Policy Research Working Paper 10330 Energy Demand during a Pandemic Evidence from Ghana and Rwanda Justice Tei Mensah Aimable Nsabimana James Dzansi Alexandre Nshunguyinka Africa Region Office of the Chief Economist February 2023 Policy Research Working Paper 10330 Abstract The COVID-19 pandemic caused significant disruptions households and firms to the COVID-19 pandemic, and to economies around the world. In response to this, some the role of utility subsidies during the period. Findings developing countries offered reliefs such as electricity sub- from the paper indicate that the pandemic led to higher sidies while others did not. How did the pandemic affect consumption of electricity in both countries, albeit with the electricity consumption of households and firms? Did variations across countries and sectors. While residential the utility subsidies enable a quick recovery from the pan- consumption soared, consumption of non-residential cus- demic? And what are the distributional impacts of the utility tomers such as hotels and industries declined during the subsidies? This paper leverages unique administrative billing period. Electricity subsidies in Ghana during the pandemic data on electricity consumption from two African coun- explain the sharp increase in residential consumption. These tries, Ghana and Rwanda, with differing policy responses findings highlight the potential effects of pandemic relief to the pandemic to document the demand response of measures on household welfare. This paper is a product of the Office of the Chief Economist, Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at jmensah2@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Energy Demand during a Pandemic: Evidence from Ghana and Rwanda∗ Justice Tei Mensah, Aimable Nsabimana, James Dzansi, Alexandre Nshunguyinka Keywords: COVID-19, Electricity, Subsidies, Africa JEL Codes: Q41, L94, L97 ∗ Mensah: Office of the Chief Economist, Africa Region, The World Bank. Email: jmen- sah2@worldbank.org; Nsabimana: Rwanda Polytechnic. Email: aimeineza@gmail.com; Dzansi, Interna- tional Growth Center. Email: james.dzansi@theigc.org ; Nshunguyinka: Rwanda Utilities Regulatory Au- thority (RURA) anshunguyinka@gmail.com 1 Introduction The COVID-19 pandemic has caused severe disruptions to the global economy, with sig- nificant effects on livelihoods particularly in developing and emerging economies. To stem the spread of the virus and save lives, governments around the world imposed restrictions on movement of people. Notable among these restrictions were stay-at-home (SAH) or- ders or lockdowns and social distancing. As a consequence, many (non-essential) employ- ment activities also transitioned to home-based-work (HBW). For work environments, where home-based-work was not possible, workers often worked in shifts to reduce the capacity of people at the work place at any given point in time. In the developing world, where majority of people work in the informal sector, these restrictions and SAH orders were less strict due to the economic and logistic constraints. Nonetheless, overall, the restrictions and associated impact on the global economy have implications on energy demand. For instance, the transition to HBW is likely to increase residential energy con- sumption while decreasing non-residential consumption, especially given the important role of energy in the modern economy. At the same time, pandemic relief measures such as electricity subsidies implemented in several countries are likely to have intended and un- intended consequences on electricity consumption (Saltane et al., 2022; Berkouwer et al., 2022). In this paper, we examine the electricity demand response of households and firms to the COVID pandemic in Africa using unique administrative data on billing records from Ghana and Rwanda. The choice of these countries is motivated by two unique features: (i) Ghana has one of the highest electrification rates (86%) in Africa, while access rate in Rwanda remain relatively low (47%) but rising;1 and (ii) Ghana offered electricity sub- sidies to customers during the pandemic while Rwanda did not. Thus, data from these countries will provide comprehensive evidence of the effect of the pandemic and associ- ated policy measures on households and firms in the context of developing economies. Specifically, we seek to address two key questions: first, how did the pandemic affect elec- tricity consumption by households and firms in the developing world? Second, to what extent did pandemic relief measures such as electricity subsidies matter for mediating the impact of the pandemic on households? Causal identification of the impact of COVID on outcomes is beset with several chal- lenges. The main challenge relates to the universal nature of the pandemic and the wide reach of the lockdowns (SAH) in many countries. This limits the ability of researchers to 1 https://data.worldbank.org/indicator/EG.ELC.ACCS.ZS?locations=GH-RW 1 identify "credible" treated and control units for causal identification of impacts. In addi- tion, energy use patterns entail significant amount of seasonalities, hence, a simple before- after differences in consumption levels may not reflect the impact of the pandemic. To address these challenges, we leverage the detailed (panel) nature of our datasets and model electricity consumption at the customer level before and after the roll-out of the COVID lockdowns (SAH) compared against the typical (average) consumption lev- els during the same period in the previous years (Irwin et al., 2021). Essentially, for each customer, we use the typical electricity consumption levels in the previous (non- COVID) years as the counterfactual, i.e., the