Improving the Expansion of Electricity Services Considering Affordability Issues: A Case Study in Brazilian Low-Income Households

Abstract

The assurance of universal access to electricity refers to not only physical access to electricity, heating, and/or cooling services but also making them affordable to families. This is the case in Brazil, where physical electricity access reaches more than 99% of the urban population, but a high proportion of households are disconnected due to lack of payment. Affordability is a key factor in planning the expansion and maintenance of the electricity grid. In this paper, therefore, we propose the development of a strategy, called the Payment Capability Indicator (PCI), that provides new information about the relationship between energy cost and family income. The classic indicator, the Ten Percent Rule (TPR), was the base, and empirical research was conducted in two low-income neighborhoods in Paraná state, southern Brazil. The results consider variables that add new information to guide local public policies, combining interventions to improve affordability, energy efficiency, alternatives to energy services and consumer behavioral changes.

1. Introduction

Goal 7 of the Sustainable Development Goals (SDGs) [1] is about affordable and clean energy. This goal claims to ensure access to affordable, reliable, sustainable, and modern energy for all. It consists of three subgoals: ensure universal access to electricity and to clean and modern cooking fuels/technologies, increase the share of renewable energy in the global energy mix and improve the rate of energy efficiency.

According to The Energy Progress Report [2,3], one factor has been identified as a determinant of achieving the goal of universal access to electricity: affordability. According to the report [3], the consumption cost of 30 kWh/month (kilowatt–hours per month) represents more than 5% of the income of the 40% poorest countries with an access deficit. This corresponds to 285 million people who are unable to pay for a minimum amount of electricity at the global level.

In Brazil, universal electricity access is a reality. According to data from the last census in 2010, the country has a rate of 98.7% of the population with electricity access in urban and rural areas [4]. The National Household Sample Survey (PNAD) [5] showed that 99.8% of households already have electricity. Of these households, 99.5% have electricity from the distribution network, and 99.2% have full-time availability.

However, the affordability problem gained prominence after successive economic recessions and the COVID-19 pandemic. There was an increase in electricity tariffs, unemployment, budget cuts to social programs, growth in the default rate, and reduction in income, among others [6,7], and Brazilians’ ability to pay for essential services was directly affected [8,9,10,11]. For example, electricity nonpayment in May 2020 was 24 times higher than in the same month in 2019 especially because of the COVID-19 pandemic [8]. In Brazil, there is a phenomenon called income management. The supply is turned off after three months. Therefore, it is common for low-income families to not pay for electricity in order to buy food, for example and to only pay the electricity bill in the following month or in the second month.

In this context, knowing/measuring and monitoring electricity affordability in Brazil is essential to addressing energy poverty [12], providing conditions to realize essential capabilities [13], and developing public policies geared toward the Sustainable Development Goals. In [7], we developed an affordability indicator for each utility concession area in Brazil. However, we realized that regions in Brazil with high levels of socioeconomic complexity tended to be more exposed to problems of nonpayment, but this was not absolute and represented a limitation in that study because we did not have accurate household-level data. Therefore, we decided to analyze data on a smaller scale—at the household level—as suggested in [2,3].

Although there are several energy poverty indicators in the literature [14,15,16], we have found limitations in their application in Brazil. First, we are a country of continental dimensions where regions have different socioeconomic and climatic realities. In addition, southern developing countries must deal with clandestine connections and insecurity in socioeconomically complex regions. Also, the problem of inequality in income distribution distorts indicators based purely on cost and income. Therefore, we developed a strategy called PCI (Payment Capability Indicator) that set a methodology to be applied in different realities, identifying new variables that have an impact on the relationship between energy cost and family income. For this, considering the lack of official data correlating socioeconomic information to electricity in Brazil, empirical research was carried out. A case study was conducted in two low-income communities in the north of the state of Paraná, southern Brazil, to collect cross-section data for the development of the PCI through an econometric model. The chosen samples came from the same metropolitan region and therefore had the same climatic characteristics. In addition, the families were beneficiaries of government housing programs and thus had similar socioeconomic characteristics.

