Advertisement

Energy Efficiency

, Volume 11, Issue 8, pp 2017–2032 | Cite as

Cool or school?: the role of building attributes in explaining residential energy burdens in California

  • Hal T. NelsonEmail author
  • Nick Gebbia
Original Article
  • 106 Downloads

Abstract

This paper quantifies the dimensions of an important energy efficiency conversation: the energy burden of low-income households. Due to budgetary constraints, low-income households face a stark consumption trade-offs described as “cool or school”. This study is the first to apply multivariate building energy regression modeling to assess the independent effects of various building characteristics and appliances on a household’s energy burden in the USA. We find that more attic insulation and newer air conditioners significantly predict lower energy burdens. Furthermore, homeowners enjoy ~ 27% more attic insulation compared to renters. Our results offer empirical support for programs that offer deep retrofits to low-income households. We conclude by offering suggestions for leveraging weatherization funding to fund building energy retrofits.

Keywords

Energy consumption Energy conservation Energy efficiency policy Policy implementation Weatherization Housing Poverty Social justice 

Abbreviations

ARRA

American Reinvestment and Recovery Act

DSM

Demand side management

HERO

Home energy renovation Opportunity

IOU

Investor-owned-utility

LIHEAP

Low-income Home Energy Assistance Program

PACE

Property assessed clean energy

RASS

Residential Appliance Saturation Survey

SFR

Single family residence

WAP

Weatherization Assistance Program

Introduction

Energy headlines today tend to center around the appropriate role of renewable energy in the grid or about the relative merits of “fracking”, a means of drilling for oil and natural gas. What has been much less central in the debate is the energy burden of low-income households.1 Energy burden is the percent of household income spent on energy bills. Households with large energy burdens have typically been labeled as living in fuel poverty, typically defined as energy expenditures exceeding 10% of income.

Energy burdens are important because energy services are price inelastic, meaning that it is difficult to reduce demand when prices rise. In order to heat or cool their homes, low-income families must often cut back on food, education expenses, clothing, or other necessities. Consumption decisions between children’s education and thermal comfort can be viewed as a “Cool-or-School” decision. In spite of public education programs in the USA, poor households’ expenditures on education were nearly 3% of total expenditures, with clothing adding a further 3.7% (BLS 2011). In 2011, expenditures on food for US households in the two lowest income quintiles averaged over 15% of total expenditures, which compete with energy expenditures (BLS 2011). These choices have been also summarized as “Heat or Eat” (Bhattacharya et al. 2003).

In addition to tough nutrition and education decisions, the opportunity costs for low-income households from high energy burdens extend to medical problems as well. A recent survey of Americans receiving federal energy assistance showed that

“37 percent [of households] went without medical or dental care, and 34 percent did not fill a prescription or took less than their full dose of prescribed medication. In addition, 19 percent became sick because the home was too cold.” (Choate and Wolfe 2011, p.3)

Low-income households are especially at risk for difficult decisions related to thermal comfort versus medical outcomes, as there has been a pronounced structural trend towards higher health care inequality in the USA. Residential energy conservation efforts have been linked to improvements in general health, sinusitis, and asthma symptoms (Wilson et al. 2013). However, in 2012, those living below the poverty line were roughly four times as likely to delay medical care due to cost, and six times as likely to forego treatment altogether, compared to those above 400% of the poverty line (Burgard and King 2014). The young, elderly, and handicapped are the most at risk when living in energy poverty (Bird and Hernández 2012). In sum, lower-income households require policy and program interventions to reduce their energy burden. Successful interventions are bound to spill over into improved outcomes in other public policy domains such as health and education.

Energy burdens in California and the United States

Our research focuses on California, since that is the source of our data. But prior to investigating the scale of the energy burden problem in the USA and California, we examine its components—income and energy expenditures—independently. The official poverty rate in the USA was 15.0% in 2012, meaning that there were about 46.5 million people below the federal poverty level of $23,000 for a family of four (US Census 2013). In California, the poverty rate was 16.9% as of 2011. However, given California’s high cost of living, the state’s cost of living adjusted poverty rate was even higher, estimated at 23.5% for 2009–2011. Statewide, all households’ mean income for 2010 was $83,483 (US Census 2016).

Given these poverty statistics, now let us look at energy expenditures. Household energy expenditures are shown in Fig. 1 for the USA, the Pacific region (AL, CA, OR, WA), and California. Due in part to higher relative electricity prices, California’s household energy consumption is lower than the US average. Also, much of the state’s population lives on the coast with a moderate climate, which reduces demand for space conditioning energy. However, California’s energy efficiency policies are also likely to have contributed to lower energy expenditures (Ettenson 2011), thus reducing the energy burden of low-income Californians. However, the role of energy efficiency policies in keeping per capita energy consumption flat in California over the last four decades has been contested (Levinson 2013).
Fig. 1

2009 California Household Energy Use Data (source: US EIA 2009, p.1)

The average energy expenditure data hides a wide variation in the share of energy expenditures over total expenditures among income groups. Given inelastic demand for energy, the energy burden falls most heavily on the poor, elderly, and single earner households. As the US Department of Energy (2010) notes, “Low-income households typically spend 14.2% of their total annual income on residential energy costs, compared with 3.3% for other households. The average expenditure in low-income households is estimated to be $1,800 annually” (p.1).

California’s energy burden has analyzed at the state, county, as well as metropolitan statistical area levels. Fisher, Sheehan and Colton (2017) show that for median incomes at 50% or less of the federal poverty level, the energy burden in California was estimated at 24.90% of median income. For the second quartile, the statewide energy burden estimate for 2016 was 13.27%. However, the statewide data masks a sizable range of energy burdens; from with a high of 37.8% (Mono County) and low of 22.4% (Los Angeles County) for the first quartile, and 20.2 to 12.4% for the same counties in the second median income quartile. In a city-level analysis using Census Bureau (2013, 2011) housing survey data, Drehobi and Ross (2016) estimate that California’s largest cities have lower energy burdens than many other cities in the USA. Not only did rich coastal cities like San Diego and San Francisco have low energy burdens for households reporting 80% or less of the area median income (their definition of low income), but so did hotter inland cities like Riverside and Sacramento (p. 21). The favorable results for California were in spite of residential electricity prices that were a whopping 45% higher than the national average (US EIA 2016).

The remainder of this paper examines the factors affecting a household’s energy burden—particularly for low-income households—as well as the implications for energy policies. Section “Literature review” reviews the literature, and section “Data and methods” develops our methods and data. Section “Results” presents the main empirical results. These comprise the impact of various building attributes on energy burdens and on home thermostat settings, as well as the relationship between homeownership and energy burden. Section “Discussion” elaborates on the empirical results and discusses relevant public policy. Section “Conclusion” concludes.

Literature review

For some time, energy burdens have been discussed as fuel poverty. Boardman (1991) gave the first formal definition of fuel poverty as “[Fuel poor households] are unable to obtain an adequate level of energy services, particularly warmth, for 10 percent of its income” (p. 207). However, Moore (2012) is highly critical of measuring fuel poverty as a percent of income. Thomson and Snell (2013) provide an analysis of fuel poverty in the EU. Excellent surveys of definitions of fuel poverty for interested readers are Liddell et al. (2012) and Boardman (2012). More recent research has attempted to quantify the non-heating share of fuel poverty (Simcock and Walker 2015). Built environment and rental market factors have also been shown to affect fuel poverty (Thomson and Snell 2013). Aranda et al. (2017) utilize building energy simulations to identify energy efficiency measures that can reduce energy burdens in multifamily buildings. Our investigation of energy burdens in California attempts to quantify the effects of building shell, heating and cooling equipment, thermostat setting behavior, as well as plug-load equipment on low-income residential energy burdens using building energy regression modeling.

We posit two commonly accepted, stylized facts about energy burdens related to building attributes. First, older homes are less energy efficient because these homes were built prior to building energy codes that reduce energy expenditures. Consider the residential energy reductions from California’s Title 24 Building Energy Efficiency Standard for homes, which is the strictest building energy code in the country. Enacted in 1975, Title 24 has consistently reduced the electricity and fuel consumption per square meter in each successive code version. As a result of Title 24, as well as federal and California appliance energy consumption standards, newer residential buildings are much more energy efficient, which reduces energy expenditures for those households.

