Energy for Sustainable Development 70 (2022) 430–441 Contents lists available at ScienceDirect Energy for Sustainable Development The direct rebound effect for two income groups: The case of Paraguay Martín Bordón-Lesme a,⁎, Jaume Freire-González b, Emilio Padilla Rosa a a Department of Applied Economics, Univ. Autonoma de Barcelona, 08193 Bellaterra, Spain b Institute for Economic Analysis (CSIC) and Barcelona School of Economics, 08193 Bellaterra, Spain a r t i c l e i n f o a b s t r a c t Article history: We estimate the direct rebound effect (DRE) for all energy services requiring electricity for their provision in Received 29 September 2021 Paraguayan households. Using recent panel data from 2001 to 2017, we estimate the magnitude of the DRE at Revised 12 July 2022 the province and municipality levels. Because we estimate the DRE through the own-price elasticity of electricity Accepted 1 August 2022 demand, we not only provide the first empirical evidence of the DRE for Paraguay, a developing country, but also Available online 26 August 2022 update the study of Paraguay's residential electricity demand. Our findings suggest a positive DRE emerges after Keywords: an improvement in energy efficiency, but the magnitude of the DRE does not completely reduce the resulting Direct rebound effect energy savings. We find a lower DRE in low-income households, which may be explained by two factors: electric- Electricity consumption ity is not the main source of energy for most low-income households, and most clandestine electricity connec- Energy services tions are from low-income households. Paraguay is one of the countries with the highest generation of Low-income households electricity per capita through hydroelectric plants. However, this electricity supply does not match electricity Panel data consumption, especially in low-income households, because of distribution issues in relation to energy sources. We derive from our findings some policy measures to improve the situation. © 2022 The Authors. Published by Elsevier Inc. on behalf of International Energy Initiative. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Introduction developing country. Second, it updates the study of the Paraguayan res- idential electricity demand (Westley, 1984) because we estimate the Empirical evidence shows an improvement in energy efficiency DRE through the own-price elasticity of electricity demand. Third, it leads to a lower than proportional reduction in energy savings due to provides updated and useful information to Paraguayan policymakers behavioral responses from consumers, which is known as the rebound at the province and municipality levels. Finally, it fosters the debate effect. This effect reduces the amount of energy savings but also involves about the potential difference in the size of the DRE for different income an improvement in social welfare because energy service consumption groups. No unanimous definition of energy poverty has been estab- increases as the effective cost is reduced by the energy efficiency im- lished to date; however, the existing literature combines the concepts provement (direct rebound), and less income must be used to purchase of an energy ladder (from the less advanced to most advanced energy the same energy services, which increases the income available to con- sources) and energy equity to define energy poverty (Sovacool, 2014). sume other energy goods and services (indirect rebound). Thus, ap- The study of the size of the DRE for different income groups may facili- praisal of the rebound effect depends on the size of consumers' tate better assessments of this issue. benefits relative to the environmental costs of the energy savings reduc- This paper is arranged as follows. The DRE in developing and middle- tion and the associated pollutant emissions (Sorrell, 2018). In addition, income countries section reviews empirical evidence related to this re- the literature suggests the rebound effect is greater in low-income search; the Methodology and variables section explains the methodol- groups because their demand for energy services is far from their satia- ogy and the variables used to estimate the DRE for Paraguay, as well tion levels (Milne & Boardman, 2000; Sorrell, 2007). Therefore, the pur- as the data employed for each variable; the Econometric model pose of this article is to estimate the direct rebound effect (DRE) for estimation section shows the econometric model estimated and in- electricity in two income groups in Paraguay at the department (prov- cludes a discussion of results; and the Policy implications section pre- ince) and district (municipality) levels. sents the conclusions and policy implications. This paper contributes to the literature in four ways. First, this paper provides the first empirical evidence of the DRE for Paraguay, a DRE in developing and middle-income countries ⁎ Corresponding author. To contextualize our study, this section summarizes empirical evi- E-mail address: martinadolfo.bordon@autonoma.cat (M. Bordón-Lesme). dence of the DRE for developing countries with a particular focus on https://doi.org/10.1016/j.esd.2022.08.002 0973-0826/© 2022 The Authors. Published by Elsevier Inc. on behalf of International Energy Initiative. This is an open access article under the CC BY license (http://creativecommons.org/ licenses/by/4.0/). M. Bordón-Lesme, J. Freire-González and E. Padilla Rosa Energy for Sustainable Development 70 (2022) 430–441 Paraguayan neighboring countries. We focus on studies applying econo- devices in explaining residential electricity consumption, which may be metric methods to estimate the DRE for residential energy services. the case in Paraguay. Similarly, Liddle and Huntington (2020) analyzed Given the existing empirical evidence, the DRE for most energy ser- the residential electricity demand for high- and middle-income coun- vices is expected to be around 30 % in developed countries (Greening tries and found smaller price elasticity, greater income elasticity, smaller et al., 2000). However, according to the literature, the rebound effect heating elasticity, and larger cooling elasticity for middle-income coun- in developing countries tends to be greater. Possible explanations for tries than for high-income/Organisation for Economic Co-operation and this finding include the following: Development (OECD) countries.2 Regarding price elasticity, according to Liddle and Huntington (2020), many non-OECD countries have sub- (i) In developing countries, the demand for energy services is far sidies for electricity price, thus diminishing the price response. Further- from residents' satiation levels (Sorrell, 2007). more, they argued most non-OECD countries present only the average (ii) Residents experience rapid accumulation of energy-using tech- electricity price and not the actual price charged to the different types nologies and more energy-intensive consumption due to a high of subscribers, which may affect the results for middle-income growth rate (van den Bergh, 2011). countries. We explain how we addressed this issue in the next section. (iii) Energy is relatively more expensive given residents' low wages; To our knowledge, no consensus has been established among re- thus, energy conservation may induce a larger re-spending effect searchers regarding how a change in price or an improvement in energy (van den Bergh, 2011). efficiency (depending on the case) may affect energy consumption for different income groups. Because empirical evidence indicating Labidi and Abdessalem (2018) estimated a DRE of 81.7 % for electric- whether the DRE is greater in low-income households is inconsistent, ity end uses in Tunisia through a panel data model with fixed effects for the results of this research can provide insight into this topic. Most re- 21 cities and 5 nonconsecutive years (1995, 2000, 2005, 2010). The vised studies have highlighted the relevance of the stock of electric con- magnitude of the effect is relevant for the energy service of refrigeration version devices in explaining the consumption of residential electricity, because this service has accounted for the greatest share of residential as well as the subsidized prices, especially for low-income households. electricity consumption since 1984 (Labidi & Abdessalem, 2018). Alvi However, relating the energy poverty literature to DRE estimation et al. (2018) found DREs of 42.9 % and 69.5 % for residential electricity may facilitate an understanding of the potential differences among the consumption in Pakistan in the short term and the long term, respec- reactions of different income groups to improvements in energy effi- tively. They used an error-correction model with time series data from ciency or to changes in energy prices. 