This article is organized as follows: Initially, we present a bibliographical review that portrays electricity affordability in Brazil, providing the reader with the necessary information for comparability. We address the existing subsidies and incentives as well as the challenges and technological alternatives for measuring and trading electricity. Then, we present a literature review of energy poverty and the methodology used to define the PCI strategy, which includes the process of collecting data in the field and the essential definitions. Finally, we discuss different ways to use and interpret the PCI.

2. Brazil: Subsidies, Incentives, and Challenges

Connection to electricity service, either through a distribution network or through distributed generation, increases families’ capacities to improve their livelihood, health, and education [13]. In this regard, one way for governments to act is to guarantee free access to electricity for the population [17,18].

In Brazil, policies on access to electricity have been successful. The first measures were introduced in the 1960s, with the creation of cooperatives for rural electrification [19], followed by other successful public policies in subsequent decades (especially the 1990s, with the National Program for Energy Development of States and Municipalities (PRODEEM) and the Light in the Countryside Program) [19,20,21]. However, the major breakthrough came with the creation of the Light for All (LPT—Luz para Todos) Program in 2003 [22,23]. The LPT’s current challenge is to stop electricity exclusion in Brazil, primarily benefiting populations in extreme poverty such as the quilombola and indigenous communities, settlements, and riverside dwellers, among others [22,23,24,25,26].

The purpose of the 7th SDG for the universalization of electricity refers to guaranteeing not only physical access but also the continuous use of electricity, that is, ensuring conditions for maintaining access to electricity service through tariffs affordable and adequate to the population’s ability to pay [1,27]. This energy vulnerability factor [28] has become crucial in Brazil in view of the country’s current economic recovery process [29,30,31]. Thomson et al. [32] concluded that there is an urgent need for research and policy activities in Latin America to support the purpose of the 7th SDG. Bezerra et al. [33] calculated a Multidimensional Energy Poverty Index (MEPI) for Brazil considering three different energy dimensions: physical access, appliances ownership, and affordability. The results indicated that affordability is the main issue characterizing energy poverty in the country.

Aiming to improve the relationship between expenditure on electricity and the population’s income, the Brazilian government provides subsidies and incentives. The main measure in order to reduce expenditures was the implementation of the Low-Income Tariff or Social Tariff (ST), which grants discounts for the low-income population according to their consumption [19,23,27,34]. It was regulated in 2011 and is based on cross-subsidy; that is, the benefit is financed by the population that pays the conventional tariff. According to Simões and Leder [35], most low-income consumers enjoy the minimum advantage from this subsidies policy, and Maciel et al. [36] proposed a renewal of ST based on macro-data.

At the subnational level, there are also initiatives to subsidize electricity to low-income households. One example is the Fraternal Light Program (FLP), established by the government of the state of Paraná (state program), where the empirical research was conducted. The FLP was established on 31 July 2013 and exempts from the payment of the electricity bill families benefiting from the Social Tariff and with monthly consumption lower than or equal to 120 kWh. In 2021, the name of the program was changed to Solidarity Energy, and the monthly consumption limit was increased to 150 kWh. The program serves the 399 municipalities of the state, and the energy distributors automatically grant the concession to families registered in the Social Tariff and who meet the consumption criteria [37,38].

In addition, the country has been supporting actions to promote the efficient use of electricity, which allows consumers to reduce their expenditures and, consequently, improves the payment capability of families [19]. Since the implementation of Law 991 in 2000, at least 0.5% of the concessionaires’ net operating income must be earmarked for energy-efficiency programs. Currently, part of these resources can be used in consuming units belonging to low-income communities, rural or benefiting from the Social Tariff [39]. These programs introduced technologies for measuring electricity; the replacement of electronic equipment by more efficient ones; solar water heating [40,41]; and educational campaigns on the efficient use of electricity [41].

Also, in order supplement families’ income and improve their ability to pay, the Bolsa Familia Program (BFP) was created in 2003 [42]. The purpose of this program is to address poverty and inequality in Brazil through three main lines: promoting immediate poverty relief through direct cash transfers to families; strengthening the exercise of basic social rights in the areas of healthcare, education, and social assistance through the fulfillment of conditionalities; and promoting opportunities for the development of families through actions that promote the overcoming of vulnerability and poverty by BFP beneficiaries [43,44].

