121.54.186.

from informal network a faith organization, or a nonprofit

Seeks payment assistance Respondent reports seeking payment assistance from their utility Binary 0.0579 0.0057 0.0466–0.0692
from utility provider through a payment plan or other utility support
Takes out a loan Respondent reports taking out a loan or payday lending to pay an Binary 0.0193 0.0033 0.0127–0.0258
energy bill
Engages in bill balancing Respondent reports engaging in bill balancing Binary 0.1835 0.0108 0.1624–0.2046
Takes on utility debt Respondent reports taking on utility debt Binary 0.2719 0.0118 0.2487–0.2950
No. of coping strategies No. of the following strategies used by the household: risky Ordinal 0.9764 0.0310 0.9156–1.0370
temperature behavior, forgoes expenses, seeks payment assistance,
engages in bill strategy (ranges from 0 to 4)
Independent variables
Lagged, struggled to pay Respondent reported difficulty paying energy bill in the previous wave Binary 0.1418 0.0098 0.1225–0.1611
energy bill, t-1
Previously disconnected Respondent reported being disconnected at some previous time Binary 0.0675 0.0090 0.0499–0.0851
period, either in a previous wave or any time the year before the
first data administration in May 2020
Children in house are There is at least 1 child living in the house who is younger than 5 y old Binary 0.1517 0.0105 0.1310–0.1724
younger than 5 y
Member(s) of household There is at least 1 individual living in the house who is older than 65 y Binary 0.3292 0.0121 0.3054–0.3529
are older than 65 y
Member(s) of household There is at least 1 individual living in the house who has a medical Binary 0.4042 0.0129 0.3790–0.4295
have a disability disability
Member(s) of household There is at least 1 individual living in the house who relies on an Binary 0.1799 0.0100 0.1603–0.1996
rely on electronic medical device electronic medical device
Black household Household is of Black racial composition Binary 0.1698 0.0110 0.1482–0.1914
Hispanic household Household is of Hispanic racial composition Binary 0.1972 0.0144 0.1689–0.2255
Household identifies as other race Households is of other racial composition Binary 0.0909 0.0081 0.0748–0.1069
White household (omitted) Households is of other White racial composition Binary 0.5414 0.0150 0.5120–0.5707
Household respondent has Respondent reports having only a high school diploma or less Binary 0.5529 0.0138 0.5258–0.5801
high school education or less
Household is within 100 to Respondent reports that household is within 100 to 150% of the FPL, Binary 0.2966 0.0112 0.2746–0.3185
150% of FPL based on income categories provided by the authors in the survey
instrument
Household is under 100% Respondent reports that household is under 100% of the FPL, based on Binary 0.3999 0.0131 0.3742–0.4255
of the FPL income categories provided by the authors in the survey instrument
Household is 150 to 200% Respondent reports that household is within 150 to 200% of the FPL , Binary 0.2122 0.0094 0.1937–0.2307
of the FPL (omitted) based on income categories provided by the authors in the survey
instrument
Household respondent is Respondent reports being employed at time of the survey Binary 0.3164 0.0132 0.2906–0.3423
employed
Broken or inefficient house At least 1 of the following house conditions is present: mold in the home, Binary 0.1778 0.0074 0.1633–0.1922
conditions reported poor insulation, drafty air, holes in the wall, bad plumbing, exposed
electric sockets, broken AC, or nonworking stove or refrigerator
The respondent(s) rent Respondent rents their residence Binary 0.5021 0.0142 0.4743–0.5299
their residence
The respondent(s) claim Respondent neither rents nor owns their residence Binary 0.0740 0.0072 0.0599–0.0882
“other” for home ownership type
The respondent(s) own Respondent owns their residence Binary 0.5509 0.0146 0.5223–0.5796
their residence (omitted)
Home is a trailer Dwelling is a trailer Binary 0.1147 0.0094 0.0962–0.1332
Home is an apartment or Dwelling is an apartment or condominium Binary 0.3334 0.0140 0.3059–0.3608
condominium
Home type is “other” Dwelling type is reported as “other” Binary 0.0027 0.0008 0.0012–0.0043
Home type is single-family Dwelling is a detached or attached single-family home Binary 0.4242 0.0059 0.4126–0.4358
home (omitted)
Summer (wave 2) Data gathered in the second wave Binary 0.4255 0.0060 0.4138–0.4372
Fall/Winter (wave 3) Data gathered in the third wave Binary 0.3164 0.0043 0.3079–0.3249
Winter/Spring (wave 4) Data gathered in the fourth wave Binary 0.2605 0.0052 0.2504–0.2707
Previously struggled to pay bill Respondent reported difficulty paying energy bill either in any Binary 0.2505 0.1290 0.2252–0.2758
previous wave or any time the year before the first data
administration in May 2020