Research on building practices have shown much greater penetration of energy efficiency measures for homes built in 2001 or later. For example, nearly all homes built in 2001 or later have wall insulation, compared to about 70% of those built before 2001 (KEMA 2010). In a study that uses an older version of the California dataset utilized here, Gillingham et al. (2012, p.56) find that each 10 years of home age predicts a 4 and 8% smaller probability of attic and wall insulation, respectively. So, ceteris paribus, to the extent that low-income families reside in older homes, they are going to have larger energy bills than their peers in newer accommodations.

Market barriers and failures

The second stylized fact explaining higher low-income energy burdens arise from the economic theories of market barriers and market failures for energy efficiency related to low-income households. The UK government showed that a majority of the households living in fuel poverty lived in the most energy inefficient buildings (DEFRA/BERR 2008, p. 55). Similar results have been found in the US (Cluett et al. 2016). Adan and Fuerst (2016) measure actual outcomes of energy efficiency upgrades in UK residencies that show a substantial reduction in energy burdens. For this paper, we expand on two relevant theoretical mechanisms to explain the energy efficiency gap, or why low-income households are less likely to live in energy efficient housing stock, and thus experience a greater energy burden.

First, barriers of access to capital to fund more efficient measures such as furnaces or windows are especially prevalent in low-income households (Gillingham et al. 2009). Consider that even with utility incentives, participants in demand side management (DSM) programs have to shoulder a large portion of the cost of efficient appliance upgrades. For instance, a new high efficiency furnace might cost $2400 plus installation, but the utility incentive would only cover less than 10% of that cost (Southern California Gas 2014).

The second mechanism to explain high energy burdens in low-income residences is the landlord-tenant market failure. Low-income households are often renters, and therefore face this split incentive market failure: landlords are responsible for investments in more efficient capital stock, but they do not pay the utility bills (Bird and Hernández 2012; Brown 2001). Here, the tenant hires the landlord to provide housing, but high-quality housing services are expensive to provide. As a result, the landlord “shirks” and provides lower-quality services since s/he does not pay the utility bills. This market failure has received considerable theoretical and empirical attention (Murtishaw and Sayanthe 2006; Jaffe and Stavins 1994) and has seen empirical support (de T’Serclaes and Jollands 2007). Davis (2010) demonstrates that landlords that do not pay the electricity bill are less likely to have purchased efficient appliances. Phillips (2012) estimates that landlords’ willingness to pay for better insulation is only one-half that of owner-occupied residents.

Another aspect of the landlord-tenant market failure is temporal: renters are less likely to pay the large, up-front capital costs for energy efficiency equipment because their tenure in each residence is shorter. Homeowners’ average home tenure is 9 years as opposed to 2 years for renters (The Economist 2009). Low-income households cope with high energy burdens by making investments in low-cost efficiency enhancing equipment such as window sealing or heavy curtains that retain winter heat (Brunner et al. 2012). These measures are also more likely to be taken with the renter when their tenure at the rental unit ends.

These sources of market barriers and failures result in large energy inefficiencies in rental housing. In California, Gillingham et al. (2012) find that renter-occupied dwellings are about 20% less likely to have attic insulation, implying that many low-income households pay a disproportionate share of their limited income on energy.

There exists a considerable literature on why residential consumers have not increased the energy efficiency of their housing to its cost-effective level (de T’Serclaes and Jollands 2007; Brown 2001; Hirst and Brown 1990). Social networks can play a role (Southwell and Murphy 2014). Consumers have been shown to have a poor understanding of energy consumption in their homes and the costs of retrofits (Mills and Schleich 2012). Consumers have also shown high discount rates for short run investments (Harris and Laibson 2001) and are averse to losses (Tversky and Kahneman 1992). Uncertainties about paybacks from efficiency investments, coupled with loss aversion combined, can lead to undervalued efficiency markets (Greene 2011).

Finally, low-income households and occupants of poor quality housing can reduce their energy burden in one important way outside of efficiency measures: turning down the furnace in the winter or the AC in the summer, thereby reducing their thermal comfort.2 As noted above, thermal stress is associated with poor health outcomes. The factors driving thermostat-setting behavior can be categorized as follows: (a) external to the occupants such as climate, (b) internal to the occupants such as income and preferences, and (c) building related. Higher income has been shown to be a strong predictor of greater energy use (Vringer and Blok 2007). Occupants “adapt” to indoor temperatures based on these factors when adjusting the thermostat (Nicol and Humphreys 2002). Owner-occupied residences and those with attic insulation are more likely to adapt to unwanted indoor temperatures by changing thermostat settings more often, controlling for other factors (Gillingham et al. 2012). Low-income households cope with high energy burdens by making investments in low-cost efficiency enhancing equipment such as window sealing or heavy curtains that retain winter heat (Brunner et al. 2012).

Data and methods

We now turn to estimating the quantitative effects of various building characteristics on California households’ energy burden. Multivariate building energy regression modeling is applied in order to assess these effects. Specifically, within the sample, we regress household energy burden on building characteristics including attic insulation, age of the A/C, window performance, wall insulation, additional building traits, as well as further controls for household characteristics and income. In this way, independent relationships between the explanatory variables and household energy burden are identified.

Normalizing energy expenditures by income is the commonly accepted definition of energy burden and recognizes the complexities associated with household income and energy expenditures, including the influence of income on housing market choices, the impact of low income housing subsidies, social transfers, and other factors. In sum, this formulation recognizes energy burdens as a systems phenomenon, rather than silo-ing energy expenditures and income.

…. Importantly, we do not include household income as a control variable, as it would be monotonically related to energy expenditures. However, we account for the independent effects of income using multiple instruments for income, such as home size (m2), English language, number of occupants, and whether the head of household attended college. Renter status is included in the full set of explanatory variables, in order to highlight any empirical evidence of the split incentive problem operating through variables we are unable to account for. In an effort to further identify a mechanism linking building attributes with energy burden through energy use, evening thermostat settings are regressed on the same list of variables, with income as an additional control. Finally, upon demonstrating the importance of attic insulation in particular to energy burdens, we regress attic insulation on homeownership as an additional test for split incentive effects.

Typically, engineering simulations are considered the “gold standard” for assessing the impacts of a potential retrofit measure, and there are a host of residential building energy simulation tools available (Polly et al. 2011). These tools have contributed to the development of building energy codes that are specific to unique climate zones that face different heating and cooling burdens (California Energy Commission 2013). However, when estimating the effects of building attributes on energy consumption, a regression-based approach provides two primary advantages over engineering simulations: first, engineering simulations are data intensive, and thereby expensive and impractical for use on a large sample of residences. Information on each building being simulated, including the window area and type, duct location and type, and a host of other location-specific information, is required (Bourassa et al. 2012). Multivariate regression models are more appropriate for large samples of houses and have proven reliable enough to be used for post building retrofit measurement and verification studies (Kissock et al. 2003).

Second, our regression approach is operationally based, employing information on actual household energy use, whereas asset-based engineering models may systematically over- or under-predict actual energy consumption (Earth Advantage Institute 2009). Operationally based measures allow us to estimate the effects of inefficient building envelope measures or appliances net of the effects of occupancy and behavioral factors that impact energy consumption. Occupancy and behavior account for a large share of actual household energy consumption, evidenced by increasing numbers of DSM behavioral efficiency programs (Mazur-Stommen and Farley 2013).