The main challenge for electricity distributors is ensuring continuous payments and avoiding electricity theft. In Brazil, if someone fails to pay the bill, access is suspended within 90 days from the due date [45]. To restore the service, the consumer needs to pay off overdue debts and to pay a reconnection fee. Negotiating this debt may not be the first option for the low-income population, so many of the consumers end up resorting to the illegal connection [46]. Reversing this situation is even more complicated, especially if electricity theft occurs in socioeconomically complex areas [47].

Debts arising from defaults in the low-income household sector are considered as irrecoverable revenues and, together with losses due electricity theft, make up the so-called non-technical losses of concessionaires, which have a direct impact on the tariff review [48,49]. That is, to compensate for the losses of the distributor, there may be an increase in tariffs, which further makes it difficult for the consumers to pay the bill on time.

Also, the theft problem may not be related to the families’ payment capability but rather to a problem of collective action. That is, if the theft of electricity is a frequent practice in the area, people do not care about doing it [50]. The problem of collective action has two dimensions: the individual stops fighting corruption and starts to expect benefits; and the individual who would like to leave corruption does not do so due to lack of confidence, remaining without incentives to act honestly. It is a behavior determined by the individual’s impression, and it derives from the sociopolitical structure and institutional governance. Thus, the problem is common in developing countries [50,51,52,53,54], and it becomes a challenging task to find effective measures [50]. Therefore, considering the impact of default and theft of electricity on the tariff, it is essential to address the problem to improve electricity affordability.

Improving energy affordability is not an easy task, and we propose in this paper a strategy to understand the problem at the household level. The methodology can be applied in different places, directing energy planning and public policies, considering new variables that could better explain the relationship between energy cost and income.

3. Energy Poverty Measurement

The Brazilian National Commission for the Sustainable Development Goals does not have an affordability indicator [55]. In a country with so many inequalities like Brazil, this indicator becomes even more subjective [56]. The authors of Tracking SDG 7—The Energy Progress Report proposed that this variable is ideally measured at the level of residence [2], and this is the purpose of the article.

Given the challenge of promoting electricity affordability in Brazil, it is essential to have a measurable reference to determine the current conditions; the advances obtained by subsidies and incentives; and the changes arising from the use of new technologies, in addition to enabling the planning of new actions and public policies. For this, [57] proposes a sequence of steps to define an energy poverty measure. It starts with the definition of the concept, i.e., what is energy poverty? Thus, the most appropriate type of approach can be thought of, as well as the procedure to be followed, verifying the availability of data or defining a sample. By knowing the approach, the available data and the indicators existing in the literature; and in partnership with experts in the field, researchers and leaders, suitable indicators can be defined and tested.

From this perspective, the first step is to define the concept of energy poverty. However, there is no consensus due to the different realities (climate, socioeconomic, cultural, etc.) of the regions and countries. Most authors describe energy poverty as a level of energy insufficient to meet basic needs [16,28,58]. The United Nations Development Program (UNDP) [59] defines this concept as the impossibility of choice for energy services in conditions that promote support for economic and social development of families and individuals. In turn, in The World Energy Outlook (WEO) [60,61], energy poverty is evaluated as the lack of access to electricity and the dependence on the traditional use of biomass for cooking.

Other authors address this concept, such as [16], who state that energy poverty occurs when the average monthly expenditure of the house on energy account for a substantial portion of the average family’s monthly income. In turn, [62] consider that the main promoters are the socioeconomic situation of the household, the energy performance of the residence and the price of energy. One of the broadest and most accepted definitions is the one of [63], which defines energy poverty as the absence of sufficient choices in accessing adequate, affordable, reliable, high-quality, safe and environmentally benign energy services to support economic and human development.

Lastly, Ref. [64] define energy poverty based on the capability approach [13], which has currently been applied in studies of the EU Energy Poverty Observatory [65]: “an inability to realize essential capabilities as a direct or indirect result of insufficient access to affordable, reliable and safe energy services, and taking into account available reasonable alternative means of realizing these capabilities”. As the definition of [64] depends on the determination of essential capabilities, it has been shown to be very adaptable to the realities of different countries.

