How Valuable is the Reliability of Residential Electricity Supply in Low-Income Countries? Evidence from Nepal.

• Author: Alberini, Anna

1. INTRODUCTION

Access to reliable electricity services is essential for poverty reduction and economic growth (World Bank, 2017). However, many developing countries, particularly low-income countries in South Asia and Sub-Saharan Africa, face severe electricity shortages, leading to frequent power shedding. (1) Underpricing of public electricity services combined with high technical and commercial losses in these countries limits the utilities' ability to recover capital and operational expenditures and affects the reliability of the power supply (Blimpo and Cosgove-Davies, 2019; Zhang 2019). Valid estimates of the willingness to pay (WTP) for reliable power supply are thus critical for both power system planning decisions and regulatory policies aimed at improving the quality of the electricity services. This is especially important for the residential sector, where low energy consumption makes electricity cost recovery a challenging problem (Trimble et al., 2016; Blimpo and Cosgove-Davies, 2019; Lee et al., 2020).

This paper estimates the value of lost residential electricity service in Nepal, a low-income country that has experienced chronic load shedding in the last decade. The load-shedding crisis over the period 2007-2016 has imposed high economic costs on Nepal's economy. (2) At the end of 2016, however, the daily load shedding of residential customers ended due to improved electricity dispatch, increased electricity production, and imports from India, though households may still experience some unscheduled outages (World Bank, 2019).

Assessing the value that households place on a reliable power supply is a difficult problem, especially in developing countries. Technical constraints and ethical considerations make it very difficult to randomly assign load-shedding schedules and prices, preventing experimental inference of the WTP increasingly common in other infrastructure settings (Berry et al., 2020; Do and Jacoby 2020; Grimm et al., 2020).

We use contingent valuation to elicit the willingness to pay of residential customers in Nepal for improved power supply. This allows us to get around the issue that it is not possible for us to structurally estimate properly specified residential electricity demand functions and infer the value of lost electricity service from them.

Stated preference methods have been previously used in developing countries to estimate the value of access to the electricity grid and/or a more reliable electricity supply (Kateregga, 2009; Abdullah and Mariel, 2010; Twerefou, 2014; Oseni, 2017, Blankenship et al., 2019, Deutschmann et al., 2021). Our unique empirical setting, the specific valuation scenario, and the econometric estimation techniques advance the existing literature in several ways. First, as Deutschmann et al. (2021), and unlike other developing country studies (e.g., Abdullah and Mariel, 2010; Bose et al., 2006; Twerefou, 2014; Oseni, 2017; Blankenship et al., 2019), we rely on a nationally representative survey of Nepali households, thus avoiding WTP sample selection bias or issues of limited regional coverage.

Second, our experimental design exploits the fact that the survey was done shortly (i.e., less than a year) after the residential load shedding had been eliminated. The respondents were asked to indicate their willingness to pay to avoid the number of days with outages they had experienced before the termination of the load shedding schedule of October 2016. This takes advantage of the respondents' actual experience with improved reliability of power supply, and by construction avoids unfamiliarity with the good to be valued (Mitchell and Carson, 1989), a problem frequently faced in similar stated preference studies in developed economies (see, e.g., Layton and Moeltner, 2005; Carlsson and Martinsson, 2007, 2008; Hensher et al., 2014; Ozbafli and Jenkins, 2016) and when the valuation task requires some technical knowledge (Broberg et al., 2021). (3) Finally, using supplemental data on actual outages at the transformer substation level, we can validate the quality of respondents' recall--an issue in contingent valuation studies, for which typically there is no easy solution (Hanemann, 1994).

Our analysis starts with calculating the WTP per kWh lost (i.e., the Value of Lost Load, VoLL) given assumptions about the load or exact information about the kWh used in a typical day. We then calculate the WTP per outage-day avoided and analyze its key drivers. Finally, we assess the internal validity of our estimates by regressing the WTP on the number of outage-days reported by the respondents, controlling for a variety of households' characteristics. To our knowledge, our study is the first contingent valuation study to address measurement error in the good to be valued, thus avoiding biased inference, by instrumenting for it. Specifically, we instrument for the number of outage-days using the frequency of all types of outages at the substation level.