Lagged, disconnected, t-1 Respondent reported being disconnected from their utility provider Binary 0.0365 0.0060 0.0247–0.0483
in the previous wave

8 of 10 https://doi.org/10.1073/pnas.2205356119 pnas.org
 Materials and Methods waves, all at wave two values; we then compare those who stopped after
the third wave with those who went on to complete the final wave, all at
Survey Design and Sampling. This study was approved by the Indiana Univer- wave three values. The results show that there are no large differences for
sity Human Subjects & Institutional Review Board as exempt status. We pro- any variable, either across subsample or within wave. All means are well
duced the data for this analysis via an original survey designed by the authors within the SDs of one another.
and administered by YouGov, a private polling firm. To build the sample, we sur- The final unbalanced sample is 2,247 in wave two (summer), 1,670 in wave
veyed a representative group of respondents of those within 200% of the FPL. three (fall/winter), and 1,378 in wave four (winter/spring). With some nonres-
This threshold is a common percentage for analyses of low-income households ponse to survey questions, the final sample size in the regressions is 5,187. We
(see, for example, ref. 36). In addition, federal energy assistance programs such presented the weighted distribution of respondents across regions and climate
as LIHEAP tend to use 150% of the FPL as the eligibility threshold; therefore, our regions in Table 2.
sample allows us to analyze differences across households that are under and We administered the survey online and it lasted ∼10 to 15 min. The survey
over this specific threshold. included questions on respondents’ personal and household characteristics,
YouGov generates a random and representative sample through a two- housing conditions, heating and cooling behavior, rates of energy insecurity,
step process. First, the firm draws a random sample from the full population and various coping strategies. In each of the waves, we asked respondents to
Downloaded from https://www.pnas.org by "UNIVERSITY OF MASSACHUSETTS BOSTON, HEALEY LIBRARY" on October 8, 2024 from IP address 158.121.54.186.

of all households within 200% of the FPL. Next, the firm uses a matching reflect on the past several months to identify whether, at any point over that
algorithm to select matching households from their proprietary panel of time period, they experienced an inability to pay their energy bill or a utility
approximately 2 million US participants. Scholars have widely validated disconnection.
YouGov’s sampling techniques (37–39). For the present analysis, we gener-
ated demographic survey weights to ensure that the sample is fully repre- Variables. In this analysis, we used these various measures to test the relation-
sentative of the population of low-income households and applied these ship between household sociodemographic variables and household conditions,
weights in all regressions. YouGov awards points per completed survey, respectively, and various coping strategies. All of the dependent variables
which translates into financial compensation. included in Figs. 2 to 4 are binary, coded as a “1” if a respondent reported that
We administered the survey at four separate points in time, roughly coincid- their household performed that action in a given wave, and “0” otherwise. The
ing with different seasons, over the course of the year from mid-2020 to two energy insecurity independent variables are coded as follows: the first is a
mid-2021. In the first wave, which we administered in May 2020, we asked single-wave lag if the respondent reported having difficulty paying their energy
bill in the last wave; the second is a comprehensive measure of a previous utility
questions about baseline conditions of energy insecurity, both in the month of
disconnection, in the event that a respondent reported in any previous time
May and over the past year, the year that predated the COVID-19 pandemic. In
period that they were disconnected, including the prewave-one time period of
the present analysis, we only use this wave to construct lagged energy insecurity
the year before the pandemic. We present the variables that we include in the
variables.‡ Beginning in wave two, which we administered in August 2020, we
analyses, the way that we operationalized them, and their descriptive statistics
asked questions that pertained to several months at a time. In our August sur-
in Table 3.
vey, we covered the months of June, July, and August. We administered wave
The timing of the data collection also coincided with several policies that
three in January 2021, which pertained to the months of September through
aimed to reduce hardship for all Americans due to the COVID-19 pandemic.