Next, several types of approach must be evaluated to choose the most suitable one for the proposed analysis: expenditure-based; agreement-based or self-reported; and results-based approaches. The most common type of approach is the expenditure-based one, whose indicators can capture the main elements and the severity of energy poverty by using different thresholds [57], such as Ten Percent Rule (TPR) [62,66,67], Two Times Median Expenditure Share (2M) [15,68], etc. Other indicators are based on agreement or self-reported and consider the basic needs of several types of households [69], for example, the High-Cost Low Income (HCLI) and the Minimum Income Standard (MIS-based) [67,68,69]. Finally, there are the results-based indicators, which are the least used to measure energy poverty. In this case, the indicators are built from output data from companies of energy supply or public health. Implementing them is complex because it is difficult to measure the results related to health and social issues. In addition, they present uncertain responses since they focus exclusively on the outputs and do not consider the causes of energy poverty [57].

In turn, the implementation of the approach depends on the availability of official data and their level of stratification. In case of unavailability of data, the procedure to be adopted is the empirical research. In this case, a statistical planning is necessary for the survey to be valid.

Finally, once the concept of energy poverty, the approach and the availability of data are defined, then the reference indicator can be chosen among those existing in the literature, enabling the validation of the indicator to be proposed.

4. Methodology

The methodology presented in this article begins with a field study due to insufficient granular data regarding energy expenditure, appliances, and income at the level of residence in the country. For the definition of the reference indicator, used to elaborate the PCI, we followed the steps indicated by [57]. Thus, the concept of energy poverty was first defined for the reality of the context, making it possible to establish the approach and the indicator to be tested. Next, the variables proposed by [16], as factors influencing electricity affordability, were used as independent variables for building the econometric model that defined the PCI. Finally, the developed indicator presented new variables that should be observed to improve electricity services. Figure 1 presents the structure of the developed methodology. The equation used to quantify the PCI is presented in Section 5.

4.1. Empirical Research

Some social groups were left out of the urbanization process due to not having proper access to basic human needs. Therefore, two low-income communities were selected for data collection. The communities are in the northern part of Paraná state, where the climate is subtropical, with hot summers and cold winters, and well-distributed rainfall throughout the year. The state’s monthly per capita income is the sixth highest in Brazil, and the research region has the tenth highest income in the state (2.7 minimum wages).

The first community (Community A) comprises 1272 houses, and the second one (Community B) comprises 185 houses. A simplified questionnaire was developed and applied to the families between November 2017 and March 2018. The sample sizes are 202 elements in Community A (error in the range of 5 to 6%, degree of confidence 90%) and 115 elements in Community B (error in the range of 4 to 5%, degree of confidence 90%). The response rate for Community A is lower in comparison to B because the research team in Community A was smaller and a random sample was selected, while in Community B, 100% of the houses were visited by a larger team.

Table 1. Summary of relevant data obtained during the field research for the two samples.

	Community A	Community B
Total households	1272	185
Sample	202	115
Not legal connection	45%	55%
Did not report income	2%	0
Unreadable electricity bill	2%	0
Consumer tracks his electricity consumption	21%	13%
Effective sample	103	52
Measurement electricity technology	Cyclometric display measurement by reader	Digital display centralized measurement
Families below the poverty line	41%	24%
Unemployment	34%	29%
Common electronics items	Lighting, refrigerator, electric shower, TV, and washing machine.	Lighting, refrigerator, electric shower, and TV.

summarizes the general data obtained with the questionnaires applied door to door. The principal factor observed in both samples was the extent of electricity theft. Since those families do not have the bill indicating consumption, expenses in currency, etc., they were excluded from the mathematical test along with those who did not report income and showed unreadable electricity bill.

4.2. Data Processing

This step considers all the preliminary definitions pointed out in [57]. The first of them refers to the concept of energy poverty. For the development of this work, we chose the definition of [64]. This description is very flexible, which adapts to different contexts based on the establishment of what would be the essential capacities of that population, and allows actions to address energy poverty.

The relationship between electricity, services and results can be built based on the information obtained from the two samples. Both communities have electricity distribution services operated by the same concessionaire. According to the criteria established by the Brazilian Electricity Regulatory Agency (ANEEL), the low-income communities that comprise the study samples receive a reliable and safe service provision. Regarding the secondary capacities dependent on electricity, which are considered essential for families to guarantee the basic capacities of health, education, respect and social relationships, the penetration of household appliances was analyzed. Thus, we observed that illumination, preserving refrigerated food, heating bath water, accessing information/entertainment, and charging electronic devices are essential capabilities for more than 90% of the total sample.