level of consumption in the absence of COVID. We include customer fixed effects and location-year fixed effects to purge customer spe- cific and temporal factors that influence consumption. In addition, we explicitly control for monthly average temperature and total precipitation to absorb the effect of seasonal weather patterns on electricity consumption. Six main results emerge from the paper. First, residential electricity consumption dur- ing the pandemic increased in both Ghana and Rwanda, albeit the rate of increase was higher in the former. Specifically, average household consumption increased by about 23% in Ghana compared to 4% in Rwanda. Second, consumption in the non-residential sector declined by about 11% in Rwanda, while rising marginally in Ghana by about 6% in Ghana. Third, industrial consumption declined in both countries, highlighting the effect of COVID-induced business closure and supply chain disruptions on the activities of the industrial sector. Fourth, the effect of COVID on electricity consumption was pronounced in lockdown and non-lockdown communities, as well as both rural and urban communi- ties. This highlights the effects of the pandemic across various segments of the popula- tion. Fifth, pandemic relief measures such as electricity subsidies administered in Ghana played a key role in raising consumption. While the subsidies were important to sustain- ing households’ consumption from falling below pre-pandemic levels, they also induced an unintended consequence: excess consumption among households. Finally, contrary to earlier criticisms that the design of the subsidy program in Ghana was regressive (Berk- ouwer et al., 2022), our back-of-the-envelope calculations suggest that the total subsidies to low-income (lifeline) customers were about 67% higher than total subsidies received by middle/high-income (non-lifeline) households (customers). This is largely driven by the extension of the subsidy for longer duration for low-income households. This paper contributes to two strands of the literature. First, the broader literature on the effects of COVID-19 on households and firms in emerging economies. These stud- ies range from the effects on food security(Amare et al., 2021; Abay et al., 2020), wel- 2 fare (Adams-Prassl et al., 2020; Schmidt et al., 2021), firms(Cong et al., 2021), employ- ment(Cortes and Forsythe, 2020; Albanesi and Kim, 2021; Klein, 2022), energy use(Cicala, 2020), water consumption(Irwin et al., 2021), pollution(Zhang et al., 2021), as well as the effectiveness of the government policies during this period (Gentilini et al., 2020). Ci- cala (2020) for instance shows that the COVID pandemic led to a substantial increase in electricity consumption by residential customers in the United States. Commercial and in- dustrial users on the other hand experienced a decline in consumption, with the transition to home-based work as the main reason for the differential impacts across the customer groups. Second, our paper contributes to the strand of literature on the (distributional) impact of energy subsidies on household and firm behavior (Komives et al., 2005; Davis, 2014; Younger et al., 2016; Basurto et al., 2020; Hahn and Metcalfe, 2021; Berkouwer et al., 2022). Perhaps, closest to our paper in this respect is Berkouwer et al. (2022) who assessed the distributional impact of the electricity subsidies in Ghana during the COVID period. We show that while the assessment by Berkouwer et al. (2022) that the (initial) design of the subsidies was (plausibly) regressive, our back-of-envelop calculations based on admin- istrative data on actual consumption suggest that the eventual extension of the subsidy period for low-income households made the program progressive than regressive. The rest of the paper is organized as follows. Section 2 presents an overview of the COVID situation in the two countries and the policy responses to the pandemic. Sections 3 and 4 describe, respectively, the data and empirical strategy for the paper. Results are presented in Section 5. Section 6 concludes the paper with a summary of findings. 2 Context This section provides a contextual overview of the COVID situation in the two countries and the policy measures implemented to contain the spread of the virus and ameliorate the socioeconomic impact of the pandemic. 2.1 Rwanda The first case of COVID-19 in Rwanda was recorded on March 14, 2020. To stem the spread of the virus, the government introduced a plethora of measures including, but not limited to, closure of schools and religious centers, restrictions on large gatherings, social distanc- ing, and urging employees to work from home. 3 Further, on March 21, 2020, a total lockdown or stay-at-home (SAH) order was im- posed across the country. During this period, movements of people were strictly prohib- ited except for essential services such as food and health care services. Industries operated at minimum capacity as employees were only allowed to work in shifts. To ameliorate the effects of the SAH order and other related restrictions on household welfare, the govern- ment provided free food and other basic essentials to vulnerable households. In addition, the government established an Economic Recovery Fund to support the recovery of busi- nesses hardest hit by the pandemic to enable them to survive, resume operations and safe- guard employment, thereby cushioning the economic effects of the pandemic. The fund was particularly targeted at hotels and operators in the tourism sector which were heavily hit, as global travel plummeted. Although the SAH was expected to last for two weeks, it was eventually extended until May 4, 2020. The measures were relaxed to some extent in May 2020, although movements of people still had some restrictions. For instance, evening curfews were imposed, thus preventing people from staying outside in the evenings, and movement across certain jurisdictions required permission from local authorities. However, some services including public and private businesses, operators in the manufacturing and construction sectors were allowed to re-open with only essential staff while non-essential staff continued to work from home. Markets were reopened with essential vendors not exceeding 50% capacity, while hotels and restaurants were only allowed to work at a limited scale. Since then, the country made regular assessments of the COVID cases and various targeted restrictions and preventive measures were applied depending on diagnostics by health officials. 