January, and wave four in late May and early June of 2021, which pertained to
For example, the distribution of the Child Tax Credit began during wave two,
the months of February through June. This analysis focuses only on waves two
summer 2020, and the second and third round of stimulus checks, $600 and
through four.
$1400, respectively, were dispersed during wave three, fall/winter 2020. The
The decline in sample size across waves of the survey was anticipated,
timing and provision of these payments likely explains why the overall pro-
given our expectations about attrition. We also set approximate quotas for
portion of the sample population that forwent expenses reduces over the
each wave to achieve nationally representative samples, and the survey was
duration of the analysis. In addition, during this time, state- and utility-level
closed when these quotas were met, in the interest of keeping the surveys
policies were implemented to protect households from being disconnected if
open for as short of a duration of time as possible. This approach allowed us they were unable to pay their utility bills. While this collection of policies may
to document one person’s response, for example, completed on the first have had an impact on household coping strategies, we do not include them
day that the instrument was open, as in the same time period as another’s, as covariates in the present analysis. The stimulus and tax credit affect all eli-
who completed it on the last day that it was open. If we left the instrument gible families the same across space and time. By contrast, the utility discon-
open for too long, then the first respondent may have experienced nection policies do vary over space and time; however, these protections are
completely different circumstances (e.g., policy, coronavirus, personal cir- monthly and at times daily protections, which cannot cleanly be aggregated
cumstances) than the second respondent, but we would still claim them to the wave level at which we collected the data and present our results. In
both as occurring in the same wave. To avoid this problem, we set threshold addition and importantly, it is likely that any variation in coping strategies
for responses with YouGov ahead of time and closed the survey once we caused by these policies will be soaked up by the state and wave fixed effects
met those thresholds. that we include in all of the models, allowing the specification to provide
Thus, there are individuals who dropped from the sample not due to unbiased correlations between the sociodemographic indicators and house-
nonresponse but due to us closing the survey. As a result, it is impossible hold coping strategies.
for us to identify who dropped from the sample due to attrition and who
dropped due to the survey closing. To analyze whether anyone was discon- Regression Analysis. We use a linear probability model to produce the results
tinued in the study not at random, however, we present basic descriptive presented in Figs. 3 and 4 and a Poisson model for Fig. 5. For each regression,
statistics in SI Appendix, Tables S8 and S9. First, for any sociodemographic we use the full suite of covariates with survey weights. The full reporting of
variable that is likely to remain constant over the waves, we present means modeling results is contained in SI Appendix.
and SDs by subsample: those who stopped after the second wave, those Data, Materials, and Software Availability. Anonymized replication data
who stopped after the third wave, and those who made it through all of the and statistical code data have been deposited in the Harvard Dataverse (https://
waves. Second, for any variable that was likely to change over time, which doi.org/10.7910/DVN/XKGOCW) (40).
includes housing, energy insecurity, and coping strategy variables, we esti-
mate means and SDs within waves. Here, we compare those who stopped ACKNOWLEDGMENTS. This research was supported by funding provided
after the second wave with those who went on to complete subsequent by the NSF; the Alfred P. Sloan Foundation; the Environmental Resilience
Institute, funded by Indiana University’s Prepared for Environmental
‡
Wave one was different than the other waves in several ways. First, wave one only cov-
Change Grand Challenge initiative; and the Indiana University Office of the
ered 1 month of time, while the others covered several months. Second, wave one Vice President of Research. We thank Dr. Diana Hernandez for sharing her
occurred during the heart of the stay-at-home orders and mass shutdowns. Thus, includ-
ing this wave of data in a substantive way could further limit the external validity of the
survey instrument on energy insecurity, which informed our own survey
findings. design.