In this sense, a family residing in a low-income community with the same constructive, cultural, and climatic characteristics of the sample under study can be considered energy poor if they are unable to have electric illumination, preserve refrigerated food, heat water for bathing, access information/entertainment through TV and charge electronic devices (cell phones).

Thus, once the concept of energy poverty is defined, the second step is to identify the best approach to measure it. The expenditure-based approach is the most appropriate one because it considers the household income and the energy expenditures, which are essential for an affordability analysis [57].

The last step, which is defining the indicator, started from the study of [14]. The authors evaluated the response of the most popular indicators regarding the dynamic properties of the payment capability. The best answers were found for TPR and MIS. The authors suggest the use of the TPR if the sample is restricted to the low-income class due to its simplicity and ease of use. Since it is specifically the sample in this work, the TPR indicator was used as a reference.

4.3. Mathematical Test

In this step of the methodology, the energy poverty indicator Ten Percent Rule (TPR) [66] was calculated for each of the sample elements according to Equation (1),

TPR = EC/IN,

(1)

where EC is the electricity cost and IN is the family income.

The multiple linear regression model considers the TPR as the dependent variable. The determination of the independent variables was performed based on the article of [16]. The authors produced a critical analysis of energy poverty indicators in Europe, and proposed new topics that should be addressed. For the payment capability factor, the following items were suggested: detailed information on household income, including social benefits; total energy cost for all needs and services; issues about how consumers perceive the payment capability/weight of energy services in the residence (including heating, cooling and services not related to temperature, such as illumination); nonpayment and delays in energy bills, and if suspension in the supply occurred; information about payment method (prepaid, cash, direct debit) and tariffs.

Considering those information [16], the following variables were proposed for the model,

Table 2. Model independent variables.

Item	Variable	Unit	Statistical Treatment
Family income including social benefits	inc	R$	endogenous variable—excluded
Total electricity costs	cost	R$	endogenous variable—excluded
Impacts of electricity costs on total household costs	EHcost	R$/R$	endogenous variable—excluded
Electricity nonpayment	Npay	Yes or Not	dummy
Power cut	cut	Yes or Not	dummy
Tariff	tax	R$	endogenous variable—excluded
Monitoring of consumption	tracks	Yes or Not	dummy
Number of home electronic items	items	Quantity	-
Number of people who live at a residence	resid	Quantity	-
At least one employed family member	emp	Yes or Not	dummy
Social benefit	social	Yes or Not	dummy
Total household costs	Hcost	R$	endogenous variable- excluded
Electricity consumption	Econs	kWh	endogenous variable- excluded

.

Thus, the regression model obtained, by means of Ordinary Least Squares (OLS), is represented by Equation (2) [1].

$$y={\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+\dots +{\beta }_k{x}_k+u,$$

(2)

where $y$ defines the dependent variable under analysis, in this case, TPR; ${\beta }_0$ is the model constant, also called the intercept; ${\beta }_1\dots {\beta }_k$ is a vector of slope parameters of the curve; $u$ is the error term that, by definition, has zero mean, normal distribution and constant variance; $x_1\dots x_k$ is an array of independent variables of the model, which represent Npay, cut, tracks, items, resid, emp and social.

After defining the classic regression model, specification tests were performed to assess whether the linear estimator is unbiased, consistent, and asymptotically normal. Considering that the data are cross-section series, it was necessary to evaluate the presence of heteroskedasticity, multicollinearity and normal distribution of the residues [70]. The data were available through Supplementary Materials.

To detect the presence of heteroskedasticity, the model was first estimated, according to Equation (2), and the homoscedasticity tests were performed on the residuals. The heteroskedasticity test has as the null hypothesis the presence of homoscedasticity; that is, in this case, one does not want to reject the null hypothesis. The White and Breusch–Pagan tests were then applied. Multicollinearity tests were also performed, using the Variance Inflated Factor (VIF) test, and the normal distribution of residuals was evaluated by the Shapiro–Wilk W test and the Skewness/Kurtosis test [70,71].