2.2 Ghana Ghana recorded its first case of COVID-19 on March 12, 2020.2 Subsequently, authorities in the country introduced several preventive measures to halt the spread of the disease. These measures include; contact tracing, a ban on all public gatherings,3 closure of schools, closure of all land borders with neighboring countries, and shutting the main international airport in Accra. The Ministry of Health also heightened routine surveillance and carried out mass education campaigns to create awareness about the virus, and to promote social 2 As of September 22, 2022, there was a total of 168,813 confirmed cases, 167,206 recoveries, and 1,459 deaths. Tragic as these figures are, particularly for those who lost loved ones, the fatality rate in Ghana pales in comparison to the global average. Ghana owes the relatively low confirmed fatalities to a suite of measures that were instituted before the virus reached the shores of the country and social interventions which were implemented once the first two cases were confirmed 3 Such as conferences, workshops, funerals, festivals, political rallies, and religious activities 4 distancing and hand washing. Due to the constant increase of positive cases in the first month of COVID-19 in the country, a 3 week (partial) lockdown (SAH) was introduced in the two largest cities, Accra and Kumasi, considered the epicenters of the pandemic, between March 31 and April 21, 2020. During the lockdown period, the movement of people in the affected communities was restricted. Only the workers in essential services (healthcare, media, food restaurants, security agencies) were allowed to move around in these two cities. Outside the two cities, people were free to move around, although most institutions and places of worship were closed due to fear of contagion.4 Although the SAH/lockdown lasted only three weeks, the other restrictions continued. It was not until July 2020, that the country started easing the restrictions. Within weeks, after the implementation of the restrictions, it became apparent that the COVID-19 pandemic and measures taken to contain it were taking a toll on the economy and livelihoods. Across the private sector, businesses especially in the education, hospital- ity, manufacturing, and agriculture sectors came under severe strain. For example, schools were closed with the incomes of teachers in private educational institutions severely re- duced or threatened. In the hospitality sector, occupancy rates fell below 5 percent, with many hotels closed.5 Productivity in factories fell to historic lows due to mandatory so- cial distancing requirements, which meant only a fraction of employees could be engaged at any time. In the agriculture sector, there were labor supply problems as both internal movement of farm labor and labor coming in from neighboring countries have been cur- tailed. It is also becoming difficult for extension officers to visit farmers due to the fear of the virus. Lastly, agribusiness exporters have been hit with the loss of markets, and with some firms throwing away perishable products. To mitigate the economic effects of the pandemic and associated restrictions, the gov- ernment instituted several policy measures under the Coronavirus Alleviation Program (CAP)6 to cushion households and businesses from the economic costs of the pandemic and the measures taken to contain it. Prominent among these measures were subsidies for electricity and water.7 Provision of free food and other household essentials were also distributed to households in communities affected by the lockdown.8 The electricity subsidies are particularly important to this paper. The subsidy program 4 https://rstmh.org/news-blog/news/the-covid-19-situation-in-ghana#:~:text=In%20Ghana%2C% 20the%20first%20official,country%20from%20Norway%20and%20Turkey. 5 https://mofep.gov.gh/sites/default/files/news/care-program.pdf 6 https://mofep.gov.gh/sites/default/files/news/care-program.pdf 7 https://presidency.gov.gh/index.php/briefingroom/speeches/1560-president-akufo-addo-speaks-on-updates-to-ghan 8 https://citinewsroom.com/2020/07/government-justifies-extension-of-free-water-electricity-to-ghanaians/ 5 offered a 100% subsidy to lifeline (low-income) customers who consume 0-50 kWh of elec- tricity a month, while non-lifeline customers (consuming above 50 kWh a month) would enjoy a 50% subsidy (Berkouwer et al., 2022). The subsidies were announced on March 29 and took effect from April 1. Eligibility status was determined based on March consump- tion. The policy was initially announced to last for three months, however, the subsidy for [only] lifeline customers was eventually extended for 12 months. 