PNAS 2022 Vol. 119 No. 36 e2205356119 https://doi.org/10.1073/pnas.2205356119 9 of 10

 1. C. Pollock, Gov. Greg Abbott says power is almost fully restored statewide and grocery stores will 21. S. Cong, D Nock, Y.L. Qiu, B. Xing, Unveiling hidden energy poverty using the energy equity gap.
soon be restocked. The Texas Tribune, 21 February 2021. https://www.texastribune.org/2021/02/ Nature Comm. 13, 2456 (2022).
21/texas-power-outage-grocery-stores-greg-abbott/. Accessed 30 September 2021. 22. D. Hernandez, Understanding ‘energy insecurity’ and why it matters to health. Soc. Sci. Med.
2. P. Aldhous, S. Lee, B. Hirji, The Texas winter storm and power outages killed hundreds more 167, 1–10 (2016).
people than the state says. Buzzfeed, 26 May 2021. https://www.buzzfeednews.com/article/ 23. J. Bhattacharya, T. DeLeire, S. Haider, J. Currie, Heat or eat? Cold-weather shocks and nutrition in
peteraldhous/texas-winter-storm-power-outage-death-toll. Accessed 30 September 2021. poor American families. Am. J. Public Health 93, 1149–1154 (2003).
3. G. Nieto del Rio, N. Bogel-Burroughs, I. Penn, His lights stayed on during Texas’ storm. Now he 24. D. Gibbons, R. Singler, Cold Comfort: A Review of Coping Strategies Employed by
owes $16,752. NY Times, 1 March 2021. https://www.nytimes.com/2021/02/20/us/texas-storm- Households in Fuel Poverty (Inclusion Research Consultancy & Energywatch, Aarhus, Denmark,
electric-bills.html. Accessed 30 September 2021. 2008).
4. B. Stone Jr. et al., Climate change and infrastructure risk: Indoor heat exposure during a 25. R. Levy, J. Sledge, A Complex Portrait: An Examination of Small-dollar Credit Consumers
concurrent heat wave and blackout event in Phoenix, Arizona. Urban Clim. 36, 100787 (2021). (Financial Health Network, New York, 2012).
5. J. W. Busby et al., Cascading risks: Understanding the 2021 winter blackout in Texas. Energy Res. 26. D. Hernandez, J. Laird, Surviving a shut-off: U.S. households at greatest risk of utility
Soc. Sci. 77, 102106 (2021). disconnections and how they cope. Am. Behav. Sci. 66, 856–880 (2021).
6. D. J. Bednar, T. G. Reames, Recognition of a response to energy poverty in the United States. Nat. 27. B. E. Harrington et al., Keeping warm and staying well: Findings from the qualitative arm of the
Energy 5, 432–429 (2020). Warm Homes Project. Health Soc. Care Community 13, 259–267 (2005).
7. M. A. Brown, A. Soni, M. V. Lapsa, K. Southworth, M. Cox, High energy burden and low-income 28. Kaiser Family Foundation (KFF), Distribution of total population by federal poverty level. https://
energy affordability: Conclusions from a literature review. Prog. Energy 2 042003 (2020). www.kff.org/other/state-indicator/distribution-by-fpl/?currentTimeframe=0&sortModel=%7B%22
8. T. Memmott, S. Carley, M. Graff, D. M. Konisky, Socioeconomic disparities in energy insecurity colId%22:%22Location%22,%22sort%22:%22asc%22%7D. Accessed 27 June 2022.
Downloaded from https://www.pnas.org by "UNIVERSITY OF MASSACHUSETTS BOSTON, HEALEY LIBRARY" on October 8, 2024 from IP address 158.121.54.186.

