2. Literature review

In this paper, we are interested in two interrelated types of literacy, which we argue are important in determining whether households are able to correctly value the benefits from the adoption of energy-efficient durables (such as appliances). The first is the idea of financial literacy, or computational ability in general, which would facilitate the performing of an investment analysis. The effects of higher levels of financial literacy have been extensively studied for both developed and developing countries, with regard to retirement decisions (such as pensions) as well as the adoption of other financial products and services (Lusardi and Mitchell, Reference Lusardi and Mitchell2007; Lusardi et al., Reference Lusardi, Mitchell and Curto2009; Lusardi and Mitchell, Reference Lusardi and Mitchell2011, Reference Lusardi and Mitchell2014; Xu and Zia, Reference Xu and Zia2012; Karakurum-Ozdemir et al., Reference Karakurum-Ozdemir, Kokkizil and Uysal2018). This literature finds an important role for financial literacy programs in determining financial knowledge-related outcomes, but also in determining financial behavior. For instance, education through financial literacy programs has been found to have a profound effect on the likelihood of opening savings accounts (and the take-up of financial products in general), but also on financial proficiency scores and knowledge (Xu and Zia, Reference Xu and Zia2012; Bruhn et al., Reference Bruhn, Lara Ibarra and McKenzie2014). However, except for Blasch et al. (Reference Blasch, Boogen, Daminato and Filippini2018), few studies have considered whether this measure may be correlated with the adoption of energy-efficient technologies (or durables, in general).

Another aspect of literacy that may play a role in facilitating the ability of households to perform an investment analysis, which may then influence their adoption of energy-efficient durables, is their knowledge of energy-related matters, such as whether they are cognizant of the operating costs of appliances, or of possible savings from the use of energy-efficient technologies. This has been understudied in the literature, in comparison to the determinants of financial literacy, even in developed countries. Previous papers have studied the effects of environmental attitudes and behavior (Ameli and Brandt, Reference Ameli and Brandt2015), as well as prices (Jacobsen, Reference Jacobsen2015), on energy-efficient adoption decisions, but less is known about the role of households' knowledge in determining these choices. Studies such as Brounen et al. (Reference Brounen, Kok and Quigley2013) and Kalmi et al. (Reference Kalmi, Trotta and Kazukauskas2017) have found that households suffer from low levels of energy literacy (or energy knowledge) in the Netherlands and Finland respectively. One study that emphasized the importance of knowledge is Mills and Schleich (Reference Mills and Schleich2012), who find that knowledge about energy consumption and energy-efficient technology options has a strong correlation with the household-level adoption of energy conservation practices, but not necessarily with the adoption of energy-efficient appliances. In a recent study, Prest (Reference Prest2020) finds that the most critical determinant of electricity demand for Irish households in response to a policy of peak-load pricing was whether consumers were aware of the policy.

Other studies have looked at the link between being able to compute the lifetime costs of appliances and a household's decision to purchase energy-efficient appliances. Blasch et al. (Reference Blasch, Boogen, Daminato and Filippini2018) evaluate the determinants of energy-related financial literacy, and also show that higher levels of energy-related financial literacy are correlated with the share of energy-efficient technologies (in this case, light bulbs) adopted by households. They use data from a sample of European countries (Italy, the Netherlands and Switzerland) for their analysis, and propose the concept of ‘energy-related financial literacy’, which encompasses both energy-related knowledge and cognitive skills, to process available information in order to take informed energy-related investment decisions.

A stream of literature based on developed countries such as Switzerland has found that many households have poor levels of energy-related financial literacy, and that this is often associated with households taking suboptimal decisions regarding identification of the appliances having the lowest lifetime costs (Blasch et al., Reference Blasch, Boogen, Filippini and Kumar2017a,Reference Blasch, Filippini, Kumar and Martinez-Cruzb, Reference Blasch, Filippini and Kumar2019). These studies have also found a largely positive impact of enhancing both a household's energy knowledge and their ability to do financial calculations, on their ability to identify the appliances having the lowest lifetime costs. Blasch et al. (Reference Blasch, Boogen, Filippini and Kumar2017a) show that more literate households are more likely to consume less electricity, and Blasch et al. (Reference Blasch, Filippini, Kumar and Martinez-Cruz2017b, Reference Blasch, Filippini and Kumar2019) find that households which have higher levels of literacy are more likely to calculate investment costs for themselves, rather than relying on simple rule-of-thumb heuristics.