3 Data This paper uses administrative data on electricity billing records from Ghana and Rwanda. The Ghana data comes from the Electricity Company of Ghana (ECG), which is the largest distributor in the country with operations in the southern and middle belts. It accounts for nearly 70% of all electricity customers in the country. We use data on billing records of the universe of electricity customers of the ECG from January 2018 to December 2020. The data identifies customer types based on the tariff applicable: residential (households), non-residential, and heavy industries. For each customer and year-month, it records the amount (kWh) of electricity consumed, the monetary value in Ghana Cedis (GHS), meter type (postpaid vs prepaid), and location (district) of the customer. In all, the data contains 42 million customer-year-month observations. The Rwanda data comes from the Energy Utility Corporation Limited (EUCL), the main distributor, via the Rwanda Utilities Regulatory Authority (RURA). The dataset con- tains the billing records of the universe of electricity customers in Rwanda from January 2018 to December 2020. The data identifies customer type based on the tariff applicable: residential, non-residential (commercial, hotels, health centers, and public works ( water storage and pump stations, broadcasters)), and small-and-medium industries. Also, all customers in the dataset use prepaid meters: Rwanda has a universal roll-out of prepaid meters, with large and heavy industries the only exception who are allowed to use post- paid meters. Our data exclude these customers (i.e. large and heavy industries). For each customer, we have monthly records on the amount (kWh) of electricity consumed (pur- chased), the monetary value in Rwandan Francs (RWF), location (community/district), and rural-urban status. In all, the data contains 21 million customer-year-month observa- tions. We complement the electricity data with monthly data on temperature and total pre- cipitation from the Copernicus Climate Change Service.9 9 https://cds.climate.copernicus.eu/cdsapp#!/search?type=dataset 6 4 Empirical Strategy Causal estimation of the effect of the COVID-19 pandemic on electricity consumption is beset with at least two challenges of identification. The first relates to the role of season- alities in energy demand. Temperature and rainfall patterns are key drivers of energy demand around the world, and given that these weather patterns vary based on the time of the year, monthly changes in electricity consumption to a large extent are influenced by the prevailing weather conditions. Thus, pre-post differences in household electricity con- sumption may reflect the seasonal patterns in electricity consumption and not necessarily the effect of the pandemic and its associated stay-at-home orders. Secondly, the global nature of the COVID-19 pandemic and the wide reach of the stay-at-home orders pose se- vere challenges to the identification of the counterfactual, as one cannot identify a relevant control group. To address these issues, we follow the approach of Irwin et al. (2021) and implement a difference-in-difference design by comparing the differences in electricity consumption during the post-COVID months and consumption levels in the same month in the previous years (2018-2019) with the average consumption in the months just before the pandemic (i.e. January and February 2020) and the years before (2018-2019). Essentially, the ap- proach uses the average monthly consumption of each household (firm) in the previous years as the counterfactual level of electricity consumption in the respective months in the absence of the COVID pandemic. The identification strategy is similar in spirit as Knutsen et al. (2017) and Isaksson and Kotsadam (2018).10 To express this formally, consider the following regression specification: n lnYimt = αi + λjt + ω β2020 (1(2020) × 1(month > M arch = ω )) ω =0 n 2019 (1) + ω βk (1(k ) × 1(month > M arch = ω )) + ϕXjmt + ϵimt ω =0 k=2018 where lnYimt is the log of the monthly electricity consumption (purchases) of household 10 Knutsen et al. (2017) for instance, compares corruption experiences of citizens living in close proximity to gold mining firms with the corruption experiences of citizens living in areas close to a yet to be opened mine in a difference-in-difference design to estimate the effect of mining on corruption in Africa. 7 (firm) i in month m(= 1, 2, .., 12) and year t(= 2018, 2019, 2020).11 αi denotes the house- hold (firm) fixed effects to capture time-invariant heterogeneity across households (firms); λjt denotes district-year fixed effects to absorb (temporal) seasonal patterns that are spe- cific to each district. 1(2020) × 1(month > M arch = ω ) are monthly event dummies that are set equal to 1 if the consumption occurred in month m relative to the start of COVID-19 SAH (lockdown) which in the case of Rwanda was March 2020.12 However, in the case of Ghana, the lockdown went into effect on March 31, 2020, and lasted until April 20, 2020. Hence, in this setup, we treat April as, effectively, the start of the SAH order in Ghana.13 Similarly, n k=2018 βk (1(k ) × 1(month > M arch = ω )) represent monthly event 2019 ω ω =0 dummies for the months between March and December in the respective years. Ximt is a vector of time-varying controls, mainly, monthly temperature and precipitation in the household’s (firm’s) locality (district), to absorb the effects of changing weather patterns on electricity demand. ϵimt is the idiosyncratic error term. The parameter, β2020 ω , estimates the pre-post (first) differences in electricity consumption in the months after the SAH relative to the months immediately before the SAH in 2020.14 As indicated earlier, there is evidence of seasonalities in electricity consumption, which may confound the β ω estimates. Therefore, to