among low-income households before and during the COVID-19 pandemic. Nat. Energy 29. T. R. Karl, W. J. Koss, Regional and National Monthly, Seasonal, and Annual Temperature
6, 186–193 (2021). Weighted by Area, 1895-1983. Historical Climatology Series 4-3 (National Climatic Data Center,
9. G. Acs, M. Karpman, Employment, income, and unemployment insurance during the Covid-19 pandemic. Asheville, NC, 1984).
https://www.urban.org/sites/default/files/publication/102485/employment- 30. US Department of Energy (DOE), Weatherization Assistance Program. https://www.energy.gov/
income-and-unemployment-insurance-during-the-covid-19-pandemic.pdf. Accessed 30 September 2021. sites/default/files/2022-06/wap-fact-sheet_0622.pdf. Accessed 27 June 2022.
10. J. Lou, Y. L. Qiu, A. L. Ku, D. Nock, B. Xing, Inequitable and heterogeneous impacts on electricity 31. T. G. Reames, A community-based approach to low-income residential energy efficiency
consumption from COVID-19 mitigation measures. iScience 24, 103231 (2021). participation barriers. Local Environ. 21, 1449–1466 (2016).
11. C. Campbell, J. Pearlman, Access to social network support and material hardship. Soc. Currents 32. L. Perl, LIHEAP: Program and Funding, CRS RL31865 (Library of Congress Congressional Research
6, 284–304 (2018). Service, Washington, DC, 2018).
12. R. Finnigan, K. D. Meagher, Past due: Combinations of utility and housing hardship in the United 33. M. Flaherty, S. Carley, D. M. Konisky, Electric utility disconnection policy and vulnerable
States. Sociol. Perspect. 62, 96–119 (2018). populations. Electr. J. 33, 106859 (2020).
13. C. Heflin, A. S. London, E. K. Scott, Mitigating material hardship: The strategies low-income 34. C. Fischer, Renewable portfolio standards: When do they lower energy prices?
families employ to reduce the consequences of poverty. Sociol. Inq. 81, 223–246 (2011). Energy J. (Camb. Mass.) 31, 101–119 (2010).
14. S. Halpern-Meekin, K. Edin, L. Tach, J. Sykes, It’s Not Like I’m Poor: How Working Families Make 35. P. Stott, Climate change. How climate change affects extreme weather events. Science 352,
Ends Meet in a Post-welfare World (University of California Press, Berkeley, CA, 2015). 1517–1518 (2016).
15. K. L. Hanson, L. Connor, C. M. Olson, G. Mills, Household instability and unpredictable earnings 36. P. She, G. A. Livermore, Material hardship, poverty, and disability among working-age adults.
hinder coping in households with food insecure children. J. Poverty 20, 464–483 (2016). Soc. Sci. Q. 88, 970–989 (2007).
16. K. Edin, L. Lein, Work, welfare, and single mothers’ economic survival strategies. Am. Sociol. Rev. 37. L. Vavreck, D. Rivers, The 2006 cooperative congressional election study. J. Elec. Pub. Opinion
62, 253–266 (1997). Parties 18, 355–366 (2008).
17. K. S. Seefeldt, Constant consumption smoothing, limited investments, and few repayments: The role 38. S. Ansolabehere, D. Rivers, Cooperative survey research. Annu. Rev. Polit. Sci. 16, 307–329
of debt in the financial lives of economically vulnerable families. Soc. Serv. Rev. 89, 263–300 (2015). (2013).
18. K. Brunner, M. Spitzer, A. Christanell, Experiencing fuel poverty. Coping strategies of low-income 39. S. Ansolabehere, B. F. Schaffner, Does survey mode still matter? Findings from a 2010 multi-mode
households in Vienna/Austria. Energy Policy 49, 53–59 (2012). comparison. Polit. Anal. 22, 285–303 (2014).
19. R. Chard, G. Walker, Living with fuel poverty in older age: Coping strategies and their problematic 40. S. Carley, M. Graff, D.M. Konisky, T. Memmott, T. Replication Data for: Behavioral and financial
implications. Energy Res. Soc. Sci. 18, 62–70 (2016). coping strategies among energy-insecure households. Harvard Dataverse. https://dataverse.
20. W. Anderson, V. White, A. Finney, Coping with low incomes and cold homes. Energy Policy harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/XKGOCW. Accessed 12 August
49, 40–52 (2012). 2022.

10 of 10 https://doi.org/10.1073/pnas.2205356119 pnas.org