Our paper is methodologically similar to that of Blasch et al. (Reference Blasch, Boogen, Daminato and Filippini2018), in that we are also seeking to understand the determinants of energy-related financial literacy, but in an LMIC context.Footnote ³ For developing countries, there is neither much evidence on the barriers to greater adoption of energy-efficient technologies in general, nor evidence on the determinants of energy-related financial literacy or on the ability of households to compute lifetime costs. We find this to be a big gap in the literature, especially because these countries suffer from already low levels of education and literacy (Klapper et al., Reference Klapper, Lusardi and Van Oudheusden2015). Moreover, as their populations and incomes increase, LMIC households are more likely to purchase appliances and vehicles, which is likely to influence energy consumption pathways in the coming decades (Wolfram et al., Reference Wolfram, Shelef and Gertler2012).

This paper attempts to fill this gap at least partially, by: (a) providing a descriptive analysis of appliance stock, financial literacy and energy-related financial literacy, as well as replacement attitudes for our sample; and (b) analyzing the determinants of energy-related financial literacy for a sample of urban households in the eastern lowlands of Nepal. While we do not have information on electricity consumption or other types of behavior, such as actual adoption of an energy-efficient appliance, which may be interpreted as a direct outcome of having a high level of energy-related financial literacy, we also seek to: (c) evaluate the association between higher levels of literacy and attitudes of households regarding replacement of old and inefficient appliances. The results of this study provide important findings that enable us to undertake future experiments to study the effects of enhancing literacy in influencing households' adoption decisions.

3. Data

The data is drawn from a household-level survey conducted by the Centre for Energy Policy and Economics (CEPE), ETH Zurich, in collaboration with the Nepal Electricity Authority (NEA).Footnote ⁴ The survey was administered during the first quarter of 2018. A total of 2,042 respondents and households were part of the survey that was conducted in some of the urban municipalities in the Morang and Sunsari districts of Nepal, e.g., Urlabari, Itahari, Biratnagar, Duhabi and Dharan. These municipalities belong to the eastern Terai region (lowlands) of Nepal.Footnote ⁵ The survey was paper-based and in the local language (Nepali) and consisted of 31 questions spread over 6 pages. At the beginning of the survey, there was an introductory page with project information, incentives for participation and a few basic instructions. The survey was administered by the local distribution offices of the NEA with the help of their on-field meter-readers – that is, the utility's staff who are responsible for physically visiting the addresses having electricity meters, checking the meter readings and handing out the electricity bills for the last month (or last quarter if it is a remote location).Footnote ⁶ Further details on the project and the household survey can be found in the report, ‘A descriptive overview of literacy, attitudes and behaviorus towards energy consumption in Nepal’ (Kumar, Reference Kumar2018).

The questions included in the survey were designed to collect information on households' sociodemographic characteristics, dwelling-related attributes, capital stock, energy-related knowledge, computational skills, attitudes and behavior towards electricity consumption and adoption of energy-efficient appliances. Below we present some descriptive statistics of the data.

3.1. Descriptive statistics

3.1.1. Sociodemographic and dwelling characteristics

Table 1 shows the percentage distribution of the basic sociodemographic characteristics of the respondents and households. Both genders are almost equally represented in the data. The typical respondent is married, lives in a joint family, both the respondent and the spouse are educated at the primary level (below high school), and the monthly household income is less than Rs.20,000.Footnote ⁷ One in every three households declares that they regularly receive foreign remittance from a family member who works abroad.Footnote ⁸ The median age of respondents in the household survey is found to be 43 years. The most frequent response with regard to the total number of people regularly living in the household is 4.

Table 1.

Basic sociodemographic attributes of the survey sample

^# Sample excluding missing values from a total of 2,042 observations.

Table A1 in the online appendix presents some information on the features of the dwelling. Most respondents live in a single family house (97.5 per cent) and in most cases the house is owned by a member of the household (95.2 per cent). Almost half of the households belong to the town of Urlabari, the second largest city (after Biratnagar) in the Morang District in the lowlands of Nepal. The most frequent response on the total number of bedrooms and living rooms within the dwelling is 4.