estimate the effect of COVID-19 on consump- tion, we use households’ (firms’) own electricity consumption in the previous non-COVID years (2018-2019) as a counterfactual (Irwin et al., 2021). The intuition is that, in the ab- sence of COVID, household’s (firm’s) monthly consumption in 2020 would have evolved along a similar pattern as their respective average monthly consumption in the preceding years (Irwin et al., 2021). To this end, we (arbitrarily) roll back the March (April) SAH order to the previous non-COVID years (2018-2019) and estimate the pre-post differences over the period (denoted by β2018 ω and β2019 ω respectively), and use these estimates to adjust 11 It is important to emphasize that for customers on prepaid electricity meters, purchases may not be necessarily equal to consumption in a given month, as a consumer can purchase more electricity than she can consume in a given month, effectively consuming the remainder in the following month. However, this is less likely to be an issue given that the majority of households make small purchases, several times a month, effectively purchasing what they need for a given period (day/week), probably due to income constraints. Hence, at the margin, purchases are likely to approximate consumption. 12 The stay-at-home order (SAH) in Rwanda was implemented between March 21 and May 4, 2020 13 Accordingly, our baseline equation for estimating the impact in Ghana can be specified as follows: n Yimt = αi + λjt + ω β2020 (1(2020) × 1(month > April = ω )) ω =0 n 2019 + ω βk (1(k ) × 1(month > April = ω )) + ϕXjmt + ϵimt ω =0 k=2018 14 i.e., January and February 8 for non-COVID related effects that may confound the β2020ω estimates. Therefore, by using the average pre-post (first) difference in consumption in the pre- vious non-COVID years (2018-2019) as the counterfactual, our difference-in-differences estimate is expressed as: 2019 ˆω − 1 ω DID = β ˆω β (2) 2020 2 k=2018 k where DIDω essentially measures the changes in electricity consumption during the COVID months relative to the typical consumption levels in the same periods during the previous years. Standard errors are clustered at the household (firm) level. 5 Results In this section, we present the results on the effect of the COVID pandemic on electricity consumption in the two countries, as well as the role of electricity subsidies in Ghana. 5.1 Rwanda We start by showing the average effects of the COVID pandemic on electricity consump- tion in Rwanda, estimated separately for the various user types. For each user type, we estimate two variant specifications: with and without climate controls (average monthly temperature and total precipitation). Our preferred specifications are where we include the full set of climate controls, customer and district-year fixed effects, albeit the results are largely robust to the exclusion of the climate controls. Results are shown in Table 1. The parameters of interest are the DID estimates, which compare the pre-post differ- ences in consumption during the period March-December 2020 with the average pre-post differences in consumption over the same period in the pre-pandemic years. Thus, con- ditional on the fixed effects and the climate controls, pre-post differences in consumption over the same period in the pre-pandemic years are assumed to be the changes in con- sumption levels in the absence of the pandemic, i.e., counterfactual. The results in Table 1 suggest that the pandemic led to a modest, 1.7% increase in (monthly) electricity consumption in Rwanda (column 2). However, there appears to be significant heterogeneity in the impact of COVID on consumption patterns across var- ious users. For instance, while electricity consumption in the residential sector increased 9 by 3.8% (column 4), it declined by 10.7% among non-residential customers (column 6).15 The effect on consumption by small and medium industries is negative albeit not statisti- cally significant (column 8). Beyond the average impact, we examine how the impact of COVID on electricity con- sumption varies over time. To this end, we estimate the dynamic model specified in equa- tion (1) for the various user categories as well as the full sample. Results on the DID estimates are presented in Table 2, while the first difference estimates are shown in Table A1 in the Appendix. The results from Table 2 are summarized and displayed in Figure 1. The results in Figure 1 show interesting trends, that plausibly reflects the dynamic re- sponses of various users to the effects of the pandemic. For instance, in Figure 1a, while the consumption of electricity by all users increased by about 2% during the start of the COVID lockdown in Rwanda, consumption declined by approximately the same level in the second month of the lockdown (April 2020), before rising steadily thereafter, to about 5%, seven months after the start of the pandemic. This trend plausibly highlights the ef- fects of easing COVID restrictions on the economy. Consumption by the residential sector increased however throughout the period, rang- ing between 2.2% and 6.8%. The SAH orders and associated Home-Based-Work, are pos- sible reasons for the increase in consumption by residential users. Interestingly, Figure 1c shows a consistent decline in non-residential electricity consumption from March through December 2020. The largest decline was recorded in one month after the start of the SAH (lockdown), with consumption falling by 24% relative to the average consumption during the same period in the preceding years.16 Despite signs of recovery from months 2-9 after the start of the SAH order, consumption levels never reached the pre-pandemic levels. The