Data

We utilize the 2009 California Residential Appliance Saturation Survey (RASS) (KEMA 2010). The RASS survey data is unique in that includes highly detailed customer demographics, building envelope and appliance data, as well as actual (not self-reported) 12-month natural gas and electricity usage. The survey was administered in 2009–2010 to 25,721 households in California Investor-Owned-Utility (IOU) service territories. California IOUs have inclining block rate tariff designs that increase the $/kWh electricity and $/therm natural gas (100,000 British Thermal Units) rate for a series of tiers above the baseline amount developed for each climate zone. We also collect data on each IOU tariff rate, baseline, and tier price in order to estimate electricity and natural gas bills for each customer in the database. Finally, we match each respondent’s zip code and billing period to National Oceanic and Atmospheric Agency historical daily climate data to control for heating and cooling energy consumption in the regression models. In the process of our analysis, we convert some categorical variables in the survey data to continuous variables by assigning midpoints of recorded ranges and top- or bottom-coding extreme categories.3

Consistent with our focus on low-income households, we limit the RASS sample to include households in the bottom two of the income quartiles. Our sample only includes households situated in inland California, away from the temperate coastline that reduces space heating and cooling energy consumption. More information on our climate zones (1–8) employed in the analysis can be found at the CEC (2015) website. Our focus on single family residences (SFR) excludes condos and townhouses that have shared walls that might bias our results. We excluded 54 houses in the inland sample with above 4700 square feet of floor space to reduce the effects of outliers. Finally, we include only households with individual meters, thereby excluding households that do not directly pay for their own energy consumption. While the RASS data include some imputed values for missing data, we do not include these observations as they impose an unknown relational structure on the data prior to our analysis. Our final sample includes 969 households; of these, 597 have recorded thermostat setting data, which provides our subsample for thermostat setting regressions.

Table 1 shows average electricity and natural gas bills of $1269 and $619 for the sample, respectively. This represents an average energy burden of about 7% of income, given the mean sample income of $35,562. The average single family home in our sample is 37 years old, has roughly 148 m2 (1600 ft2) of living space, enjoys 10.2 cm (4.0 in.) of attic insulation, and uses heating and cooling systems that are, on average, 13 and 11 years old, respectively.
Table 1

Descriptive statistics

 

Mean

Min

Max

SD

Skewness

Kurtosis

Fuel bills and income

Annual electricity bill, estimate

$ 1269

$26

$7514

$1008

2

8.5

Annual gas bill, estimate

$ 619

$38

$1877

$284

1.1

5

Annual total fuel bill, estimate

$1888

$176

$7729

$1125

1.6

6.3

Annual income

$ 35,562

$5000

$55,000

$14,345

− .18

2.1

Fuel fraction of income

.069

.0032

.81

.077

4.9

37

Home characteristics

Square footage of home

1643

375

4700

555

1

5.4

Age of home

37

1

73

20

− .11

2.1

Years of residence

16

1

30

11

.072

1.4

Home number of stories

1.2

1

3

.41

2.2

7

Inches of attic insulation

4

0

10

2.9

.31

2.2

Window panes

All or most windows are single pane

.37

0

1

.48

.56

1.3

All or most windows are double pane

.53

0

1

.5

− .1

1

Mixture of single and double pane windows

.11

0

1

.31

2.5

7.3

Exterior wall insulation

All exterior walls insulated

.56

0

1

.5

− .23

1.1

Some exterior walls insulated

.2

0

1

.4

1.5

3.3

No exterior walls insulated

.24

0

1

.43

1.2

2.4

Appliances

Age of primary refrigerator (years)

7.4

1

20

5.1

.76

2.5

Total number of televisions

2.5

0

7

1.3

.78

3.9

Total number of desktop PC’s

.91

0

3

.7

.74

4.1

Heating and cooling

Age of main heating system (years)

13

.5

30

9.7

.41

1.7

Age of main AC system (years)

11

.5

30

8.4

.77

2.4

Number of room air conditioners

.36

0

3

.71

2.1

7.1

Number of central air conditioners

.92

0

3

.45

− .076

6.2

Main heating is electric (vs. gas)

.068

0

1

.25

3.4

13

Demographics

Primary language: English

.9

0

1

.3

− 2.6

7.9

Head of Household Has Some College

.76

0

1

.43

− 1.2

2.4

Number of household occupants

2.6

1

8

1.6

1.3

4.2

Electricity provider

Pacific gas and electric

.32

0

1

.47

.79

1.6

San Diego Gas and Electric

.068

0

1

.25

3.4

13

Southern California Edison

.51

0

1

.5

− .035

1

L.A. Dept. of Water and Power

.11

0

1

.31

2.6

7.5

Gas provider

Pacific gas and electric

.3

0

1

.46

.89

1.8

San Diego Gas and Electric

.069

0

1

.25

3.4

13

Southern California Gas Company

.63

0

1

.48

− .56

1.3

Notes: Sample is constant and contains 969 observations

Results

The results of the ordinary least square (OLS) regressions with the dependent variable of energy burden are presented in Table 2. Due in part to categorical recording of income, thermostat setting, and insulation measures in the data, Shapiro-Wilk tests show non-normality in the residuals of the regressions. In order to ensure that reported levels of statistical significance are not influenced by non-normality, we utilize bootstrapped standard errors for all models presented below.
Table 2

Natural Gas and Electricity Bills as a Fraction of Income

 

(1)

(2)

(3)

(4)

(5)

Inches of attic insulation

− 0.00214***

− 0.00262***

− 0.00252***

− 0.00248**

− 0.00234**

 

(0.007)

(0.006)

(0.009)

(0.011)

(0.017)

Mixture of single and double pane windows

 

0.000911

0.000574

0.00141

0.00230

  

(0.886)

(0.928)

(0.826)

(0.719)

All or most windows are double pane

 

0.00138

0.00140

0.00138

0.00266

  

(0.815)

(0.811)

(0.813)

(0.649)

Some exterior walls insulated

 

− 0.00227

− 0.00369

− 0.00438

− 0.00442

  

(0.776)

(0.644)

(0.583)

(0.577)

All exterior walls insulated

 

0.0107

0.00974

0.00938

0.0101

  

(0.236)

(0.279)

(0.300)

(0.265)

Age of main heating system (years)

  

− 0.000427

− 0.000419

− 0.000426

   

(0.227)

(0.231)

(0.220)

Age of main AC system (years)

  

0.000725*

0.000685*

0.000686*

   

(0.080)

(0.090)

(0.088)

Age of primary refrigerator (years)

  

− 0.000798*

− 0.000837*

− 0.000809

   

(0.064)

(0.052)

(0.060)

Number of room air conditioners

  

0.00291

0.00253

0.00189

   

(0.481)

(0.533)

(0.644)

Main heating is electric (vs. gas)

  

− 0.00109

− 0.000751

− 0.00246

   

(0.897)

(0.929)

(0.776)

Total number of televisions

   

0.00346*

0.00379**

    

(0.071)

(0.046)

Total number of desktop PC’s

   

0.00422

0.00482

    

(0.242)

(0.173)

Own/buying (vs. rent/leasing)

    

− 0.0224**

     

(0.011)

Constant

0.0107

0.00313

0.00762

0.00102

0.00178

 

(0.607)

(0.889)

(0.748)

(0.965)

(0.474)

Observations

969

969

969

969

969

Controls

Yes

Yes

Yes

Yes

Yes

R 2

0.0896

0.0940

0.0999

0.105

0.111

RMSE

0.0739

0.0739

0.0738

0.0737

0.0735

Notes: p values in parentheses. Base for windowpane indicators is “All or most windows are single pane”. Base for exterior wall insulation indicators is “No exterior walls insulated”

*p < 0.10, **p < 0.05, ***p < 0.01

The models’ F-statistics (not reported) show that the combined effects of the variables in each of the models are significantly different from zero. Controls not shown in the regressions include house size (m2) and age, household occupancy, climate zone, and whether or not English was the primary language spoken in the household. We also include dummy variables for each of the electricity and natural gas utilities. The models explain between 9 and 11% of the variation in the dependent variable of annual energy burden, which we consider respectable since energy expenditures are normalized by the income variable, which is initially recorded as categorical.