Tables A2 and A3 in the online appendix provide these summary statistics for the regression sample of table 6 of our results.

3.1.2. Capital stock

The survey asked respondents if they owned certain types of household appliances, a motorcycle, or an inverter-battery system (or backup-system), and whether or not they were planning to buy or replace these durables in the next two years. Table 2 shows the share of responses to this question as percentages. We find that the share of households owning an electric rice cooker is more than 60 per cent, followed by fridge (58.4 per cent), motorcycle (39.5 per cent), an inverter-battery based electricity backup system (21.4 per cent), washing machine (8.9 per cent), electric geyser (8.5 per cent) and air-conditioner (4 per cent). The share of households planning to add or replace one of these appliances in the next two years varies from 2–5 per cent.

Table 2.

Capital stock and plan to add or replace in the next two years

^# Sample excluding missing values from a total of 2,042 observations.

3.2. Definition of literacy scores

Measurement of the level of energy-related knowledge and computational skills was one of the main objectives of the survey which consisted of several questions for this purpose. Below we present an overview of these questions and use the responses to construct three types of scores, following the approach in Blasch et al. (Reference Blasch, Boogen, Daminato and Filippini2018). Our first score is an index that measures energy knowledge, and the second is a score of that we term ‘computational skills’, which is a combination of financial literacy along with the ability to do a lifetime cost calculation. The final measure has been referred to as energy-related financial literacy, which is defined as the sum of the energy knowledge as well as computational skills scores of the respondents.

3.2.1. Energy-related knowledge

Questions on energy-related knowledge checked whether respondents knew the cost of electricity (in monetary terms) for some of the daily electricity consuming activities, or energy services, at home, and whether they were able to compare two energy services and determine which one consumed more electricity. The respondents were also asked how much they thought the percentage of energy savings was when using an LED bulb as opposed to an incandescent bulb.Footnote ⁹ These questions are described in table A4 in the online appendix; summary statistics on the correct responses are presented in table 3.

Table 3.

Energy-related knowledge

Notes: All variables are binary with 1 implying a correct response.

The mean values in table 3 are on the lower side (in an absolute sense). In general, this highlights a low level of energy-related knowledge among the Nepalese respondents in our sample. Among all questions, respondents seem to be better informed about comparison and savings on LED bulbs compared to regular incandescent bulbs. This is likely due to the fact that the NEA has been actively promoting the benefits of LED bulbs through advertisements and information campaigns since 2016 (Nepal Electricity Authority, 2017).Footnote ¹⁰

The index capturing energy-related knowledge is created as the sum of the number of questions which the respondent answered correctly (theoretically, it can vary from 0 to 6, given that we ask 6 questions to measure the respondent's knowledge and energy-related awareness).Footnote ¹¹ In our sample, a plurality (almost 40 per cent) of households incorrectly answered (or did not know the answer to) every question related to energy consumption.

3.2.2. Computational skills

The questionnaire contained a few questions to assess the level of financial literacy and computational skills. Some of these questions pertain to calculations of simple interest, compound interest as well as to the effects of inflation, which are standard in the financial literacy literature (Lusardi and Mitchell, Reference Lusardi and Mitchell2014).Footnote ¹² Additionally, we asked the respondents a question that sought to identify their ability to perform an investment analysis by means of comparing two appliances. This question captures whether households can compare two fridges, and identify the one having a lower lifetime cost (defined as the sum of the purchasing cost, and the operating costs).Footnote ¹³ These questions are described in table A4 in the online appendix; the summary statistics on correct responses are presented in table 4.

Table 4.

Financial literacy and computational skills

Notes: All variables are binary with 1 implying a correct response.

Table 4 depicts a generally low level of financial skills. Nearly 30 per cent of respondents appear to be aware of how to calculate a compound interest rate. Understanding of the interplay of rate of inflation and bank interest rate appears to be lower. The poor performance in the simple interest question when compared to the compound interest question is somewhat strange. It might be due to the use of a larger monetary amount (Rs. 10,000) and a longer time duration (5 years) in the simple interest question, which, if true, again points to cognitive limitations of consumers in processing large numbers. On the other hand, this may also have happened due to the layout of the questionnaire, as this particular question started on page 4, and ended on page 5 in the paper survey.Footnote ¹⁴ Lastly, respondents perform very poorly on the question concerning identification of the appliance with the lower lifetime cost – only about 1 in 10 answer this correctly – which highlights the inability of consumers to perform investment analysis in the domain of appliance choice. We originally planned to conduct a randomized controlled trial using this question, by providing half of the respondents with a one-page educational brochure on how to calculate the lifetime cost of a durable before they answered this question. Unfortunately, because of logistical limitations, this could not be implemented. However, about 20 per cent of the respondents still ended up receiving the survey with the brochure attached to the end of the questionnaire, but this was not communicated to them. Later in the empirical section, we also account for this aspect.