1973 to 2016. If consumers respond the same way to a decrease in en- ergy prices as they do to more efficient energy systems (given both de- Methodology and variables crease the effective cost of energy services), then the own-price elasticity of electricity demand can be used as a proxy of the DRE. Re- Because we estimate the DRE through the own-price elasticity of garding Paraguayan neighboring countries, Casarin and Delfino (2011) electricity demand, we should consider the assumptions involved estimated own-price elasticity values of 10 % and 20 % for the residential when analyzing our results. These assumptions are as follows (Sorrell, electricity demand in Greater Buenos Aires (Argentinian capital) in the 2007; Sorrell & Dimitropoulos, 2007): short term and the long term, respectively. They found increases in the 1. Symmetry—Consumers respond the same way to a decrease in en- stock of air conditioners and regulatory tariffs that fixed the electricity ergy prices as they do to more efficient energy systems3 because price for several years tended to increase residential electricity de- more efficient systems reduce the effective cost of energy services. mands. These results may also be relevant for Paraguay because it has 2. Exogeneity—Energy prices do not affect energy efficiency. To fulfill a warmer climate than Buenos Aires. In addition, the National Adminis- this assumption, the period analyzed must be characterized by stabil- tration of Electricity (ANDE, 2020) fixed electricity prices for ity or decreases in energy prices because increasing energy prices Paraguayan households. Villareal and Moreira (2016) estimated an may induce an improvement in the energy efficiency of energy sys- own-price elasticity of electricity demand between 23 % and 44 % for tems. Brazilian households. These values are relevant for the energy services 3. Constant energy efficiency—The efficiency of an energy system does of electric showers and refrigeration because they account for a consid- not change with the amount of energy service used. erable share of residential electricity consumption in Brazil (EPE, 2020). Comparing the electricity demand among these countries is particularly The main definition of the DRE is the efficiency elasticity of the de- pertinent because Paraguay shares the ownership of two hydroelectric mand for useful work (Berkhout et al., 2000; Sorrell, 2007; Sorrell & plants, including one with Argentina (Yacyreta) and the other with Dimitropoulos, 2007). Nevertheless, we use the own-price elasticity of Brazil (Itaipu). Furthermore, in relation to the empirical evidence of electricity demand as a proxy for the DRE given data availability issues the DRE for other developing countries, the magnitude of the DRE for (Freire-González, 2010; Wang et al., 2014, 2016; Bordón-Lesme et al., Buenos Aires–Argentina and for Brazil is relatively small. However, ac- 2022). See Sorrell (2007, 2009) and Sorrell and Dimitropoulos (2007) cording to the Handbook of Statistics of the United Nations Conference for further DRE estimation methods. The primary definition of the DRE on Trade and Development (UNCTAD), these two neighboring countries is as follows: of Paraguay may be considered to have more advanced economies among developing countries (UNCTAD, 2020). ηε ðEÞ ¼ ηε ðSÞ  1 ð1Þ Regarding studies differentiating among income groups, Zhang and Peng (2017) presented a study similar to ours by estimating the DRE of China's residential electricity consumption for two income regimes The first term, ηε(E), represents the efficiency elasticity of the and for two cooling degree day (CDD) levels. In line with the literature, demand for energy, and the second term, ηε(S), is the efficiency which suggests a higher DRE in low-income groups, they found a elasticity of the demand for useful work. For the residential case, greater DRE under a low-income regime (68 %) than under a high- examples of useful work are residential energy services such as income regime (55 %). Moreover, for CDD levels, the authors found a heating, lighting, or cooking. greater DRE under a high CDD level (90 %) than under a low CDD level Following previous research on the topic (Alvi et al., 2018; Belaïd (75 %).1 They also highlighted the relevance of the stock of space cooling et al., 2018; Chitnis et al., 2013; Freire-González, 2010; Labidi & 1 2 They also found a greater DRE under a heavy rainfall regime (86 %) than under a light Paraguay was included as a middle-income country in their data set. 3 rainfall regime (68 %). In our case, the energy systems are residential energy conversion devices. 431 M. Bordón-Lesme, J. Freire-González and E. Padilla Rosa Energy for Sustainable Development 70 (2022) 430–441 Abdessalem, 2018; Zhang & Peng, 2017), we use a double logarithmic significant. Moreover, we retain the LPG price variable in Model 4 be- functional form to estimate the DRE for residential energy services re- cause it is significant at the 10 % level. We include the lagged dependent quiring electricity in Paraguay. The model is as follows: variable in the models to deal with autocorrelation. We also add cross- section weights in the models to address potential cross-section hetero-   Eit  skedasticity. Similar specifications have been used widely in previous ln ¼ α þ β1 ln P Eit þ β2 ln P LPGit þ β3 lnCDDit hhit  1 research (Sorrell & Dimitropoulos, 2007).   Eit  1 þ β4 lnHDDit þ β5 ln Y it þ β6 ln þ εit ð2Þ hhit  1 Results where it represents the data of each geographic subdivision Table 4 shows the results of the estimations of the residential elec-   (i) per time period (t) for each variable. The dependent variable hh Eit is the average tricity demand models. Models 1 and 3 correspond to non-low- it electricity consumption; PLPG it denotes the price of liquefied petroleum income households, whereas Models 2 and 4 correspond to low- gas (LPG); Yit represents the income variable; and CDDit and HDDit are income households. climate variables—CDDs and heating degree days (HDDs), Regarding the models for non-low-income households (Model 1 and respectively. Our variable of interest, the price of residential electricity, Model 3), the coefficients of the own-price elasticity of electricity de- changes depending on the amount consumed and is therefore charged mand have a negative sign between 23 % and 60 %, with a significance after consumption (PE it−1). Finally, (Eit−1/hhit−1) is the lagged level of 0.001; that is, an increase in the price of electricity reduces its dependent variable, and εit represents the error term. consumption. The climate variables show a positive sign at a signifi- Regarding the income groups, two price categories exist in Paraguay, cance level of 0.001. Thus, as the temperature reaches below or above and both have their own price levels according to the amount of elec- some thresholds, the consumption of electricity for cooling and heating tricity consumed by a household. Low-income households are regis- devices increases. Furthermore, the CDD coefficient is significant at both tered under price category 141, which corresponds to a subsidized geographic levels, whereas the HDD coefficient is not significant at the price at the ANDE.4 Appendix 1 and Appendix 2 illustrate the price cat- province level. At the municipality level, where both are significant, egories per consumption level and the corresponding discount rate for the CDD coefficient has a greater magnitude. Thus, the impact of space