Model 1 in Table 2 includes only attic insulation and the vector of controls defined above and explains nearly 9% of the variation in energy burdens. Model 2 also includes SFR building envelope measures whose coefficients are not significantly different from zero and add little to the variation in energy burdens explained. While there is likely to be some measurement error in these indicators as respondents might not know about wall insulation levels, better thermal efficiency of walls and windows did not significantly reduce SFR energy burdens. The − .0026 coefficient for attic insulation in model 2 predicts that if a house without attic insulation were to be retrofitted to the mean level of 4.0 in. (10.2 cm), energy burden would be reduced by about one percentage point, or about a 13% reduction in annual energy burden for the mean household currently without insulation. This estimate is independent of the vector of control variables, as well as the other building envelope measures.

Model 3 shows the evidence of the relationship between the age of prime movers and large appliances with energy burden. The results for the age of heating equipment is not significant. Older refrigerators significantly predict lower energy burdens, which is counter-intuitive. The age of the main A/C system measures the efficiency of the system. Older A/C’s consistently predict higher energy burdens. The significant results predict that replacing an 11 year-old A/C system (the mean age of A/C’s in the sample) with a new unit could result in a reduced energy burden of nearly .008%. This represents a decrease of over 11% in the average energy burden.

Model 4 adds plug loads and shows that each additional TV is linked with a 3/10ths of a percent increase in energy burden, a surprisingly high impact. Since income is not a control variable (it is used to normalize energy expenditures), it is possible that the number of TV’s is picking up variation in energy expenditures via income.4 As such, controlling for number of TVs means that the other variables in the model reflect only their own independent effects.

Model 5 reflects the effects of the split incentive between owners and renters, operating through covariates which are unmeasured in the models, and for which we are therefore unable to control. The results show that the average renter’s annual energy burden is about 2.2 percentage points (34%) higher than that for a homeowner, controlling for house age, m2, climate, and other covariates. We interpret this as reflective of split-incentive effects operating through factors outside our observation, such as poorer equipment maintenance or unsealed leaks for rented homes. The owner vs. rental indicator is by far the most substantive covariate in the results and reflects the power of the split incentive market failure in explaining higher energy burdens in rental properties. Also worth noting is that the estimated coefficients for attic insulation and A/C remain statistically significant and stable across all model specifications, demonstrating that the relationship between attic insulation and A/C age with energy burdens is robust.

Next, Table 3 adds in the average evening thermostat settings to analyze the tradeoffs in consumption baskets that low-income households face. Recall that in many low-income households, a comfortable indoor climate is a luxury that cannot be afforded: “Heat or Eat” and “Cool or School” show the competition for caloric intake or educational expenditures, respectively (Bhattacharya et al. 2003). We use evening thermostat settings, as this is when space conditioning is most often required. In many higher inland California locations, overnight cooling might not be needed in the summer. Conversely, in more temperate inland regions, overnight heating may not be needed. Similarly, evenings are when houses are most likely occupied by residents as opposed to daytime thermostat settings.
Table 3

Home thermostat settings (evening)

 

(1) Heating

(2) Cooling

Annual income (thousands)

0.0799

− 0.0112

 

(0.237)

(0.813)

Annual income (thousands) squared

− 0.00134

0.000252

 

(0.153)

(0.701)

Inches of attic insulation

− 0.114

0.133***

 

(0.103)

(0.007)

Mixture of single and double pane windows

0.976

− 0.488

 

(0.124)

(0.286)

All or most windows are double pane

0.388

− 0.0928

 

(0.443)

(0.776)

Some exterior walls insulated

− 0.951

− 0.156

 

(0.111)

(0.700)

All exterior walls insulated

− 0.823

− 0.0850

 

(0.146)

(0.832)

Age of main heating system (years)

0.0159

− 0.00928

 

(0.720)

(0.762)

Age of main AC system (years)

− 0.0403

0.00816

 

(0.400)

(0.798)

Age of primary refrigerator (years)

0.0716**

0.0204

 

(0.043)

(0.404)

Number of room air conditioners

− 0.515

− 0.474

 

(0.171)

(0.134)

Main heating is electric (vs. gas)

1.188

− 0.498

 

(0.126)

(0.325)

Total number of televisions

0.178

0.144

 

(0.245)

(0.135)

Total number of desktop PC’s

0.00148

0.172

 

(0.996)

(0.371)

Own/buying (vs. rent/leasing)

− 0.536

− 0.0612

 

(0.421)

(0.891)

Constant

68.35***

72.54***

 

(0.000)

(0.000)

Observations

597

597

Controls

Yes

Yes

R 2

0.0892

0.160

RMSE

4.364

2.956

Notes: p values in parentheses. Base for windowpane indicators is “All or most windows are single pane”. Base for exterior wall insulation indicators is “No exterior walls insulated”

*p < 0.10, **p < 0.05, ***p < 0.01

Table 3 explains nearly 9 and 16% of variation in heating and cooling season thermostat setting behavior, respectively. For the lower-income households included in the sample, annual income by itself is not a significant predictor of either heating or cooling behavior. Although the coefficients have the theoretically consistent signs, controlling for a quadratic relationship with income via income-squared is also not significant. This result is consistent with Gillingham et al. (2012).

Model 1 in Table 3 shows no statistically significant effect from attic insulation on evening heating (furnace) settings. Model 2 predicts that an additional 4.0 in. (10.2 cm) of attic insulation is associated with setting the thermostat roughly and half a degree (F) higher for cooling purposes—resulting in lower energy bills.

Finally, Table 4 provides evidence toward the effect of the split incentive problem on levels of attic insulation. Even after controlling for differences in income, renters have nearly 3/4ths of an inch (1.9 cm) less insulation than homeowners, constituting an 18% shortfall for the average renter. It also shows that for the low-income households in the sample, more window A/C units are associated with less attic insulation, all other factors held constant.
Table 4

Split incentive effects on levels of attic insulation

 

(1)

Own/buying (vs. rent/leasing)

0.73**

 

(0.018)

Annual income (thousands)

0.0366

 

(0.227)

Annual income (thousands) squared

− 0.000370

 

(0.382)

Age of main heating system (years)

− 0.0173

 

(0.133)

Age of main AC system (years)

− 0.00901

 

(0.502)

Age of primary refrigerator (years)

0.0170

 

(0.337)

Number of room air conditioners

− 0.390***

 

(0.003)

Main heating is electric (vs. gas)

0.134

 

(0.741)

Total number of televisions

− 0.125*

 

(0.052)

Total number of desktop PC’s

0.128

 

(0.351)

Constant

1.759*

 

(0.067)

Observations

969

Controls

Yes

R 2

0.180

RMSE

2.709

Notes: p values in parentheses

*p < 0.10, **p < 0.05, ***p < 0.01

Of vital importance to understanding our results is the potential effect of measurement error, resulting from the categorical nature of the initial data. In an OLS regression setting, it is well known that measurement error in the outcome variable (energy burden) will attenuate point estimates toward zero, while measurement error in the regression will decrease efficiency, thereby increasing standard errors (Gujarati and Porter 2009, pp. 483–484). As a result, we consider the standard errors for home age and square-footage, age of heating and cooling systems, refrigerator age, and attic insulation to be potentially overstated. Furthermore, as our dependent variable is affected by the measurement error inherent in categorical recording of income, our coefficient estimates are likely attenuated toward zero. This increases the risks of a type II error, or falsely rejecting significant effects of the explanatory variables. Combined, these two considerations about measurement error should give the reader additional comfort about the validity of the statistically significant variables in the Table 4.

Discussion

Public policies designed to help alleviate high energy burdens fall into two main categories: weatherization and payment assistance. For example, the US Department of Energy’s (DOE) Weatherization Assistance Program (WAP) retrofits existing homes to make them more efficient and claims $1.80 in energy savings for each $1.00 invested (Eisenberg 2010). The federal government also provides other funding to states through the State Energy Program. In addition to federal funding, 26 states have adopted long-term energy efficiency policies that can contribute to low-income weatherization or payment assistance (ACEEE 2014).

However, compared to the scope of the problem, WAP has been underfunded since its inception under the Energy Conservation and Production Act of 1976. The American Reinvestment and Recovery Act PL111–5 (ARRA) 2009 stimulus funding quintupled existing weatherization funding by allocating $5B to the program. After a slow start, the ARRA funding weatherized over a million homes (Caperton et al. 2012). For example, Linda Higgins got a new front door, a furnace to replace her electric space heaters, and two new windows. These measures reduced the Higgins’ monthly utility bill to about $60 from more than $200 (Amaro 2012). Yet, the Higginses are only 1 of the 38 million households that are eligible for federal weatherization services, with eligibility contingent on a household’s annual income falling at or below 200% of the poverty level.