Our score measuring computational skills is constructed by summing up the number of above described questions that the household answered correctly in the survey, i.e., it varies from 0 to 4, and considers both financial knowledge and ability to perform a simple lifetime cost calculation of an appliance. Thus, the index attempts to represent the computational ability of the respondents. While it is closely related to a typical measure of financial literacy, we prefer it over financial literacy because it is more comprehensive – it incorporates a critical component of the ability to perform an investment analysis, namely, the ability to calculate and compare lifetime costs of appliances. Nevertheless, for the empirical analysis, we also consider the financial literacy score that varies from 0 to 3 (i.e., without considering the lifetime cost comparison question).

3.2.3. Energy-related financial literacy

The total number of questions answered correctly by the respondent pertaining to both energy knowledge and computational skills are the two components of the index of energy-related financial literacy. Theoretically, the index can vary from 0 to 10 (there are six questions aimed at capturing energy knowledge and four questions for computational skills). In our data, the maximum score that households got is 8. Interestingly, 30 per cent of the households got a score of zero, meaning that they were unable to answer a single question correctly.

Following the arguments listed in Blasch et al. (Reference Blasch, Boogen, Daminato and Filippini2018), one way to justify combining these different components together in creating the energy-related financial literacy index, at least statistically speaking, is to check for: (a) the correlation between them, and (b) the internal consistency of the scales that have been used to measure the two components. The pairwise correlation coefficient between the two different indices that we have used to measure energy-related financial literacy, namely energy knowledge and computational skills, is found to be 27.24 per cent. The correlation coefficient between computational skills and energy knowledge is comparable to that of about 27 per cent for the Swiss, Italian and Dutch population (Blasch et al., Reference Blasch, Boogen, Daminato and Filippini2018). This suggests that there may be common determinants of the two factors, such as income and education, which are closely related to one another in the context of developing countries.

In addition, table A5 in the online appendix includes information on Cronbach's alpha measures which tests the internal consistency of the indices. Internal consistency refers to the strength of association between multiple items that are used to measure the same construct (often the components of a Likert Scale index, or a score). Higher levels of internal consistency (usually measured using Cronbach's alpha) indicate that the components of the index are closely related to one another, and thus invoke more reliability in the measure. From the results in table A5, we find that the measures of Cronbach's alpha are high for each combination of measures that we use. This suggests scale reliability as well as internal consistency of these measures when they are used as a group (Tavakol and Dennick, Reference Tavakol and Dennick2011).

Table 5 presents the summary statistics on the energy knowledge index, the index for computational skills, the energy-related financial literacy index, as well as the classical financial literacy index. Given the maximum score that was achievable, we see that the mean scores are very low for our sample, along both dimensions of energy knowledge and computational skills, and that no household attains scores of energy-related financial literacy which are higher than 8 (the theoretical maximum is 10). Moreover, the low median scores suggest that a large share of the sample fares poorly.

Table 5.

Summary statistics on the scores

4. Empirical model

In this section, we first present the empirical model pertaining to determinants of the different literacy scores. Next, we illustrate the empirical model to analyze the replacement attitudes. For the first model, we adopt an approach similar to that of Blasch et al. (Reference Blasch, Boogen, Daminato and Filippini2018), in using an ordered probit estimation to model energy knowledge, computational skills, and energy-related financial literacy (defined as the sum of the indices for energy knowledge, and computational skills). This is intuitive, given that the dependent variables for this set of estimations are scores, with a natural ordering (a higher score indicates a more ‘literate’ respondent). This methodology assumes that there is a continuous, latent measure underlying the observable ordinal variable, and that it can be expressed as

(1)\begin{equation} y^{*}_{i}= X_{i}\beta + \epsilon_{i}, \end{equation}

where $y^{*}_{i}$ is a latent (continuous) variable measuring the literacy score of the respondents, $X_{i}$ is a vector denoting the socioeconomic characteristics of household i such as age, income, level of education and dwelling-specific characteristics, and β is the vector of parameters that need to be estimated. $\epsilon _{i}$ denotes the stochastic error term, which is assumed to be independently and identically distributed across households. This model is estimated using maximum likelihood estimation.