effect on the industrial sector, however, appears to be short-lived (Figure 1d). Consump- tion fell significantly by 13.5% and 21% during the first and second months of the SAH, but bounced back to pre-pandemic levels after the lifting of the SAH.17 To fully understand the significant decline in non-residential electricity consumption during the pandemic, we conduct separate analyses for the main groups within the non- residential category, namely: hotels, health centers, and others. The DID estimates are presented in Table 3 and Figure 2.18 The results on the effects of COVID on electricity consumption of commercial units (Figure 2a) mimics the changes in consumption for the overall non-residential sector (Figure 1c). This is largely a result of the fact that com- 15 In a log-linear model, the effect size is computed as (expβ −1)×100. Hence (exp−.113 −1)×100 = −10.7% 16 (exp−.279 − 1) × 100 = 24% 17 (exp−.239 − 1) × 100 = −21.1% 18 The corresponding first difference estimates are shown in Table A2 in the Appendix 10 mercial units account for about 90% of the non-residential sample in our study. Figure 2b also indicate that hotels (tourism sector) were heavily affected by the pandemic as electricity consumption declined significantly during the three months lockdown period. While consumption improved after the lockdown was lifted, it still remained lower than pre-pandemic levels. This result is in line with economic data from several countries in- dicating the devastating impact of the pandemic on the tourism sector’s contribution to GDP19 due to SAH and restrictions on international travels (Mulder, 2020; Abbas et al., 2021). Rural-Urban differences Finally, we explore the differences in the impact of COVID on electricity consumption between rural and urban households. The average effects are shown in Table 4. The DID estimates suggest electricity consumption among urban households increased by about 5.3% (column 2) compared to a 3.2% (column 4) increase among rural households. To trace the dynamics in the effects, we show estimates of the monthly changes in electricity consumption in Figure 3 (see Table 5 for details20 ). The results in Figure 3 show an in- creased level of consumption among both rural and urban households, albeit the rate of increase was consistently higher among urban households for the majority of the period. Taken together, the analysis of the Rwandan electricity data suggests a modest increase in electricity consumption associated with the COVID-19 pandemic, albeit, non-residential electricity consumption experienced large declines. This partly reflects the differing re- sponses to the impact of the pandemic across various sectors. 5.2 Ghana We now turn our attention to the Ghana data. Similar to the analysis in Section 5.1, we be- gin by looking at the average effects of the pandemic on electricity consumption in Ghana, estimated separately for the various user groups, as shown in Table 6. The DID estimate in column 2 suggests that overall, relative to the pre-COVID lockdown period, electric- ity consumption increased by about 20.8%.21 This increase appears to be largely driven by residential electricity consumption, which increased by 22.7% during the period (col- umn 4), while consumption by non-residential customers increased only marginally by 19 https://www.imf.org/en/Publications/fandd/issues/2020/12/impact-of-the-pandemic-on-tourism-behsudi 20 See Table A3 for the corresponding first difference estimates. 21 (exp.189 − 1) × 100 = 20.8% 11 5.9% (column 6). Industrial consumption, however, fell by 11.9% during the same period (column 8). We also examined the time-varying effects of COVID on consumption. The DID esti- mates are shown in Figure 4 and Table 7.22 The results show an overall increase in elec- tricity consumption relative to the pre-COVID period, with the highest increase observed during the first month of the COVID lockdown in Ghana (April 2022). During the first month of the start of the SAH order, the average consumption (all users) increased by about 53%23 and reducing to about 10% in the fourth month after the start of the SAH, before increasing slightly thereafter (see Figure 4a or Table 7 column 2). A similar trend is observed for residential consumption as shown in Figure 4b. The spike in consumption and subsequent decline in the rate of increase in consumption can be largely attributed to the SAH in Ghana which lasted for just three weeks. The relaxation of some COVID restrictions and gradual opening of sectors of the economy potentially led to a resump- tion of normal activities and thus households spending more time away from home than was required during the lockdown period. For non-residential customers, we observe a spike in consumption of about 21.5% during the SAH period and declining thereafter to consumption levels close to the pre-COVID period, and rising marginally between 4% and 10% between the third and eight months after the start of the SAH order (Figure 4c). In- dustrial consumption however declined steeply (≈ 30%) during the SAH period (Figure 4d). Although the rate of decline in consumption slowed after the SAH period, consump- tion levels remained lower throughout the entire period relative to the consumption levels in the pre-COVID period. Once again, the decline in electricity consumption by (heavy) industries in Ghana partly reflects the negative effect of the pandemic on the productive sectors of the economy. Due to the lack of detailed information on the sectoral composi- tion of the non-residential customers, we are unable to conduct a detailed analysis within the sector as was done in the case of Rwanda. 