In addition to weatherization programs, payment assistance programs—such as the Federal Low-income Home Energy Assistance Program (LIHEAP)—help to address the energy burden for low-income households. LIHEAP was originally authorized by the Omnibus Budget Reconciliation Act of 1981 and consists of block grants to the states administered by the Department of Health and Human Services (HHS). The LIHEAP program is also dramatically underfunded. HHS estimates that only 14% of eligible households receive heating benefits from LIHEAP, down from 23% in 1981 (HHS 2009). Nearly 80% of LIHEAP funding has gone to heating and cooling payment assistance (Murray and Mills 2014).

Overall, in the USA, spending toward payment assistance programs dwarves that for weatherization programs: weatherization spending is estimated at only 8% of payment assistance (Heffner and Campbell 2011). Our results are suggestive of the potential efficacy of expanding support for weatherization programs.

Converting payment assistance into comprehensive weatherization funding

Our findings from the regression analyses show that across a range of specifications and dependent variables, more attic insulation and A/C replacement both predict substantively lower energy burdens. Furthermore, attic insulations predict greater thermal comfort. Our results also reinforce the results of previous research that has examined market failures in residential house markets (Gillingham et al. 2012; Murtishaw and Sayanthe 2006). We show that rented homes on average have 18% less attic insulation compared to similar, owned homes. Even while controlling for the effect of attic insulation disparities in our main regressions, we find that energy burdens remain 34% higher for renters, which we attribute importantly to the split incentive effect. The policy implication is that there is a large pool of renters that are suffering from an excessive energy burden. This indicates many lower income SFR’s in inland California that are strong candidates for cost-effective, deep retrofits that include additional attic insulation and A/C replacement.

Our claim that the poor efficiency of rental SFR’s is driven by structural market barriers and failures implies that policy interventions will be required if policymakers want to mitigate the problem. Policy mechanisms to overcome these structural problems have been addressed by Murtishaw and Sayanthe (2006) and Bird and Hernández (2012), among others. The main policy implication stemming from the building energy regression modeling and low-income analysis is to convert bill assistance (consumption expenditures) into weatherization funding (conservation expenditures). By targeting the most energy inefficient houses with cost-effective weatherization measures, reduced utility bills in low-income and rental SFRs can be ensured over time. Bill reductions will occur through better building performance rather than subsidies that increase energy consumption and associated air pollutants and carbon dioxide emissions. While not every low-income household can be retrofitted due to older wiring or vermiculite, bill assistance expenditures from these homes can be redirected to other needy low-income households.

The policy inefficiencies of large consumption expenditures for qualifying households can be conceived of as a perpetual release of dollars out of the attics (and doors and walls) of these residential buildings. For example, Fig. 2 shows an infrared photo comparison of buildings with and without weatherization measures. The left photo shows winter heat losses as lighter areas around the eaves, windows, and some of the wall areas in the baseline case. Once a retrofit occurs as in the right photo, the heat losses are reduced dramatically.
Fig. 2

Infrared imaging of housing heat losses (Source: Sagewell 2014. Reprinted with Permission) (Color on web only)

Funding consumption subsidies (LIHEAP) over weatherization spending (WAP) at an 11:1 ratio might have made sense in the high interest rate environment of the 1970s and 1980s. However, with current public (and private) financing costs of 2–5%, large consumption subsidy to weatherization ratios does not make sense, and low-income programs should target weatherization funding instead. Consider that the interest payment on a $6500, 20-year loan for weatherization at a 10% interest rate is $63 a month, but at a 2% interest rate, the monthly payment is only $33.

To illustrate the costs and benefits of converting subsidies to building retrofits, in Table 5, we selected a basket of illustrative weatherization measures based on the descriptive statistics data in Table 1. The illustrative household is assumed to be 1600 ft2 (154 m2) with an annual energy expenditure of $1800, our sample averages. This house requires new attic insulation, A/C replacement, as well as air leakage control measures such as a blower door test, weather stripping, and caulking. We also assume duct sealing as a space conditioning efficiency improvement. The cost of the energy efficiency measures come from the NREL National Residential Efficiency Measures Database (2014). We use end-use energy shares from the LIHEAP Report to Congress that estimate space conditioning for LIHEAP homes at 43% of total energy use and water heating at 11% of energy use (US DHHS 2014, p. 25).
Table 5

Illustrative weatherization measure costs and annual energy savings

Measure name

Description

Savings (%)

End use

Energy savings ($/year)

Initial cost unit

Total cost

Notes

Attic insulation

Unisulated to R-30 (5.5″/14 cm)

20%

Space Conditioning

$154

$3.6 ft2/($38.75 m2)

$4133

Assumes 20% of homes have 2 levels

Energy Audit including Blower Door Test

Measure air turnover, efficiency opportunities

Included in Other Measures

Space Conditioning

Included in Other Measures

$400

$400

 

Weather Stripping and Caulking

15 ACH to 8 ACH

15%

Space Conditioning

$116

$1.6 ft2/($17.22 m2)

$371

Assumes 2 exterior doors and 12 4 × 4 windows

Insulate and Seal Ducts

Uninsulated to R-8 insulation: 30 to 7.5% losses

22%

Space Conditioning

$170

$2.3 ft2 ($24.75) duct surface

$722

Assume 1600 ft2 house with 100 linear feet of ducts

Low Flow Faucets and Showerheads

Replacing faucets and showerheads to 2.0 gpm

40%

Non-Space Conditioning Energy

$45

$6 per aerator, $10/showerhead

$32

Replacing 4 old 3.5 and 5.0 gpm showerheads and aerators with 1.5 and 2.0 gpm units

A/C Replacement

Central air replacement from 10 SEER to 21 SEER

35%

Space Conditioning Electricity

$178

$860

$860

Incremental cost of new high efficiency central A/C

Total

   

$663

 

$6519

 

The analysis in Table 5 suggests that average energy savings from the basket of illustrative weatherization measures could be approximately $663 a year or $55 a month. At a 2.0% interest rate, the monthly payments on the loan would be $33 (or $43 a month at 5.0% interest). The average benefits to the LIHEAP program from avoided payment assistance would be approximately $40 per month, assuming 2009 fuel assistance of $3.99B with 8.3 M recipients (US DHHS 2014, pp. 32–33). This analysis indicates that social improvements are likely to be obtained from reducing household energy burdens through increased weatherization funding rather than payment assistance. Nevertheless, this is a complex issue with many tradeoffs in the bill assistance vs weatherization debate. The goal of this analysis is to bring forth evidence on the impact of building attributes on energy burdens and how that can inform the policy domain.

Mechanisms for converting payment assistance into weatherization

In spite of the potential net benefits presented in Table 5, the simple payback of the weatherization package is roughly 10 years. The retrofit requires capital investment of $6500 and an additional pool of monies is required.

The USA can learn from other jurisdictions that are addressing energy burdens by weatherizing poorly performing buildings. The UK became more concerned about fuel poverty after fears of rising energy prices following its entry into the European Union’s Emissions Trade Scheme. England is attempting to reduce fuel poverty by 2016 by retrofitting “as many fuel poor as is reasonably practicable” to be more energy efficient (HM Government 2014). In recent years, there has been concern that the UK budget for tackling fuel poverty has been cut too much (Association for the Conservation of Energy 2012). A large portion of fuel poverty program funding, including payment assistance efforts, has traditionally been funded by a levy on utility sales.