The VIF test confirmed the absence of multicollinearity. However, the White and Breusch–Pagan tests demonstrated the presence of heteroskedasticity in the model. This fact does not corrupt the hypothesis of unbiased and consistent estimators, but they stop being efficient: that is, they no longer have the minimum variance (there is another estimator with lower variance and, therefore, more efficient). When this happens, it is necessary to make a correction. We used White’s correction method and bootstrap with 100 replications and random seed 54,321. Such estimators are technically conceptualized as consistent estimators of the covariance matrix for heteroskedasticity. However, these corrections do not solve the problem of non-normal distribution of residuals. This is the limitation of the model due to the small sample.

5. Results

Considering that the effective sample in Community A has 103 elements and 52 elements in Community B, excluding illegal connections, families who did not report income and those with unreadable electricity bills, we chose to join the two samples to define an econometric model PCI. The complete model, presented in

Table 3. PCI regression model with White´s correction—Community A + B.

Variable	Coefficient β	Robust Standard Error	p-Valor	Statistical Significance	95% Conf. Interval
Npay	0.071	0.032	0.030	**	0.007	0.135
cut	−0.026	0.040	0.518	-	−0.105	0.053
tracks	0.014	0.028	0.614	-	−0.041	0.070
items	−0.003	0.006	0.596	-	−0.016	0.009
resid	0.019	0.008	0.017	**	0.003	0.034
emp	−0.060	0.026	0.020	**	−0.111	−0.009
social	0.048	0.023	0.039	**	0.002	0.094
constant	0.050	0.056	0.368	-	−0.060	0.160
Number of observations = 155
Probability > F = 0.041
R2 = 0.137

At the confidence level set at 95%, ** corresponds to a statistical significance of 5%; - indicates that the variable has no statistical significance.

, lists impact factors for the payment capability: default, suspension in the supply, monitoring of consumption, number of inhabitants of the residence, number of home electronic items, employed family member and social benefit, which meets the variables suggested by [16]. The result presented in Table 3 and

Table 4. PCI regression model with bootstrap correction—Community A + B.

Variable	Coefficient β	Bootstrap Standard Error	p-Valor	Statistical Significance	95% Conf. Interval
Npay	0.071	0.032	0.029	**	0.007	0.135
cut	−0.026	0.037	0.489	-	−0.099	0.047
tracks	0.014	0.030	0.637	-	−0.045	0.073
items	−0.003	0.006	0.565	-	−0.015	0.008
resid	0.019	0.007	0.012	**	0.004	0.034
emp	−0.060	0.026	0.022	**	−0.111	−0.009
social	0.048	0.020	0.018	**	0.008	0.088
constant	0.050	0.049	0.311	-	−0.047	0.147
Number of Observations = 155
Replications = 100
Probability > chi2 = 0.007
R2 = 0.137

At the confidence level set at 95%, ** corresponds to a statistical significance of 5%; - indicates that the variable has no statistical significance.

(White and bootstrap correction, respectively) had an adjustment around of 14%.

Based on the results obtained through regression analysis, the PCI can be represented through Equation (3),

$$\begin{gathered} PCI=0.050+0.071\ast Npay-0.026\ast cut+0.014\ast tracks-0.003\ast items+0.019\ast resid-0.060 \\ \ast emp+0.048\ast social. \end{gathered}$$

(3)

The first variable to be observed is the intercept, which suggests that the ratio between the invoice and the income, when the explicative variables are zero, is 5% on average. However, this is not statistically significant at 95%; then, its statistical value is zero.

The presence of default indicates an increase in the PCI of 7.1%. It is well known that this is a widespread practice among low-income families. Since the suspension of the electricity supply takes place within 90 days of the payment delay, these consumers fail to pay for their consumption to honor other household expenses, and they only pay before the suspension occurs. The difficulties encountered with the suspension of electricity and the fees involved in the reconnection and regularization of the service are the main points because families seek to improve their conditions of payment capability and avoid the problem.

Both the income and the expenditure of a residence can be directly affected by the number of people who live in it. Therefore, it was expected that the number of residents would be a statistically significant variable in the model. The result indicates that the greater the number of people living in the residence, the higher the ICP (1.9% for each extra person). That is, for the low-income families that make up the sample, an increase in the number of residents impacts more on expenditure than on income growth, which is probably because they are of an age that is inadequate for the labor market or unemployed.

On the other hand, the presence of at least one person employed reduces the ICP by 6%. This result is expected because an active income in the home increases the family’s purchasing power. Furthermore, the family will spend less time at home, tending to reduce energy consumption.