The probability that household i has reached a literacy level of j (where j can vary from 0 to 10 in the case of the model estimating energy-related financial literacy), is given by

(2)\begin{align} &{\rm Pr}(y_{i} = j) = {\rm Pr} (k_{j-1} < y^{*}_{i} \leq k_{j}), \\ & \quad -\infty = k_{0} < k_{1} <\cdots k_{j}\cdots < k_{J} = \infty, \quad j \in 1, 2, 3, \ldots, J, \end{align}

where the $y_{i}$s are the ordinal values of the literacy score, and the $k_{j}$s are the threshold parameters. We estimate this model separately for each score that we are interested in, namely the energy knowledge of the households, their skills score, the energy-related financial literacy index; lastly, we also estimate the determinants of the financial literacy score.

In the second part, we evaluate the association between these literacy scores and the replacement attitudes among our sample of Nepalese households. While we do not observe their actual replacement behavior, the respondents are asked about their reasons for not having replaced their old and inefficient appliances. Recall that, based on the replacement attitudes, we were able to group our sample into three categories – ‘irrational’, ‘boundedly rational’, and ‘rational’. These groups exhibit an ordinal nature, i.e., ‘irrational’ < ‘boundedly rational’ < ‘rational’. We also use, in this case, an ordered probit methodology.

We first estimate the association between energy-related financial literacy and the level of rationality in replacement attitudes. This estimation can be represented by the expression for the ordered probit regression:

\begin{equation} r^{*}_{i}= E_{i}\alpha + X_{i}\beta + \nu_{i}, \end{equation}

where $r^{*}_{i}$ is a latent (continuous) variable measuring the replacement attitudes of the households, $E_{i}$ is the energy-related financial literacy score of household i, and $X_{i}$ denotes the same vector of socioeconomic characteristics of household i that we used for the previous estimation. $\nu _{i}$ denotes the stochastic error term, which is assumed to be independently and identically distributed across households.

The probability that household i has reached a level of rationality j in terms of replacement attitudes (where j can vary from 1 (irrational) to 3 (rational)), is given by

\begin{align} & {\rm Pr}(r_{i} = j) = {\rm Pr} (s_{j-1} < r^{*}_{i} \leq s_{j}) \\ & \quad -\infty = s_{0} < s_{1} < \cdots s_{j} \cdots < s_{J} = \infty, \quad j \in 1,2,3, \ldots, J, \end{align}

where the $r_{i}$s are the ordinal values of the replacement attitudes as defined above, and the $s_{j}$s are the threshold parameters.

We then also estimate this model in a specification using the components of the energy-related financial literacy score separately, i.e., introducing the energy knowledge index along with the computational skills measure as independent variables. Finally, we incorporate the financial literacy score as an independent variable in one specification.

In the estimation of equation (3), there is a concern about endogeneity which prevents us from causally identifying the effect of literacy on the level of rationality. For instance, it is not straightforward to rule out the possibility of correlated unobservables influencing both the dependent variable and our explanatory variables, such as having a bank account or membership in a co-operative (since engagement in financial services helps to accumulate computational skills).Footnote ¹⁷ In order to partially mitigate this concern, we control for important determinants of both attitudes and literacy (such as education and income, as well as the total number of investments already made by households in appliances and personal mobility).

It is also likely that credit constraints influence outcomes such as replacement attitudes, given that energy-efficient appliances are often relatively more expensive, and that households may not be able to afford them, irrespective of their levels of energy-related financial literacy. For our sample, we feel that this is relatively less likely to be a concern, because our focus in this study is on urban households that have access to electricity, and already own a few appliances. In our regressions, we attempt to control for the importance of credit and liquidity constraints by including socioeconomic controls for income, education, and stock of durables owned by households. Moreover, in response to the question on replacement attitudes regarding an old appliance, we find that only 5 per cent of respondents stated that high upfront costs prevented them from investing in a more energy-efficient appliance.