Utility levies for low-income programs are also common in the USA. In 2010, 35 states reported nearly $2 billion in low-income assistance, most of which was spent on payment assistance rather than weatherization. Between 1992 and 2010, US states were encouraged to leverage non-federal funding for LIHEAP payment assistance or weatherization funding. Eligible activities must have resulted in quantifiable home energy benefits. As of 2015, states can set aside up to 40% of their LIHEAP funds for energy efficiency. The most recent year for which data was available (2006) showed that only about 6% of LIHEAP funding was converted to WAP programs for energy efficiency (Cluett et al. 2016, p.6). The lack of federal weatherization funding as led jurisdictions to leverage funding from regional home loan banks to perform low-income weatherization (Nelson 2014). Other regional sources of public weatherization funding include the Regional Greenhouse Gas Initiative where stakeholders in those states can utilize climate change auction revenues dedicated to energy efficiency for low-income weatherization. California stakeholders can leverage the Assembly Bill 32 (AB32) climate change auction proceeds dedicated to disadvantaged residents into weatherization program funding.

However, given the scale of the market barriers and failures in low-income and rental markets, public funding is unlikely to ever be large enough to mitigate the inefficiencies. Private investment is required, and public policies must help overcome existing barriers to funding. One retrofit program that has been successful in leveraging private funding is the Home Energy Renovation Opportunity (HERO) program that originated in Riverside County in Southern California. HERO has helped fund 20,000 residential efficiency projects totaling more than $375 M and is spreading to over 213 cities and unincorporated areas in California (Hales 2014). HERO is a Property Assessed Clean Energy (PACE) provider that uses property taxes as a vehicle to make payments on loans for energy efficiency (and renewable energy) projects. PACE loans in California have surged since the legislature approved a modest reserve fund to reimburse mortgage holders in the event of PACE default. PACE programs also help mitigate the temporal market failure discussed in the theory section above by tying weatherization borrowing to long-term loans that are still in service even after if residence is sold during the life of the loan.

Overcoming the landlord-tenant market failure also difficult. Landlords are typically responsible for making weatherization investments, but typically do not pay the utility bills. In theory, on-bill financing can help mitigate market failures as a loan program that uses utility bills as a vehicle to pay back weatherization project borrowing. To facilitate the deployment of this mechanism, a reserve fund for defaults akin to the PACE fund needs to be set up to alleviate lenders’ concerns about low-income households’ ability to repay the principal. In spite of the theoretical benefits of on-bill financing in reducing the split incentive problem, it is not clear that landlords’ perceived benefits would outweigh their transaction costs associated with initiating and supervising a weatherization project on their property. Interested readers are referred to Bird and Hernández (2012) who analyze the barriers to on-bill financing for low-income residences in detail.

In spite of the opportunities from voluntary financing of low-income and rental building retrofits, there are substantial reasons to believe a minimum performance standard for hard-to-reach households might be required (Bradbook 2010). California has had ratepayer-funded low-income weatherization programs in place for decades. Yet, utilities consistently have difficulty in recruiting adequate participation and spending the authorized funding. For example, in 2013, San Diego Gas and Electric carried forward 20% of its annual low-income Energy Savings Assistance budget. By then, the accumulated carry-forward balance was equal to 1-year program budget. This carry-forward occurred despite sophisticated targeted marketing based on zip+4 demographic characteristics by the utility (Hassan 2014, Table 12, and p.8).

Mandatory building energy labeling schemes hold promise to help reduce the landlord tenant market failure by making building energy efficiency information available to renters prior to lease signing. Mandatory labeling policies in the USA exist primarily for the large commercial sector. Seattle and Austin (TX) have mandatory energy ratings for single family and 5 or more unit multifamily residential properties, respectively (Nadkarni and Micheals 2012). New York State has a “truth in heating” law where residential property owners must disclose historical heating and cooling costs to prospective (not current tenants) upon written or oral request. In contrast to the piecemeal efforts in the USA, the European Union countries are much further ahead on mandatory building labeling. This terse review of the issues associated with overcoming landlord-tenant market failure is intended to highlight the opportunities and barriers associated with making low-income and rental buildings more energy efficient.

Conclusion

Our analysis is the first that we are aware that uses multivariate methods to identify how building attributes affect energy burdens in the interior USA. Regression analysis allows us to estimate the independent effects of each variable, controlling for the effects of other relevant factors. We focus in on how building attributes affect energy burdens, but we also examine their effect on thermostat setting behavior by occupants. We find potentially significant economic and social benefits from programs that would weatherize older houses which may be under-insulated due to the split-incentive market failure. Our results support the need for programs such as Energy Upgrade California that support deep retrofits such as A/C replacement and attic insulation. Attic insulation is one of the most cost-effective energy efficiency measures (US DOE 2012). Our results show that energy burdens from such comprehensive weatherization programs can reduce low-income energy burdens by more than 24% (for attic insulation and A/C replacement alone).

Our sample includes only inland California, single-family homes, which potentially limits the generalizability of our findings to other inland regions of the USA. Utility prices are higher in California than in many other parts of the USA. Space cooling is certainly more important in inland California than in the UK, or other regions, that have studied energy burdens for decades.

However, while energy prices and incomes differ, the design of framed housing has typically not changed much over the decades between US regions. Typically, 2 × 4 inch (5 cm × 10 cm) studs were used for the exterior walls and to construct the roof and attic in the existing building stock. The stud cavities were typically not filled with cellulosic or fiberglass until the 1970s. In the USA, the 1995 Model Energy Code gave “recommended” attic insulation levels that varied by climate zone. As of 2009, nearly two of three residential buildings in the USA were built before 1980 (US EIA 2013). These are the structures that are more likely to not have adequate, or any, attic insulation. There are over 10 million SFR rentals in the USA. Improving the efficiency of these buildings could dramatically reduce national energy consumption and carbon emissions.

…. Given that low-income households face important opportunity costs for their energy expenditures, household budget constraints imply that utility expenditures could be better spent on medical, nutritional, or educational goods. Cool-or-School is a reality for low-income households in our sample. Extreme indoor temperatures predict health problems and excess mortality (Howden-Chapman 2007). Because of the links between higher energy burdens and worse socio-economic outcomes, coupled with our findings of substantial cost-effective attic insulation and A/C replacement, we posit that deep retrofit programs should be considered a key mechanism to improve social justice outcomes.

Not only can low-income weatherization programs improve social justice, they are also the basis for welfare enhancing energy policies. Win-win energy policies are possible when extensive market failures are present (Levinson and Niemann 2004). The correction of the market failures in the residential building sector would include an increase in social welfare, as well as a reduction in energy consumption and associated pollution. Additional study in the area of energy efficiency and conservation in rental markets will add to existing work relating to the evaluation of the direct and spillover market effects of energy efficiency policies, which could lead to better implementation in the future.

Footnotes

  1. 1.

    A Google News search in June of 2017 yielded the following results: “Renewable Energy” = 2,510,000 results, “Fracking” = 721,000 results. “Energy Burden” = 467 results.

  2. 2.

    A complete review of the factors predicting building climate control behavior, and related window opening behavior, is beyond the scope of this paper. Interested readers are referred to Fabi et al. (2012).

  3. 3.

    Variables converted from categorical include income, home age and square-footage, attic insulation, age of heating and cooling systems, refrigerator age, and thermostat settings.

  4. 4.

    More TV’s likely correlates with both higher income and thereby higher energy expenditures, and so is related to energy burden by both the numerator and the denominator. However, since energy expenditures are more precisely measured due to categorical recording of income, it is sensible that more TV’s would be correlated with a higher energy burden in our data.

Notes

Acknowledgements

We would like to thank the participants in the 2014 Western Political Science Association annual meeting for their comments on an earlier version of this paper. This research was partially funded under California Energy Commission Public Interest Energy Research grant # 57356A/11-1. The sponsors played no part in the study design; collection, analysis, and interpretation of data; writing of the article; or the decision to submit for publication.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.