low-income households registered in the social tariff program. The cooling devices on electricity consumption may be greater than that of LPG price is the same for all households and does not change with the space heating devices. Therefore, an increase in the stock of air condi- amount consumed. tioners may increase the residential electricity demand (Casarin & Household disposable income and the climate variables (CDDs and Delfino, 2011; Liddle & Huntington, 2020). Regarding this issue, HDDs) are available only at the province level and are therefore the Appendix 4 and Appendix 5 show space cooling accounts for a greater same for all estimations. share of electricity consumption than space heating. The income vari- The dependent variable is available for both income groups and both able is significant only at the municipality level, with a 0.01 significance geographic subdivisions. Thus, we estimate the coefficients of the equa- level, and it has a positive sign; that is, as income increases, residential tion for both types of households, low-income and non-low-income electricity consumption also increases. The lagged dependent variable households, at the province and municipality levels. For all models, suggests electricity consumption in period t − 1 has a positive effect the monetary variables are constant at 2017 prices. Table 1 depicts the on the current period because it has a positive sign significant at the data development process for all variables. 0.001 level. The LPG price coefficients have significance levels of 0.01 and Econometric model estimation 0.001 at the municipality and province levels, respectively. The sign of the coefficients indicates a substitutive relationship between electricity Following Eq. (2), we estimate our model at the province and munic- and LPG for the demand of residential energy services; that is, an in- ipality levels for both income groups (low-income and non-low-income crease in LPG prices increases electricity consumption in the residential households). The Hausman test confirms differences exist between sector in non-low-income households. Therefore, space heating and fixed and random effect estimators in all models at both geographic cooking (energy services commonly provided by LPG, firewood, or char- levels (Table 2). Therefore, we prefer fixed effect estimates for all coal) would be replaced by electricity. models. Regarding the models for low-income households (Model 2 and Evidence for Paraguay may differ from other empirical evidence of Model 4), the coefficients of the own-price elasticity of electricity de- the DRE, especially evidence for developed countries. The nominal mand also have the expected negative sign and are 18 % at the mu- price of electricity was the same from 2005 to 2016; that is, the ANDE nicipality level and 14 % at the province level. At the province level, fixed the price during those years. The only change, which occurred in the coefficient is significant at the 0.05 level, whereas at the munici- 2017, was the addition of price subcategories four, five, and six for cat- pality level, the coefficient is significant at the 0.01 level. The HDD egory 142 (non-low-income households), which correspond to the coefficient is not significant at the municipality or the province prices without social tariff discounts, as shown in Appendix 1. Thus, level. Nevertheless, the next section shows low-income households we assumed the price and consumption of electricity in Paraguay mostly use firewood for space heating services (Fig. 1), whereas could not be cointegrated over time. Nevertheless, we performed the middle- and high-income households mostly use electricity for this Pedroni residual cointegration test for the four models. As expected, al- energy service (Appendix 4 and Appendix 5). Furthermore, the most all the statistics confirmed the null hypothesis of no cointegration, CDD coefficient has a positive relationship with residential electric- as shown in Table 3. Therefore, we do not apply an error-correction ity consumption, with significance levels of 0.001 and 0.01 at the model for our estimates. municipality and province levels, respectively. The coefficients of In the models for low-income households at the municipality and income and the lagged dependent variable also have positive province levels, we exclude the HDD variable because this type of relationships with residential electricity consumption. In this case, household does not use electricity for the energy service of space the income variable is significant at both geographic levels, with sig- heating. We exclude the income variable in Model 3 because it is not nificance levels of 0.001 (municipality) and 0.05 (province). The lagged dependent variable is significant at the 0.001 level for both 4 To be registered under that price category, households must provide legal documents geographic levels. to the ANDE office that prove a certain income level. See https://www.ande.gov.py/ For low-income households, LPG price coefficients have a com- infodata.php?catid=6#.X8FQBc1Kg2w for further details. plementary relationship (negative sign) with respect to residential 432 M. Bordón-Lesme, J. Freire-González and E. Padilla Rosa Energy for Sustainable Development 70 (2022) 430–441 Table 1 Definitions of the variables of the model. Variable Definition Availability Time Data sources Expected coefficient sign   Eit Average electricity consumption. Aggregate electricity Data at the municipality and province Annual (2001 to Administración Positive for the hhit consumption per municipality and province divided by levels. Data available for low-income 2017) Nacional de lagged the registered subscribers at the municipality and and non-low-income households. Electricidad (ANDE, dependent province levels. 2020) variable   Eit  1 hhit  1 (PE it−1) The real price charged at both geographic levels. We Data at the municipality and province Annual Administración Negative calculate the real price charged to consumers by levels. Data available for low-income (2001 to 2017) Nacional de allocating the price categories for both income groups and non-low-income households. Electricidad (ANDE, according to their kWh range of consumption. For 2020) low-income households, after allocation to the price categories, we calculate the corresponding discount. See Appendices 1 and 2. PLPG it The real LPG price. Data at the national level. The same data Annual (2005 to SIEN Statistics – Negative for both income groups at both 2017) Viceministerio de geographic levels. Data available from Minas y Energía 2005 to 2017. (2020) CDDit Cooling degree days. A base temperature of 22 degrees Data at the province level. The same Daily data Dirección de Positive Celsius (see Appendix 3 for further details about the data for both income groups at both aggregated into Meteorología e calculation) (www.degreedays.com/calculation) geographic levels. annual data (2001 to Hidrología (2020) 2017, with gaps) HDDit Heating degree days. A base temperature of 21 degrees Data at the province level. The same Daily data Dirección de Positive Celsius (see Appendix 3 for further details about its data for both income groups at both aggregated into Meteorología e calculation) (www.degreedays.com/calculation) geographic levels. annual data (2001 to Hidrología (2020) 2017, with gaps) Yit Real household income. Data at the province level. The same Annual (2001 to DGEEC - Dirección Positive data for both income groups at both 2017, with gaps) General de geographic levels. Estadística, Encuestas y Censos (2020) *Some daily data for the minimum and maximum temperatures are missing for most provinces. Therefore, some provinces have data gaps in some years (unbalanced panel). There are 47 total annual gaps among the sample. *Only five provinces do not have missing annual income data. Therefore, the remaining provinces have gaps in most years for this variable (unbalanced panel). There are 138 total annual gaps among the sample. *For the prices of electricity and LPG, the consumer price index (CPI) falls under the same category according to the Central Bank of