22 The corresponding first difference estimates are shown in Table A4 in the Appendix. 23 (exp.431 − 1) × 100 = 53.87% 12 Effects of the SAH order on Consumption Unlike Rwanda where the SAH was imposed largely across the entire country, the case was different in Ghana. A lockdown was imposed only in the two large metropolitan areas: Accra and Kumasi, which jointly account for a relatively large share of the popu- lation and electricity customers. Nonetheless, it is important to emphasize that beyond the SAH order, restrictions on social gatherings and other outdoor events were imposed across the entire country from March 2020 through January 2021. In this section, we ex- plore the potential effect of the SAH order on electricity consumption. To this end, we estimate the baseline results separately for users in the lockdown areas as well as those outside the lockdown areas. Results (DID estimates) are presented in Figure 5 (details in Tables 8 and 9). For residential and non-residential customers, we find no evidence of systematic differences in changes in consumption of users in lockdown and non-lockdown communities. The relatively short duration of the SAH order and the general restrictions on social gatherings across the country are plausible reasons for the seemingly similar pat- terns in consumption. However, we observe significant differences in changes in industrial consumption between lockdown and non-lockdown communities (Figure 5d). Interest- ingly, industrial consumption in the two areas fell significantly during the SAH month, albeit with a large decline in lockdown communities. However, industrial consumption in non-lockdown communities returned to pre-COVID levels immediately thereafter, while industrial consumption in lockdown communities continued to remain lower than pre- COVID levels, despite a slight improvement. 5.3 Role of Subsidies So far, our results have shown increased electricity consumption during the COVID period in both countries, with residential users experiencing the highest gains in consumption. However, there are marked differences in the rate of increase in residential consumption between the two countries: the increase in residential consumption during the COVID pe- riod in Ghana is about five-fold the rate of increase in Rwanda. Did electricity subsidies offered by the Ghanaian government during the period play a role in elevating consump- tion? In this section, we assess the effects of the subsidies in elevating consumption during the COVID era, by leveraging a unique feature of the electricity subsidies in Ghana: where lifeline (low-income households) customers were offered 100% subsidy for 12 months 13 while non-lifeline customers enjoyed a 50% subsidy for the first three months. Specifi- cally, we adopt two strategies: First, we replicate the main analysis by estimating equation 2 separately for lifeline and non-lifeline residential customers and examine the trends in consumption over time. This allows us to see the trends in consumption for both lifeline and non-lifeline customers, especially the response of non-lifeline customers after the removal of their partial subsidy. The results as shown in Figure 6 reveal interesting trends in the responses to the subsidy. We see that relative to the pre-subsidy (COVID) period, electricity consumption by lifeline customers increased by between 35% and 70%. On average, between April and Decem- ber 2020, electricity consumption of lifeline customers (100% subsidy) increased by about 64% relative to the average consumption in the pre-subsidy period (see Table 10 column 224 ). On the contrary, for non-lifeline customers (50% subsidy) demand increased only during the first two months of the subsidy, falling sharply after the subsidy removal. In fact, for the most part of the post-subsidy period, monthly consumption was below the pre-subsidy (covid) period (see Figure 6).25 During the three-month period (April-June) where the partial subsidy was offered to non-lifeline customers, average consumption in- creased by 11% relative to the pre-subsidy period and declining thereafter by 4.3% dur- ing the post-subsidy (see Table 10 column 4). Thus, comparatively, the results provide suggestive evidence that the electricity subsidies played a key role in the relatively high increase in electricity consumption during the COVID period in Ghana. At the same time, the decline in consumption among non-lifeline residential customers after the removal of the subsidies relative to the pre-COVID era is indicative of the potential effect of COVID on household electricity consumption if the subsidies were not offered. Our second strategy exploits the differences in the duration of the subsidy across res- idential user types to estimate the effect of the subsidy on consumption. Given that the subsidy for non-lifeline customers expired after the third month, while lifeline customers continued to enjoy the subsidy, we leverage the removal of the subsidy for the non-lifeline customers to conduct a "conventional" difference-in-difference analysis by using consump- tion data of the two groups between April-December, 2020 and comparing the differences in consumption between lifeline and non-lifeline customers before and after July 2020 when the subsidy for non-lifeline customers expired. So essentially, we estimate the ef- fect of subsidy removal to identify the extent to which subsidies contributed to the stark increase in electricity consumption in Ghana as shown by the preceding results. Note that 24 (exp.494 − 1) × 100 = 63.8% 25 An exception is month t = 8, i.e. December 2020, where we see a spike in consumption. 