References

  1. ACEEE. (2014). State Energy Efficiency Resource Standards (EERS). February. http://www.aceee.org/files/pdf/policy-brief/eers-02-2014.pdf.
  2. Adan, H., & Fuerst, F. (2016). Do energy efficiency measures really reduce household energy consumption? A difference-in-difference analysis. Energy Efficiency, 9, 1207–1219.  https://doi.org/10.1007/s12053-015-9418-3.CrossRefGoogle Scholar
  3. Amaro, Y. (2012). Weatherization Program Keeps Cold at Bay. Merced Sun-Star. December 18.Google Scholar
  4. Aranda, J., Zabalza, I., Conserva, A., & Millán, G. (2017). Analysis of energy efficiency measures and retrofitting solutions for social housing buildings in Spain as a way to mitigate energy poverty. Sustainability, 9(10), 1869.CrossRefGoogle Scholar
  5. Association for the Conservation of Energy. (2012). National Fuel Poverty Budgets. May. http://www.ukace.org/wp-content/uploads/2012/11/ACE-Briefing-2012-05-National-fuel-poverty-budgets.pdf. Accessed 1 November 2014.
  6. Bhattacharya, J., DeLeire, T., Haider, S., & Currie, J. (2003). Heat or eat? Cold-weather shocks and nutrition. American Journal of Public Health, 93(7), 1149–1154.CrossRefGoogle Scholar
  7. Bird, S., & Hernández, D. (2012). Policy options for the split incentive: Increasing energy efficiency for low-income renters. Energy Policy, 48, 506–514.CrossRefGoogle Scholar
  8. Boardman, B. (1991). Fuel poverty: From cold homes to affordable warmth. London: Belhaven Press.Google Scholar
  9. Boardman, B. (2012). Fuel poverty synthesis: Lessons learnt, actions needed. Energy Policy, 49(0), 143–148.Google Scholar
  10. Bourassa, N. J., Rainer, L., Mills, E., & Glickman, J. (2012). The Home Energy Scoring Tool: A Simplified Asset Rating for Single Family Homes. Proceedings of the 2012 ACEEE Summer Study on Energy Efficiency in Buildings.Google Scholar
  11. Bradbrook, A. J., & Gardam, J. G. (2010). Energy and poverty: A proposal to harness international law to advance universal access to modern energy services. Netherlands International Law Review, 57(1), 1–28.Google Scholar
  12. Brown, M. A. (2001). Market failures and barriers as a basis for clean energy policies. Energy Policy, 29(14), 1197–1207.CrossRefGoogle Scholar
  13. Brunner, K. M., Spitzer, M., & Christanell, A. (2012). Experiencing fuel poverty. Coping strategies of low-income households in Vienna/Austria. Energy Policy, 49, 53–59.CrossRefGoogle Scholar
  14. Bureau of Labor Statistics (BLS). (2011). Consumer Expenditures Survey. Table 1. Downloaded from: http://www.bls.gov/cex/csxstnd.htm. Accessed 14 November 2014.
  15. Burgard, S., & King, M. (2014). Health inequalities. Pathways: State of the Union: the Poverty and Inequality Report. Special Issue. http://web.stanford.edu/group/scspi/_media/pdf/pathways/special_sotu_2014/Pathways_SOTU_2014.pdf.
  16. California Energy Commission. (2013). Residential Compliance Manual. Chapter 4. HVAC. http://www.energy.ca.gov/2013publications/CEC-400-2013-001/chapters/04_Building_HVAC_Requirements.pdfu.
  17. California Energy Commission. (2015). California Energy Maps: Building Climate Zones, 2015. http://www.energy.ca.gov/maps/renewable/building_climate_zones.html Accessed 29 May, 2017.
  18. Caperton, R., James, A., & Kasper, M. (2012). Federal Weatherization Program a winner on all counts. Center for American Progress. 28 September. http://www.americanprogress.org/issues/green/news/2012/09/28/39115/federal-weatherization-program-a-winner-on-all-counts/.
  19. Census Bureau. (2011). American Housing Survey 2011 public use file. http://www.census.gov/programs-surveys/ahs/data/2011/ahs-national-and-metropolitan-pufmicrodata.html. Accessed November 2015.
  20. Census Bureau. (2013). American Housing Survey 2013 Metropolitan public use file. http://www.census.gov/programssurveys/ahs/data/2013/ahs-2013-public-use-file--puf-/2013-ahsmetropolitan-puf-microdata.html. Accessed November 2015.
  21. Choate, J., & Wolfe, M. (2011). National Energy Assistance Survey. November. http://neada.org/wp-content/uploads/2013/05/NEA_Survey_Nov11.pdf.
  22. Cluett, R., Amann, J., & Ou, S. (2016). Building better energy efficiency programs for low-income households. ACEEE Report Number A1601. March.Google Scholar
  23. Davis, L. W. (2010). Evaluating the Slow Adoption of Energy Efficient Investments: Are Renters Less Likely to Have Energy Efficient Appliances?. NBER Working Paper, (w16114). http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1626603.
  24. de T’Serclaes, P., & Jollands, N. (2007). Mind the gap: Quantifying principal-agent problems in energy efficiency. Paris: International Energy Agency and the Organization for Economic Cooperation and Development http://www.iea.org/publications/freepublications/publication/mind_the_gap.pdf.Google Scholar
  25. DEFRA/BERR. (2008). The UK Fuel Poverty Strategy: 6th Annual progress Report. http://www.bristol.ac.uk/poverty/downloads/keyofficialdocuments/Fuel%20poverty%20strategy%202008.pdf.
  26. Drehobi, A., & Ross, L. (2016). Lifting the high energy burden in America’s largest cities. ACEEE Research Report. 20 April. http://aceee.org/research-report/u1602.
  27. Earth Advantage Institute. (2009). Energy Performance Score 2008 Pilot: Findings and Recommendations Report. August. http://www.earthadvantage.org/assets/documents/EPSPilotReport_2008.pdf.
  28. Economist. (2009). Home ownership: Shelter, or burden? 16 April.Google Scholar
  29. Eisenberg, J. F. (2010). Weatherization Assistance Program-Background Data and Statistics (No. ORNL/TM-2010/66). Oak Ridge National Laboratory (ORNL). https://info.ornl.gov/sites/publications/files/Pub23632.pdf. Accessed 31 October 2016.
  30. Ettenson, L. (2011). “Energy efficiency: California's leading energy resource.” Electric Light and Power. November 9.Google Scholar
  31. Fabi, V., Andersen, R. V., Corgnati, S., & Olesen, B. W. (2012). Occupants' window opening behaviour: A literature review of factors influencing occupant behaviour and models. Building and Environment, 58, 188–198.CrossRefGoogle Scholar
  32. Fisher, S. & Colton. (2017). Current Year Affordability Gap. 2016 Data. Released April 2017. http://www.homeenergyaffordabilitygap.com/03a_affordabilityData.html.
  33. Gillingham, K., Newell, R. G., & Palmer, K. (2009). Energy efficiency economics and policy (NBER working paper 15031). National Bureau of Economic Research. http://www.nber.org/papers/w15031.
  34. Gillingham, K., Harding, M., & Rapson, D. (2012). Split incentives in residential energy consumption. Energy Journal, 33(2).Google Scholar
  35. Greene, D. L. (2011). Uncertainty, loss aversion, and markets for energy efficiency. Energy Economics, 33(4), 608–616.CrossRefGoogle Scholar
  36. Gujarati, D., & Porter, D. (2009). Basic econometrics (5th ed.). Boston: McGraw-Hill.Google Scholar
  37. Hales, R.L. (2014). Two residential PACE programs dominate California headlines. The EcoReport. 25 November. http://www.theecoreport.com/green-blogs/area/usa/california/two-residential-pace-programs-dominate-california-headlines/. Accessed 26 November 2014.
  38. Harris, C., & Laibson, D. (2001). Dynamic choices of hyperbolic consumers. Econometrica, 69(4), 935–957.MathSciNetCrossRefGoogle Scholar
  39. Hassan, K. (2014). Energy Savings Assistance Program (ESA) and California Alternate Rate for Energy (CARE) Program Annual Report 2013. Results May 1. http://www.cpuc.ca.gov/iqap/. Accessed 31 October 2016.