Paraguay (BCP, 2020). This CPI is at the national level. The income variable was already obtained with 2017 constant prices according to Dirección General de Estadística, Encuestas y Censos (DGEEC, 2020) electricity demand. That is, an increase in the price of LPG would re- and low-income households, respectively.5 Moreover, the significant in- duce electricity consumption. The potential income constraint could fluence of LPG price in explaining residential electricity consumption is explain this relationship. Moving up the energy ladder by changing consistent with the finding of Bordón-Lesme et al. (2022). They esti- their energy consumption from traditional energy sources to elec- mated the DRE of residential electricity for Spain and found other en- tricity could be expensive because of the capital cost of the more ef- ergy sources influenced it. ficient, electrically run conversion devices (Van der Kroon et al., Because we estimate the DRE for a collection of energy services that 2013). Thus, as shown in the next section (Fig. 1), low-income require electricity, our results are more relevant for the energy service households may prioritize the substantial energy services of cooking with the greatest share of electricity consumption. Hence, for low- and water heating because both can be provided by traditional en- income households, our results are more relevant for the energy service ergy sources instead of the modern energy services provided by elec- of food preservation because this energy service amounts to 37.3 % of tricity (Sovacool, 2014). the total electricity consumption of this income group. For non-low- Therefore, considering the assumptions explained in the income households, the magnitude of the DRE is more relevant for the Methodology and variables section, the DRE of electricity for Paraguay energy services of space cooling and water heating because both energy could be between 23 % and 60 % for non-low-income households and services have the greatest share of electricity consumption in the high- between 14 % and 18 % for low-income households. That is, because of income and middle-income households (29.6 % and 25.8 %), respec- an improvement in electricity efficiency with respect to a scenario tively. See Appendix 6 for further details. where there are no behavioral responses from consumers, the electric- Moreover, for all models, the exogeneity assumption should not be a ity savings would be reduced up to 60 % and 18 % in non-low-income source of bias because the period analyzed is characterized by stable electricity prices (the ANDE fixed the prices). However, the symmetry assumption may provide an upper bound for the magnitude of the DRE because consumers could easily notice the electricity prices instead Table 2 of searching for the improvements in electricity efficiency. Hausman test of Model 1 to Model 4. Correlated random effects – Hausman test Discussion of the results Cross-section random Chi-sq. statistic Chi-sq. df Prob. Model 1 326.0419 6 0.0000 The magnitude of the DRE for non-low-income households, which is Model 2 202.9900 5 0.0000 between 23 % and 60 %, falls in the range of the expected values of the Model 3 30.4818 4 0.0000 DRE for developed countries at approximately 30 % (Freire-González, Model 4 20.7580 5 0.0009 5 Note: Test performed after using random and fixed effects for each model. Usually, energy efficiency improvements are due to more efficient conversion devices. 433 M. Bordón-Lesme, J. Freire-González and E. Padilla Rosa Energy for Sustainable Development 70 (2022) 430–441 Table 3 Cointegration test for Model 1 to Model 4. Alternative hypothesis: common AR coefficients (within-dimension) Model 1 Model 2 Model 3 Model 4 Prob. Weighted stat prob. Prob. Weighted stat prob. Prob. Weighted stat prob. Prob. Weighted stat prob. Panel v-Statistic 0.9966 0.9992 0.9988 1.0000 0.1258 1.0000 0.6915 0.8790 Panel rho-Statistic 1.0000 1.0000 1.0000 1.0000 0.7572 0.6441 0.9191 0.9402 Panel PP-Statistic 0.8269 0.0000 0.0000 1.0000 0.0004 0.0105 0.1554 0.5637 Panel ADF-Statistic 0.6222 0.9389 1.0000 1.0000 0.5497 0.7351 0.6180 0.7766 Alternative hypothesis: individual AR coefficients (between-dimension) Model 1 Model 2 Model 3 Model 4 Prob. Prob. Prob. Prob. Group rho-Statistic 1.0000 1.0000 0.9822 0.9956 Group PP-Statistic 0.0000 0.7255 0.0000 0.6322 Group ADF-Statistic 1.0000 1.0000 0.8551 0.4580 Note: The test was performed after using random and fixed effects for each model. Table 4 Empirical estimates of households' electricity demand in Paraguay. Dependent variable: ln Municipality fixed effects Province fixed effects (Eit/hhit) Non-low-income households (Model 1) Low-income households (Model 2) Non-low-income households Low-income households (Model 4) (Model 3) α Coef. 4.5606⁎⁎⁎ 0.3668 −0.3251 −0.5795 Std.Err (0.3878) (0.3008) (1.0308) (1.1278) (ln (PE it − 1)) Coef. −0.5972⁎⁎⁎ −0.1786⁎⁎⁎ −0.2302⁎⁎⁎ −0.1379⁎ Std.Err (0.0232) (0.0162) (0.0634) (0.0597) lnPLPG it Coef. 0.0410⁎⁎ −0.0925⁎⁎⁎ 0.1132⁎⁎⁎ −0.0565 Std.Err (0.0156) (0.0127) (0.0275) (0.0326) lnCDDit Coef. 0.2514⁎⁎⁎ 0.2567⁎⁎⁎ 0.2508⁎⁎⁎ 0.3044⁎⁎ Std.Err (0.0183) (0.0221) (0.0389) (0.0980) lnHDDit Coef. 0.0162⁎⁎⁎ Std.Err (0.0038) lnYit Coef. 0.0164⁎⁎ 0.1287⁎⁎⁎ 0.0927⁎ Std.Err (0.0062) (0.0103) (0.0399) (ln(Eit−1/hhit−1)) Coef. 0.5751⁎⁎⁎ 0.7819⁎⁎⁎ 0.8654⁎⁎⁎ 0.8333⁎⁎⁎ Std.Err (0.0179) (0.0153) (0.0563) (0.0529) Periods 13 13 13 13 Cross-sections 189 187 16 16 Observations 1235 1207 165 85 Panel Unbalanced Unbalanced Unbalanced Unbalanced Weighted statistics R2 0.9957 0.9826 0.9922 0.9800 Prob (F-Statistic) 0.0000 0.0000 0.0000 0.0000 Durbin–Watson Stat 2.2026 2.3132 2.3336 2.5061 Unweighted statistics R2 0.9561 0.8134 0.9879 0.7730 Durbin–Watson Stat 2.4575 2.8149 2.8039 3.2068 ⁎ p < 0.05. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.001. 2017; Greening et al., 2000; Sorrell, 2009). Intriguingly, the magnitude in low-income households. We identify two factors that may explain of the DRE for low-income households is between 14 % and 18 %. this peculiarity of our results. When comparing both types of households, we observe the DRE for i. Electricity is not the main energy source for most low-income electricity is lower in low-income households because the own-price households: elasticity of electricity demand is lower for these households. This fea- ture is present at both the municipality and province levels, which may seem counterintuitive because the literature suggests the DRE As Fig. 1 shows, electricity accounts for only 16 % of total residential should be higher in low-income groups, given their demand for energy energy input consumption7 in low-income households, whereas char- services is far from their satiation levels (Milne & Boardman, 2000; coal (41 %) and firewood (34 %), which are traditional energy sources Sorrell, 2007). Appendix 7 shows the robustness checks for models 1 (Van der Kroon et al., 2013), account for 75 % of this total. However, to 4,6 which reinforce the finding that there is a lower DRE for electricity these energy sources are used primarily for cooking, an energy service that accounts for 69 % of total residential energy input consumption in low-income households. Notably, reliance on traditional energy sources 6 For comparison purposes, we leave out the variables with a lower significant level in 7 the original models, as well as some coefficients that were not present in models 2 to 4. Energy input is the energy before its transformation into useful energy. 