The results from our study are striking for a number of reasons. First, unlike studies of other low-income economies, particularly the Sub-Saharan Africa region (Blimpo and Cosgrove-Davies 2019, Lee et al., 2020), we find that households, on average, attach economically significant value to a reliable power supply. The average WTP is about 123 NR ($1.11) per month, or 65% of the actual average monthly bill at the time of the survey, even though about 26% of the households are not willing to pay anything at all, and even though respondents are likely understating their WTP. When we convert the WTP to a VoLL (i.e., the WTP per kWh lost), our preferred estimates are in the range of 5 to 15 NR/kWh (Cents4.7 to Cents14/kWh), and thus bracket the average price per kWh from the grid paid by the respondents at the time of the survey. We also find that the WTP increases significantly--but slowly--with income and education levels. This result is consistent with the recent finding that households value the consumption of subsistence goods more than the availability of electricity services when they are poor (Sievert and Steinbuks, 2020).

Second, quite surprisingly, our average VoLL estimates do not seem to be any larger than those from a survey conducted more than a decade ago, when the load-shedding crisis started (Karki et al., 2010), even though the country's GDP per capita has grown by 42% since. Third, for the sample as a whole, the VoLL is higher for service lost in unscheduled outages than for service lost during scheduled load shedding. This finding is consistent with evidence from developed economies (Carlsson and Martinsson, 2007). But when we restrict the analysis to the "attentive" respondents--namely those who appear to have recollected exactly the number of outages in the month a year before the time of the survey--the VoLL is identical for unscheduled and scheduled lost electricity consumption.

Fourth, households that use rechargeable batteries (i.e., inverters) or solar equipment as backup equipment report systematically lower WTP. This result is in sharp contrast with earlier studies (Oseni, 2017), where households that own diesel generators with high running costs reported a higher WTP for reliable power supply. However, when we adjust the VoLL of those with rechargeable batteries and solar equipment to the VoLL implicit in the purchase of such equipment, we obtain higher estimates ranging from 9 to 22 NR/kWh. These results indicate that, in the absence of effective public policies, households internalize their WTP for reliable power supply by investing in power backup equipment.

Finally, although most households have an economically meaningful willingness to pay for a reliable power supply, our VoLL figures appear to be below the marginal cost of avoided load shedding (i.e., utilizing high-cost operating reserves or importing electricity at times of high demand). These findings suggest that if the government's goal is to improve the quality of residential electricity consumption, it must either lower the cost of generation, transmission, and distribution or resort to demand response--if technologically feasible and acceptable to the public. (4)

The remainder of this paper is organized as follows. Section 2 provides background information. Section 3 describes our data. Section 4 presents the methods and section 5 the results. Section 6 concludes.

(2.) BACKGROUND

1. Nepal and Its Electricity Supply

   Nepal, a landlocked country in South Asia, is one of the poorest countries in the world. As of 2018, its GDP per capita was USD 1039 (current USD), placing the country squarely in the bottom quintile worldwide. (5) In part because of the country's low level of development, and in part because of the disruption of the electricity supply (which we describe below), electricity demand in Nepal is much less than that of its neighboring countries. Per capita electricity consumption was 146 kWh in 2014 and approximately 245 kWh in 2018. For comparison, India's was 1,181 kWh per capita and China's 4,973 kWh per capita in 2018.

   As of December 2020, Nepal has about 1,303 MW of power generation capacity for its almost 30 million population, 97% of which is hydropower. (6) Almost all hydropower plants are of the run-of-river type, and up to two-thirds of this capacity is not available during the dry season (November-June). As a result, Nepal suffered a severe shortage of electricity supply during the decade from 2007 to 2016, which forced the national power company--the Nepal Electricity Authority (NEA) (7) --to implement extensive load shedding (World Bank, 2019). (8)

   Figure 1 plots the hours with no electricity (as per the scheduled outages) each month from January to December 2016 -- one of the worst load-shedding years--based on the schedules publicly announced by the NEA. (9) In the early months of the year, people were without power for about half of the time. Load shedding occurred every single day between...