  40. Heffner, G., & Campbell, N. (2011). Evaluating the co-benefits of low-income energy-efficiency programmes. June. In Workshop Report, OECD/IEA, Paris. http://www.iea.org/publications/freepublications/publication/low_income_energy_efficiency.pdf. Accessed 31 October 2016.
  41. Hirst, E., & Brown, M. (1990). Closing the efficiency gap: Barriers to the efficient use of energy. Resources, Conservation and Recycling, 3(4), 267–281.CrossRefGoogle Scholar
  42. HM Government. (2014). Cutting the cost of keeping warm – a consultation to prepare for a new fuel poverty strategy for England. July. https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/335099/fuel_poverty_consultation.pdf Accessed 14 November 2014.
  43. Howden-Chapman, P. (2007). Effect of insulating existing houses on health inequality: Cluster randomised study in the community. British Medical Journal, 334(7591), 460.CrossRefGoogle Scholar
  44. Jaffe, A. B., & Stavins, R. N. (1994). The energy-efficiency gap what does it mean? Energy Policy, 22(10), 804–810.CrossRefGoogle Scholar
  45. KEMA, Inc. (2010). 2009 California Residential Appliance Saturation Study. California energy commission. Publication number: CEC-200-2010-004-ES. http://websafe.kemainc.com/rass2009/.
  46. Kissock, J. K., Haberl, J. S., & Claridge, D. E. (2003). Inverse modeling toolkit: Numerical algorithms (RP-1050). Transactions-American Society Of Heating Refrigerating And Air Conditioning Engineers, 109(2), 425–434.Google Scholar
  47. Levinson, A. (2013). “California energy efficiency: Lessons for the rest of the world, or not?”, NBER Working Paper No. 19123. http://www.nber.org/papers/w19123
  48. Levinson, A., & Niemann, S. (2004). Energy use by apartment tenants when landlords pay for utilities. Resource and Energy Economics, 26(1), 51–75.CrossRefGoogle Scholar
  49. Liddell, C., Morris, C., McKenzie, S. J. P., & Rae, G. (2012). Measuring and monitoring fuel poverty in the UK: National and regional perspectives. Energy Policy, 49, 27–32.CrossRefGoogle Scholar
  50. Mazur-Stommen, S., & Farley, K. (2013). ACEEE field guide to utility-run behavior programs. December 19. http://aceee.org/research-report/b132.
  51. Mills, B., & Schleich, J. (2012). Residential energy-efficient technology adoption, energy conservation, knowledge, and attitudes: An analysis of European countries. Energy Policy, 49, 616–628.CrossRefGoogle Scholar
  52. Moore, R. (2012). Definitions of fuel poverty: Implications for policy. Energy Policy, 49, 19–26.CrossRefGoogle Scholar
  53. Murray, A. G., & Mills, B. F. (2014). The impact of low-income home energy assistance program participation on household energy insecurity. Contemporary Economic Policy, 32, 811–825.CrossRefGoogle Scholar
  54. Murtishaw, S., & Sayanthe, J. (2006). Quantifying the Effect of the Principal-Agent Problem on US Residential Energy Use. LBNL-59773. http://escholarship.org/uc/item/6f14t11t.
  55. Nadkarni, N., Michaels, H. (2012). A New Model for Disclosing the Energy Performance of Residential Buildings. http://web.mit.edu/energy-efficiency/docs/EESP_Nadkarni_BuildingPerformanceDisclosure.pdf.
  56. National Renewable Energy Laboratory (NREL). (2014). National Residential Efficiency Measures Database. v3.0.0. http://www.nrel.gov/ap/retrofits/group_listing.cfm. Accessed 14 November 2014.
  57. Nelson, H. (2014). Personal communication with a staff member of the DC sustainable energy utility. 27 June.Google Scholar
  58. Nicol, J. F., & Humphreys, M. A. (2002). Adaptive thermal comfort and sustainable thermal standards for buildings. Energy and Buildings, 34(6), 563–572.CrossRefGoogle Scholar
  59. Phillips, Y. (2012). Landlords versus tenants: Information asymmetry and mismatched preferences for home energy efficiency. Energy Policy, 45, 112–121.CrossRefGoogle Scholar
  60. Polly, B., Kruis, N., & Roberts, D. R. (2011). Assessing and improving the accuracy of energy analysis for residential buildings. US Department of Energy, Energy Efficiency and Renewable Energy. https://www.nrel.gov/docs/fy11osti/50865.pdf.
  61. Sagewell. (2014). How energy efficient is your home? http://www.sagewell.com/ Accessed 12 November 2014.
  62. Simcock, N., & Walker, G. (2015). Fuel Poverty Policy and Non-Heating Energy Uses DEMAND Centre working paper 16. April. http://www.demand.ac.uk/wp-content/uploads/2016/01/working-paper-16-Walker.pdf.
  63. Southern California Gas. (2014). Rebates for your home. http://www.socalgas.com/for-your-home/rebates/ Accessed 12 November 2014.
  64. Southwell, B. G., & Murphy, J. (2014). Weatherization behavior and social context: The influences of factual knowledge and social interaction. Energy Research & Social Science, 2, 59–65.CrossRefGoogle Scholar
  65. Thomson, H., & Snell, C. (2013). Quantifying the prevalence of fuel poverty across the European Union. Energy Policy, 52, 563–572.CrossRefGoogle Scholar
  66. Tversky, A., & Kahneman, D. (1992). Advances in prospect theory: Cumulative representation of uncertainty. Journal of Risk and Uncertainty, 5(4), 297–323.CrossRefGoogle Scholar
  67. U.S. Department of Energy (DOE). (2012). Tips: Insulation. May 2nd. Access 16 April 2014. http://energy.gov/energysaver/articles/tips-insulation.
  68. U.S. Department of Health and Human Services (DHHS). (2014). Low-income home energy assistance program: Report to congress for fiscal year 2009. June 6. https://www.acf.hhs.gov/sites/default/files/ocs/rpt_fy09_liheap_rtc_final_052114_2.pdf. Accessed 1 Nov 2014.
  69. U.S. Department of Health and Human Services (HHS). (2009). LIHEAP home energy notebook for fiscal year 2009. http://www.acf.hhs.gov/sites/default/files/ocs/fy2009_liheap_notebook.pdf Accessed 16 Nov 2014.
  70. U.S. Energy information agency (EIA). (2009). Household Energy Use in California. http://www.eia.gov/consumption/residential/reports/2009/state_briefs/pdf/ca.pdf Accessed 12 November 2014.
  71. U.S. Energy information agency (EIA). (2010). Weatherization and Intergovernmental Programs. Energy Efficiency and Renewable Energy. http://www1.eere.energy.gov/library/pdfs/48098_weatherization_assisprog_fsr4.pdf.
  72. U.S. Energy Information Agency (EIA). (2013). Table HC2.2 structural and geographic characteristics of U.S. homes, by owner/renter status, 2009. Final Release, April. http://www.eia.gov/consumption/residential/data/2009/#undefined. Accessed 26 November 2014.
  73. US Census. (2013). Poverty. Updated September 17. http://www.census.gov/hhes/www/poverty/about/overview/index.html.
  74. US Census. (2016). Mean income in the past 12 months (in 2010 inflation-adjusted dollars. Variable S1902 for 2010. https://factfinder.census.gov/faces/nav/jsf/pages/guided_search.xhtml.
  75. US EIA. (2016) California: State Profile and Energy Estimates. California Price Differences from U.S. Average, Most Recent. Last updated October 20, 2016. Retrieved 31 May 2017. https://www.eia.gov/state/?sid=CA#tabs-5.
  76. Vringer, K., & Blok, T. A. K. (2007). Households energy requirements and value patterns. Energy Policy, 35, 553–566.CrossRefGoogle Scholar
  77. Wilson, J., Dixon, S. L., Jacobs, D. E., Breysse, J., Akoto, J., Tohn, E., Isaacson, M., Evans, A., & Hernandez, Y. (2013). Watts-to-wellbeing: Does residential energy conservation improve health? Energy Efficiency, 7, 151–160.  https://doi.org/10.1007/s12053-013-9216-8.CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media B.V., part of Springer Nature 2018

Authors and Affiliations

  1. 1.Portland State UniversityPortlandUSA
  2. 2.Pomona CollegeClaremontUSA

Personalised recommendations