434 M. Bordón-Lesme, J. Freire-González and E. Padilla Rosa Energy for Sustainable Development 70 (2022) 430–441 Fig. 1. The share of energy input, in kiloton of oil equivalent (KTOE), per energy service in Paraguayan low-income households of 2011. Source: Personal elaboration from BNEU (2020). for cooking and the lack of access to a bare minimum of electricity are indicates that in Paraguay, the use of traditional energy sources may be methods for measuring energy poverty (Sovacool, 2014). due to income constraints, which is an energy poverty indicator. In Thus, Paraguay presents an unusual case relative to other developing Paraguay, most low-income households are located in slums; thus, economies because it has one of the highest per capita electricity gener- most of these households do not have a property title (RAP - ation levels through hydropower, given it shares ownership of two hy- Relevamiento de Asentamientos Precarios, 2015). Without this docu- droelectric power plants with its neighboring countries: Itaipu in Brazil ment, they cannot ask for a legal electricity connection to the ANDE, and Yacyreta in Argentina (Blanco et al., 2017).8 However, this electric- which may be another reason most clandestine electricity connections ity supply does not match electricity consumption, especially in low- are from low-income households. income households (Fig. 1). Furthermore, according to the (IEA, 2020), Given points i and ii, electricity price and potential improvements in 99.3 % of the Paraguayan population has access to electricity. Thus, com- electricity efficiency would not affect low-income households as much pared to other developing areas in which the percentage of electricity as they would affect non-low-income households. access is relatively small, such as developing Asian countries and most African countries (IEA, 2020), Paraguay has high electricity supply and Policy implications access. Therefore, in Paraguay, the issue may be moving up the energy ladder by increasing electricity consumption, especially in low-income The two factors we identify in the above section would be a conse- households, rather than access to the electricity grid. quence of electricity consumption barriers affecting only the poverty- This discrepancy between the supply of and access to electricity and stricken group. The higher capital cost of the energy conversion devices electricity consumption by low-income households can be explained by run by electricity could explain why low-income households are not distribution issues related to the relevance of different energy sources affected by electricity prices or electricity improvements in home appli- for different income groups, an issue analyzed in the energy poverty lit- ances and so, they rely on energy conversion devices that require erature (Halff et al., 2014). charcoal or firewood. Given the income constraint of Paraguayan low-income households, ii. Clandestine electricity connections: they might be focused on just consuming the energy services for basic living, such as cooking and water heating. These energy services can also be provided more or less effectively without electricity, using bio- Most clandestine electricity connections are from low-income mass. On the other hand, non-low-income households might decide households. Thus, these households do not react to prices, which may their energy services requirements according to their comfort needs. result in low elasticity. Appendix 4 and Appendix 5 match the energy Thus, space cooling has a greater share of the total energy consumption input with the share of energy services for middle- and high-income for this type of households. More advanced energy carriers, such as elec- households. The consumption of modern energy sources tends to in- tricity, provide the possibility to access modern energy services, such as crease as income increases, which is consistent with the energy ladder space cooling, lighting, and more efficient cooking. model. Furthermore, high-income households still use some traditional Another type of electricity consumption barrier could be the energy input in their bundle of energy sources, as the energy stacking incompatible periodicity between irregular informal incomes and elec- model suggests (Van der Kroon et al., 2013)9; this may also be explained tricity bills. The results of a survey to low-income households in a as cultural factor (Masera et al., 2000). However, the fact that most clan- Paraguayan poverty-stricken settlement suggest that, on average, destine electricity connections are from low-income households (Fig. 2) those households spend on charcoal and firewood as much as they do 8 on electricity, and show a lack of energy conversion devices run by elec- Itaipu and Yacyreta have annual average production levels of 98,287 GWh and 20,867 GWh, respectively (Blanco et al., 2017). tricity, especially for cooking and space cooling (Llamosas et al., 2018). 9 The energy ladder and energy stacking are models analyzed in the energy poverty lit- Therefore, in the short run, it is less risky to buy a bag of charcoal daily erature. or to collect firewood whenever necessary than paying the monthly 435 M. Bordón-Lesme, J. Freire-González and E. Padilla Rosa Energy for Sustainable Development 70 (2022) 430–441 Fig. 2. Clandestine electricity connection per income level. electricity bill; regardless of whether they are registered in the social policymakers should expect a greater DRE for electricity for this type electricity tariff program and regardless of the charcoal and firewood ef- of household and, consequently, an increase in total electricity con- fectiveness versus electricity effectiveness in providing energy services. sumption. Because the electricity bill is charged after consumption, low-income The Vice Ministry of Mines and Energy estimates that, by the year households might not know for sure what will be their discount per- 2040, it would be necessary to incorporate new hydroelectric plants to centage or what range of consumption they will have. Moreover, if match the internal demand for electricity. If the DRE for low-income they have an irregular income, they might not know if they would be households is considered in a scenario where the social tariff discount able to pay the electricity bill at the end of the month. has a real benefit for this type of households (if the alternative and com- Hence, given these electricity consumption barriers, such as the cap- plementary policies suggested were applied), the internal demand for ital cost of energy conversion devices run by electricity, the income con- electricity in Paraguay might exceed the supply before 2040. Currently, straint of the Paraguayan low-income households, and the available the Paraguayan senate is passing a bill that fosters the production of energy carrier per energy service, it would be desirable to develop pol- electricity from non-conventional non-hydraulic renewable energy icies promoting the increase of the stock of energy conversion devices sources in order to diversify the production of energy (Florentín, run by electricity in low-income households. If this is implemented, 2021). Thus, if there is a positive appraisal of the Paraguayan govern- the social tariff discount would provide a real benefit to low-income ment towards increasing levels of the DRE in low-income households, households, as they would be able to consume electricity more effec- given their low levels of electricity consumption, production of energy tively. Then, the transition from charcoal and firewood to electricity through non-conventional renewable energy sources should be consid- would be easier for low-income households and so their consumption ered. of energy services would be more effective. An energy efficiency improvement of an energy service can come in The incompatibility between irregular informal incomes and elec- a variety of technical changes, such as more efficient conversion devices tricity bills could be tackled introducing an alternative payment option (energy system), more efficient energy sources (energy transition) or for the electricity bill. This can be the pay-as-you-go prepayment of en- investments in more efficient infrastructure, such as proper household ergy, which is implemented by several energy companies in the UK, insulation (energy system). Most of the rebound effect studies focus such as British Gas, E.ON, SSE, and Ecotricity, among others. Those com- on measuring the consumption of energy and/or its related greenhouse panies provide a pay-as-you-go meter which allows households to pay a gas emissions after an energy efficiency improvement. The findings of certain amount of money for electricity before using it. This payment Roy (2000) are