Finally, the presence of social benefit is statistically significant at 90% of reliability, and the results demonstrate that a person who receives a social benefit is 4.8% more energy poor. This result was also expected because these families undergo an analysis of several criteria to be eligible for social benefits. Typically, they are people on the poverty line.

In order to define the most appropriate model, we used the definition of energy poverty which is reference for this work: “A family residing in a place with the same constructive, cultural and climatic characteristics of the sample under study may be considered energy poor if they do not have the following capabilities due to a financially inaccessible electricity service: having electric illumination; keeping food refrigerated; heating water for bathing; accessing information/entertainment through TV; and charging electronic devices (cell phones)”.

The realization of the secondary capabilities, listed in the definition, depends on the possession of certain electronic items. So, those families that do not have a refrigerator, an electric shower, and a television (3 items) must be considered energy poor by definition.

Finally, after defining Equation (3), which models the affordability conditions of low-income families in Paraná State, we can affirm that nonpayment and numbers of residents are factors that increase energy poverty, besides families who receive social benefits are poorer and consequently energy poor, too. On the other side, the number of employed residents impact positively, reducing electricity poverty.

6. Conclusions

Electricity affordability is ideally measured at the household level, and defining an indicator is not an easy task due to data unavailability and different socioeconomic and climatic realities. Moreover, different realities affect the essential capabilities. For example, hot bathing is essential in southern Brazil, but it is not in the north and northeast areas. Bearing this in mind, we developed a strategy, called Payment Capability Indicator (PCI), which provides new information about the relation between energy cost and family income.

The idea is to conduct empirical research in the region to be studied. The data obtained are the reference for defining the concept of energy poverty based in [64]; for the appropriate approach to measure energy poverty; the way to implement this approach; and the choice of an indicator existing in the literature for testing. A linear regression model can be elaborated considering the testing indicator as a dependent variable and the criteria raised by [16] as explanatory variables. Thus, the PCI is defined according to the reality of the study population.

The indicator considers other variables that add new information. In the case study, the number of inhabitants and the nonpayment were also used to compose the PCI. Both variables aggravate the population’s payment capability. Considering that the number of residents impacts the electricity consumption, and Brazilian tariff benefits are granted based on consumption, many low-income families are excluded from this policy. Also, measures on energy efficiency and guidance/communication are essential. Guiding the monitoring of consumption is minimally necessary, at least in places with digital meters.

The definition of energy poverty used as reference in this study proposes other levels of intervention in order to meet particular needs and promote alternative energy services of non-domestic supply; to support alternatives to energy services; and to change expectations, customs and practices. In this sense, we suggest expanding the offer of cultural and sporting activities in the evening for the case study. Thus, we propose a change in habits by making families leave their houses, which would lead to a reduction in electricity consumption related to televisions and illumination. Regarding energy services of non-domestic supply, the creation of community laundries is recommended, in which a small fee for the maintenance of the structure would be charged. Also, placing solar-powered outlets in public places can reduce household electricity consumption due to charging electronic devices.

In addition to guiding actions to address energy poverty, the PCI also makes it possible to monitor the evolution that such changes may cause. From the socioeconomic and electricity data at the household level, one can build a consumer profile for each concessionaire. Due to the differences existing in a country with continental dimensions like Brazil, the elaboration of a profile by concession area makes the planning of measures more assertive. Furthermore, this allows the comparability of results in a historical series, formalizing a metric.

The main limitation of the study was the sample size. The lack of official data relating socioeconomic variables and electricity, at the household level, prevents the expansion of the study and the validation of the model for a Brazilian state or region. To implement the strategy as a state policy, we suggest including energy-related questions in the research conducted by Brazilian Institute of Geography and Statistics (IBGE)—Census and National Household Sample Survey (PNAD), without extra cost. Furthermore, a larger sample would improve the results of the econometric model. In future works, we intend to test the methodology on other samples and establish a definition of energy poverty in Brazil. In addition, we intend to updating the data from the two communities studied and monitoring the PCI.

Finally, PCI meets Brazil’s need to study, define, and establish metrics to understand energy poverty. The strategy presented is a valuable tool to monitor electricity affordability, direct government programs and public policies that aim to guarantee electricity access for all.