an example of it, because she measured the consump- method would allow low-income households to budget their expendi- tion of kerosene for lighting after it was improved by solar lanterns, ture on electricity consumption as the timespan of their income will which is explained above. However, the increase in welfare in these match the electricity bill. Then, they might be able to spend less of Indian households after improving their energy service of lighting was their income in energy because electricity is more effective in providing not entirely measured. the energy services of cooking and water heating than charcoal and The energy poverty literature provides us with several concepts and firewood. measurements for it. The energy ladder model suggests that households Roy (2000) studied the case of rural lighting where The Ministry of will change their energy sources for more advanced ones as their in- New Renewable Energy Sources replaced kerosene lamps with solar come increases, while according to the energy stacking model a house- lanterns. She estimated a DRE between 50 and 80 %. That is, up to 50 hold use a bundle of multiple energy sources (Van der Kroon et al., to 80 % of the potential kerosene savings were reduced. However, she 2013). In Paraguay, there might be a mix between the two models. also mentioned that if the energy service of cooking was included, the First, because in both extremes of the Paraguayan social levels, high- DRE would have been as much as 200 %. This is because the efficiency income and low-income households, there is a use of multiple energy improvement did not only cause more lighting in households, but also sources to satisfy their households' energy service needs. Second, be- the kerosene saved from the more efficient lighting was used for cause in the energy sources bundle of high-income households there cooking and, in some cases, the saved kerosene was even sold by is a higher share of advanced energy sources (electricity and LPG) some households to their neighbors (Roy, 2000). Hence, if the low- than in the one of low-income households. Moreover, a concept of en- income households get a real benefit from the social tariff discount, ergy poverty is the absence of choice in accessing energy services that given additional and complementary policies, the Paraguayan are affordable, of high-quality, environmentally benign, adequate and 436 M. Bordón-Lesme, J. Freire-González and E. Padilla Rosa Energy for Sustainable Development 70 (2022) 430–441 reliable (Reddy et al., 2000). González-Eguino (2015) clarifies this issue by distribution issues related to the share of different energy sources by stating that the households' needs are not energy consumption, but for different income groups; this is an issue analyzed in the energy pov- the provision of energy services (cooking, space heating, lighting, etc.). erty literature (Halff et al., 2014). Thus, our results introduce a new line However, as we explained at the beginning of this section, it is the diffi- of research by exploring the relationship between the DRE and energy culty in accessing advanced energy sources what might define the con- poverty. In further research, gathering available data regarding the sumption of each energy service. For example, in Paraguay, space prices of firewood and charcoal (the principal energy sources for cooling, an energy service that is provided by an advanced energy Paraguayan low-income households) would be relevant. Likewise, the source, such as electricity, does not have the same share of energy con- specific factors affecting only the poverty-stricken group, such as the sumption in high-income and low-income households. Nevertheless, two points highlighted in this research, should be considered in further the rebound effect literature acknowledge increasing magnitudes of re- analyses. bound effect as a potential increase in social welfare because of the in- Another novelty of this study is that it provides the first empirical ev- crease in energy consumption (Sorrell, 2018). idence of the DRE for Paraguay, a developing country. Empirical evi- Therefore, energy policymakers in developing countries should con- dence of the DRE in the South American region is lacking because sider linking the lessons of the energy poverty literature with those of most studies focus on the determinants of electricity demand rather the rebound effect literature because the former would determine to than residential DREs for electricity (Agostini et al., 2012; Bendezú & which energy source households should transition whereas the latter Gallardo, 2006; Casarin & Delfino, 2011; Laureiro, 2018; Orejuela et al., would provide information about how these households are responding 2015; Villareal & Moreira, 2016). Furthermore, because we estimated to the energy source change. For climate change mitigation purposes, it the DRE through the own-price elasticity of electricity demand, we are is more effective to switch to non-conventional renewable energy also updating the study of the Paraguayan residential electricity de- sources than encourage energy efficiency, since CO2 emissions are not mand. Given this estimation method, food preservation is the energy reduced as expected (Chen et al., 2021; Chitnis & Sorrell, 2015; service associated with the greatest electricity consumption for low- Druckman et al., 2011). income households; by contrast, space cooling and water heating ac- With these policy recommendations and as explained in the count for the greatest consumption for non-low-income households. Discussion of the results section, Paraguayan low-income households In further research, the DRE for each energy service should be estimated would be able to improve their current situation in a sustainable man- by differentiating each energy service according to the required energy ner by transitioning from charcoal and firewood to electricity, which is source (Hunt & Ryan, 2014). Moreover, data availability permitting, in- a cleaner energy source. First, electricity requires less energy input per troducing the efficiency of each energy service into the price of the cor- energy service; therefore, low-income households would spend less responding energy source may reduce the classical assumptions of the on energy consumption. Second, electricity reduces household drudg- DRE through price elasticity (Hunt & Ryan, 2014). ery, which is associated with increasing economic development levels. Finally, our findings suggest alternative energy policies to the social Third, electricity alleviates indoor air pollution caused by the inefficient tariff discount for residential electricity should be implemented to alle- use of biomass, which improves health, especially among women and viate energy poverty. Despite the existing electricity discounts for low- children. Finally, reduced drudgery and cleaner cooking methods pro- income households, electricity is not their main energy source. There- vided by electricity are relevant to gender equality (Barnes et al., 2014, fore, given our results, linking electricity consumption barriers to the p 16). level of electricity access may improve energy poverty measurements. These potential consumption barriers may explain why electricity is Conclusions not the main energy source for low-income households in a country rich in hydroelectricity such as Paraguay. Consequently, our research The literature on DREs suggests low-income groups show a greater provides a new path to expand the understanding of energy poverty is- increase in energy consumption than high-income groups after an im- sues as well as magnitude estimations of DREs by combining both con- provement in energy efficiency, given their level of energy consumption cepts. is far from their satiation levels (Milne & Boardman, 2000; Sorrell, 2007). However, the results of this research show a lower DRE in the Declaration of competing interest low-income group for energy services requiring electricity for their pro- vision in Paraguay. We find DREs between 14 % and 18 % in low-income The authors declare that they have no known competing financial households and between 23 % and 60 % in non-low-income households interests or personal relationships that could have appeared to influ- for residential electricity consumption in Paraguay. We identify two fac- ence the work reported in this paper. tors that may explain our results. First, electricity is not the principal en- ergy source for most low-income households. Second, most clandestine Acknowledgments electricity connections are from low-income households, leading to lim- ited reactions to price changes in this group. Moreover, considering the We acknowledge the editor and reviewers for providing insightful two points highlighted, the results of DREs for non-low-income house- comments on the manuscript. This research was funded by Becas Don holds are more comparable to other empirical DRE results than to our Carlos Antonio López (BECAL), grant agreement number 106/2017 results for low-income households. between the Treasury of Paraguay and Martín Adolfo Bordón Lesme. Interestingly, Paraguay's electricity supply provided by Itaipu and We also acknowledge support from project RTI2018-095484-B-I00 Yacyreta surpasses its electricity demand because Paraguay is one of (Spanish Ministry of Economy and Competitiveness and ERDF). Jaume the largest exporters of hydropower electricity in the world (Blanco Freire-González received support from the Beatriu de Pinós et al., 2017). The discrepancy between the supply of electricity and its postdoctoral program of the Government of Catalonia's Secretariat for consumption, especially in low-income households, can be explained Universities and Research of the Ministry of Economy and Knowledge. Appendix 1. Paraguayan electricity prices for households, category 142 Number Monthly range of consumption Price Unit of measure 1 0–50 kWh 311.55 G/kWh 2 51–150 kWh 349.89 G/kWh (continued on next page) 437 M. Bordón-Lesme, J. Freire-González and E. Padilla Rosa Energy for Sustainable Development 70 (2022) 430–441 (continued) Number Monthly range of consumption Price Unit of measure 3 151–300 kWh 365.45 G/kWh 4 301–500 kWh 403.82 G/kWh 5 501–1000 kWh 420.27 G/kWh 6 >1000 kWh 435.51 G/kWh Source: Personal elaboration with data from https://www.ande.gov.py/docs/tarifas/PLIEGO21.pdf. Appendix 2. Paraguayan electricity prices for households, category 141 Number Monthly range of consumption Price Unit of measure 1 0–50 kWh 311.55 G/kWh 2 51–150 kWh 349.89 G/kWh 3 151–300 kWh 365.45 G/kWh Source: Personal elaboration with data from https://www.ande.gov.py/docs/tarifas/PLIEGO21.pdf. Number Monthly range of social tariff consumption Discount Rate. Law N° 3480/2008 1 0–100 kWh 75 % 2 101–200 kWh 50 % 3 201–300 kWh 25 % Source: Personal elaboration with data from https://www.ande.gov.py/docs/tarifas/PLIEGO21.pdf. Appendix 3. Calculation methods for the climatic variables Condition Heating degree days formula Tmin > Tbase HDD = 0 (Tmax + Tmin) / 2 > Tbase HDD = (Tbase − Tmin) / 4 Tmax ≥Tbase HDD = (Tbase − Tmin) / 2 − (Tmax − Tbase) / 4 Tmax < Tbase HDD = Tbase − (Tmax + Tmin) / 2 Condition Cooling degree days formula Tmax < Tbase CDD = 0 (Tmax + Tmin) / 2 < Tbase CDD = (Tmax − Tbase) / 4 Tmin ≤Tbase CDD = (Tmax − Tbase) / 2 − (Tbase − Tmin) / 4 Tmin > Tbase CDD = (Tmax + Tmin) / 2 − Tbase Source: (Calculating Degree Days, n.d.). Appendix 4. Share of energy sources' input (KTOE) per energy services in Paraguayan high-income households of 2011 Source: Personal elaboration with BNEU data. 438 M. Bordón-Lesme, J. Freire-González and E. Padilla Rosa Energy for Sustainable Development 70 (2022) 430–441 Appendix 5. Share of energy sources' input (KTOE) per energy services in Paraguayan middle-income households of 2011 Source: Personal elaboration with BNEU data. Appendix 6. Data on energy sources' input per income level of 2011 Energy source Energy service Income level Data (TOE) Electricity Space cooling High income 10,375 Charcoal Cooking High income 8262 Electricity Water heating High income 7970 Electricity Other appliances High income 5861 Electricity Food preservation High income 5574 LPG Cooking High income 4349 Electricity Cooking High income 2716 Electricity Lighting High income 2031 Charcoal Water heating High income 507 Electricity Space heating High income 387 LPG Water heating High income 188 Electricity Pumping water High income 127 Firewood Cooking High income 120 Firewood Water heating High income 41 Firewood Space heating High income 13 Charcoal Space heating High income 3 Charcoal Cooking Middle income 34,776 Electricity Water heating Middle income 15,573 LPG Cooking Middle income 14,149 Electricity Food preservation Middle income 12,850 Electricity Other appliances Middle income 11,732 Electricity Space cooling Middle income 11,180 Firewood Cooking Middle income 8188 Electricity Lighting Middle income 4861 Charcoal Water heating Middle income 4828 Electricity Cooking Middle income 3878 Firewood Water heating Middle income 720 LPG Water heating Middle income 709 Biomass waste Cooking Middle income 705 Biomass waste Water heating Middle income 205 Electricity Space heating Middle income 174 Charcoal Space heating Middle income 58 Kerosene Cooking Middle income 14 Biomass waste Space heating Middle income 13 LPG Space heating Middle income 3 Charcoal Cooking Low income 52,915 Firewood Cooking Low income 49,211 Charcoal Water heating Low income 17,248 LPG Cooking Low income 11,401 Electricity Food preservation Low income 10,299 Firewood Water heating Low income 6964 Electricity Other appliances Low income 5859 (continued on next page) 439 M. Bordón-Lesme, J. Freire-González and E. Padilla Rosa Energy for Sustainable Development 70 (2022) 430–441 (continued) Energy source Energy service Income level Data (TOE) Electricity Water heating Low income 3537 Electricity Lighting Low income 3512 Electricity Space cooling Low income 3448 Biomass waste Cooking Low income 1777 Firewood Space heating Low income 898 Biomass waste Water heating Low income 817 Electricity Cooking Low income 811 LPG Water heating Low income 351 Charcoal Space heating Low income 122 Kerosene Lighting Low income 81 Electricity Pumping water Low income 75 Petrol Other appliances Low income 55 Electricity Space heating Low income 54 Petrol Pumping water Low income 22 Biomass waste Space heating Low income 10 Charcoal Other appliances Low income 6 Source: Personal elaboration with BNEU data. Appendix 7. Robustness checks Dependent variable: ln Municipality Fixed effects Province fixed effects (Eit/hhit) Non-low-income households Low-income households Non-low-income households Low-income households Low-income households (Model 1) (Model 2) (Model 3) (Model 4) (Model 4.1) α Coef. 3.9259*** 1.7465*** 3.6763*** 2.0249** 2.5176*** Std.Err (0.1629) (0.2210) (0.4001) (0.7016) (0.2899) (ln (PE it − 1)) Coef. −0.4492*** −0.3084*** −0.4220*** −0.3271*** −0.3017*** Std.Err (0.0089) (0.0105) (0.0233) (0.0337) (0.0302) lnPLPG it Coef. Without Without Without Without Without Std.Err lnCDDit Coef. 0.1877*** 0.1001*** 0.1944*** 0.0964 Without Std.Err (0.0162) (0.0283) (0.0397) (0.0936) lnHDDit Coef. Without Std.Err lnYit Coef. Without Without Without Without Std.Err (ln(Eit−1/hhit−1)) Coef. 0.6737*** 0.8570*** 0.6884*** 0.8343*** 0.8472*** Std.Err (0.0091) (0.0093) (0.0249) (0.0309) (0.0261) Periods 16 16 16 16 16 Cross-sections 189 187 16 16 16 Observations 2517 2479 210 210 256 Panel Unbalanced Unbalanced Unbalanced Unbalanced Balanced Weighted Statistics R2 0.9842 0.9230 0.9892 0.9264 0.9280 Prob (F-Statistic) 0.0000 0.0000 0.0000 0.0000 0.0000 Durbin–Watson 2.1357 1.8623 2.1376 1.7809 1.7723 Stat Unweighted Statistics R2 0.9579 0.8479 0.9867 0.8577 0.8674 Durbin–Watson 2.3057 2.2009 2.1869 2.1944 2.0791 Stat Note: *p < 0.05; **p < 0.01; ***p < 0.001. References Blanco, G., Amarilla, R., Martinez, A., Llamosas, C., & Oxilia, V. (2017). Energy transitions and emerging economies: A multi-criteria analysis of policy options for hydropower surplus utilization in Paraguay. Energy Policy, 108, 312–321. ANDE – Administración Nacional de Electricidad Ande (,). fromhttps://www.ande.gov.py/ BNEU Consolidado 11-02-14.pdf (.). fromhttps://www.ssme.gov.py/vmme/pdf/BNEU/ tarifas_vigentes.php (Accessed 15 June 2020). BNEU%20Consolidado%2011-02-14.pdf (Accessed 15 June 2020). Agostini, C., Plottier, M. C., & Saavedra, E. H. (2012). La demanda residencial por energía Bordón-Lesme, M., Freire-González, J., & Padilla, E. (2022). Do household energy services eléctrica en Chile. Economía Chilena, 15(3), 64–83. affect each other directly? The direct rebound effect of household electricity con- Alvi, S., Mahmood, Z., & Nawaz, S. M. N. (2018). Dilemma of direct rebound effect and cli- sumption in Spain. Energy Efficiency. https://doi.org/10.1007/s12053-022-10048-7 mate change on residential electricity consumption in Pakistan. Energy Reports, 4, In preparation. 323–327. Casarin, A. A., & Delfino, M. E. (2011). Price freezes, durables, and residential electric- Barnes, D. F., Samad, H., & Banerjee, S. G. (2014). The development impact of energy ac- ity demand. Evidence from Greater Buenos Aires. Energy Economics, 33(5), cess. Energy Poverty: Global Challenges and Local Solutions, 54. 859–869. BCP (.). Banco Central del Paraguay. fromhttps://www.bcp.gov.py/ (Accessed 15 June Chen, Q., Zha, D., Wang, L., & Yang, G. (2021). The direct CO2 rebound effect in house- 2020). holds: Evidence from China’s provinces. Renewable and Sustainable Energy Reviews, Belaïd, F., Bakaloglou, S., & Roubaud, D. (2018). Direct rebound effect of residential gas de- 155, Article 111888. https://doi.org/10.1016/j.rser.2021.111888. mand: Empirical evidence from France. Energy Policy, 115, 23–31. Chitnis, M., & Sorrell, S. (2015). Living up to expectations: estimating direct and indirect Bendezú, L., & Gallardo, J. (2006). Econometric analysis of household electricity demand in rebound effects for UK households. Energy Economics, 52, S100–S116. Peru. Peru: National Survey of Rural Household Energy. Chitnis, M., Sorrell, S., Druckman, A., Firth, S. K., & Jackson, T. (2013). Turning lights into Berkhout, P. H., Muskens, J. C., & Velthuijsen, J. W. (2000). Defining the rebound effect. flights: Estimating direct and indirect rebound effects for UK households. Energy Energy Policy, 28(6–7), 425–432. Policy, 55, 234–250. https://doi.org/10.1016/j.enpol.2012.12.008. 440 M. Bordón-Lesme, J. Freire-González and E. Padilla Rosa Energy for Sustainable Development 70 (2022) 430–441 DGEEC – Dirección General de Estadística, Encuestas y. Censos (.). fromhttps://www. Milne, G., & Boardman, B. (2000). Making cold homes warmer: The effect of energy effi- dgeec.gov.py/ (Accessed 15 June 2020). ciency improvements in low-income homes a report to the Energy Action Grants Dirección de Meteorología e Hidrología (.). Dirección de Meteorología e Hidrología. Agency charitable Trust. Energy Policy, 28(6–7), 411–424. https://www.meteorologia.gov.py/ (Accessed 15 June 2020). Orejuela, V., Arias, D., & Aguila, A. (2015). Response of residential electricity demand Druckman, A., Chitnis, M., Sorrell, S., & Jackson, T. (2011). Missing carbon reductions? Ex- against price signals in Ecuador. Proceedings of the 2015 IEEE thirty fifth Central ploring rebound and backfire effects in UK households. Energy Policy, 39(6), American and Panama Convention. 35. (pp. 373–378). 3572–3581. RAP - Relevamiento de Asentamientos Precarios (2015). Área Metropolitana de Asunción. EPE – Empresa de Pesquisa Energética (.). https://www.epe.gov.br/pt (Accessed 15 June https://issuu.com/matilez/docs/informe-rap (Accessed 15 June 2020). 2020). Reddy, A. K., Annecke, W., Blok, K., Bloom, D., Boardman, B., Eberhard, A., & Ramakrishna, Florentín, N. (2021). fromhttp://www.senado.gov.py/index.php/noticias/noticias- J. (2000). Energy and social issues. World Energy Assessment, 39–60. presidencia/8979-senadores-aprueban-con-modificaciones-proyecto-que-fomenta- Roy, J. (2000). The rebound effect: Some empirical evidence from India. Energy Policy, 28 la-utilizacion-de-energias-renovables-2021-11-25-22-36-08 (Accessed 13 March (6–7), 433–438. 2022). SIEN (.). https://www.ssme.gov.py/vmme/index.php?option=com_content&view= Freire-González, J. (2010). Empirical evidence of direct rebound effect in Catalonia. Energy article&id=1405 (Accessed 15 June 2020). Policy, 38(5), 2309–2314. Sorrell, S. (2007). The rebound effect: An assessment of the evidence for economy-wide en- Freire-González, J. (2017). Evidence of direct and indirect rebound effect in households in ergy savings from improved energy efficiency. UK Energy Research Centre London. EU-27 countries. Energy Policy, 102, 270–276. Sorrell, S. (2009). The evidence for direct rebound effects. Energy efficiency and sustainable González-Eguino, M. (2015). Energy poverty: An overview. Renewable and Sustainable consumption (pp. 23–46). Springer. Energy Reviews, 47, 377–385. Sorrell, S. (2018). Energy sufficiency and rebound effects Concept paper. ResearchGate. Greening, L. A., Greene, D. L., & Difiglio, C. (2000). Energy efficiency and consumption— https://doi.org/10.13140/RG.2.2.35846.22088. The rebound effect—A survey. Energy Policy, 28(6–7), 389–401. Sorrell, S., & Dimitropoulos, J. (2007). UKERC review of evidence for the rebound effect: Tech- Halff, A., Sovacool, B. K., & Rozhon, J. (2014). Energy poverty: Global challenges and local so- nical Report 2: Econometric studies. lutions. OUP Oxford. Sovacool, B. K. (2014). Defining, measuring, and tackling energy poverty. Energy Poverty: Hunt, L. C., & Ryan, D. L. (2014). Catching on the rebound: Why price elasticities are generally Global Challenges and Local Solutions, 2, 21–53. inappropriate measures of rebound effects. Surrey Energy Economics Centre (SEEC), UNCTAD (2020). Classifications. UNCTAD E-handbook of statistics. https://stats.unctad. School of Economics, University of Surrey. org/handbook/Annexes/Classifications.html (Accessed 15 June 2020). IEA – International Energy Agency (.). IEA. https://www.iea.org (Accessed 15 June 2020). van den Bergh, J. C. (2011). Energy conservation more effective with rebound policy. Labidi, E., & Abdessalem, T. (2018). An econometric analysis of the household direct re- Environmental and Resource Economics, 48(1), 43–58. bound effects for electricity consumption in Tunisia. Energy Strategy Reviews, 19, Van der Kroon, B., Brouwer, R., & Van Beukering, P. J. (2013). The energy ladder: Theoret- 7–18. ical myth or empirical truth? Results from a meta-analysis. Renewable and Sustainable Laureiro, P. (2018). Determinantes del consumo de energía eléctrica del sector residencial en Energy Reviews, 20, 504–513. Uruguay. Instituto de Economia-IECON. Villareal, M. J. C., & Moreira, J. M. L. (2016). Household consumption of electricity in Brazil Liddle, B., & Huntington, H. (2020). How prices, income, and weather shape household between 1985 and 2013. Energy Policy, 96, 251–259. electricity demand in high-income and middle-income countries. Energy Economics, Wang, Z., Han, B., & Lu, M. (2016). Measurement of energy rebound effect in households: 95, Article 104995. https://doi.org/10.1016/j.eneco.2020.104995. Evidence from residential electricity consumption in Beijing, China. Renewable and Llamosas, C., Giordanego, H., González, A., Félix, F., Oxilia, V., & Blanco, G. (2018). Energy Sustainable Energy Reviews, 58, 852–861. poverty in Paraguay: Perspectives and conceptualization—Case study of the Carmen Wang, Z., Lu, M., & Wang, J. -C. (2014). Direct rebound effect on urban residential electric- Soler Community (Limpio, Paraguay). https://hd-ca.org/publication_type/hdca- ity use: An empirical study in China. Renewable and Sustainable Energy Reviews, 30, conference-paper?publication_year=2018&post_types=publications (Accessed 15 124–132. June 2020). Westley, G. D. (1984). Electricity demand in a developing country. The Review of Masera, O. R., Saatkamp, B. D., & Kammen, D. M. (2000). From linear fuel switching to Economics and Statistics, 459–467. multiple cooking strategies: A critique and alternative to the energy ladder model. Zhang, Y. -J., & Peng, H. -R. (2017). Exploring the direct rebound effect of residential elec- World Development, 28(12), 2083–2103. tricity consumption: an empirical study in China. Applied Energy, 196, 132–141. 441