14 the classification of lifeline vs non-lifeline customers in the subsidy program was deter- mined based on March consumption, and since the subsidy was announced on March 30, consumers plausibly had no control in manipulating the extent of the subsidy they re- ceived. Also, at the start of the program, both customer classes were expected to receive subsidies for a period of three months. It was not until the end of the first phase of the program that an announcement was made that the subsidy was being extended for lifeline customers. Consequently, we estimate the following equation: Yimt = αi + λjt + β · 1(non − lif eline) × postsubsidy + ϕXjmt + ϵimt (3) where 1(non − lif eline) is an indicator variable set equal to 1 for non-lifeline customer, de- fined as consuming more than 50 kWh of electricity during the month of March, 2020 and 0 if otherwise. postsubsidy is an indicator variable equal to 1 for the period July-December, 2020 when the partial subsidy for non-lifeline customers was removed. All other variables remain as previously defined. The parameter of interest is β which measures the relative difference in consumption between lifeline and non-lifeline customers before and after the removal of the subsidies for the latter. Standard errors are clustered at the district- customer type level. We show the results in Table 11. Across the various specifications, we find robust ev- idence of a decline in consumption by non-lifeline tariffs after the removal of their subsi- dies, relative to lifeline customers that enjoyed subsidies throughout the study period. In column 2, for instance, the results show a 15% decline in consumption after the subsidy removal. Put differently, the partial subsidies induced at least a 15% increase in consump- tion. Such a sizeable increase in consumption is plausibly responsible for the gap in the residential electricity consumption between Ghana and Rwanda during the COVID pe- riod. Thus to summarize, the main takeaway from the analysis herein is that: (i) the COVID relief packages via electricity subsidies offered by the government of Ghana were useful in mitigating the effects of COVID on households, as it allowed them to sustain (electricity) consumption; (ii) however, at the same time the subsidies also induced unin- tended consequences: excessive consumption among recipient households. Finally, critics argue that the design of the COVID electricity subsidy in Ghana was regressive, with middle/high-income (non-lifeline) consumers receiving higher amounts (value) of subsidies and low-income (lifeline) consumers receiving low subsidy amounts (Berkouwer et al., 2022). The initial design of the subsidy offered, respectively, lifeline and non-lifeline consumers 100% and 50% subsidies on their consumption between April and June 2020. Thus while the rate of the subsidy is progressive, the actual amounts received 15 by households appear to be regressive since non-lifeline consumers will in principle re- ceive high-income transfers due to their relatively high consumption. However, we show that the extension of the subsidy from the initial three months to 12 months, for [only] lifeline customers, made the subsidy progressive than regressive. Using the actual consumption of both consumer groups, we compute the [expected] subsidy (implicit transfer) to the median lifeline and non-lifeline customers (households) during the period as shown in Figure ??. First, we compute the monthly subsidy received by the median household in each consumer group between April and December 2020 by multiplying the subsidy rate by their respective average monthly consumption. Averag- ing this over the period, our estimates suggest that on average, the median lifeline and non-lifeline household received GHS 12 (USD 2.14) and GHS 29 (USD 5.18) per month respectively.26 By aggregating these over the 12 months period that lifeline consumers received the subsidies, the median lifeline customer received a total of GHS 139 (USD 25). On the other hand, non-lifeline consumers received only GHS 86 (USD 15) over their three-month subsidy period. Thus, cumulatively, the total subsidies to lifeline customers were about 67% higher than those to non-lifeline customers. Admittedly, these estimates are only indicative, and may not reflect the true amount received by households largely because the subsidies were administered at the meter level, and given the prevalence of shared metering particularly among low-income households, the share of the transfer re- ceived by households may in practice be much lower (Berkouwer et al., 2022). 6 Discussion and Conclusion This paper examines the effect of the COVID pandemic on electricity consumption pat- terns among households and firms in Sub-Saharan Africa using administrative data on billing records from Ghana and Rwanda. We also explore the effects of SAH orders (lock- down) and pandemic relief measures such as electricity subsidies on household consump- tion. Six main results emerge from the paper. First, residential electricity consumption dur- ing the pandemic increased in both Ghana and Rwanda, albeit the rate of increase was high in Ghana than in Rwanda. Specifically, average household consumption increased by about 23% in Ghana compared to 4% in Rwanda. Second, consumption in the non- residential sector declined by about 11% in Rwanda, while rising marginally by about 6% in Ghana. Third, industrial consumption declined in both countries, highlighting the ef- 26 Based on an exchange rate of USD 1: 5.6 GHS 16 fect of COVID-induced business closure and supply chain disruptions on the activities of the industrial sector. Fourth, the effect of COVID on electricity consumption was pro- nounced in lockdown and non-lockdown communities, as well as, rural and urban com- munities. This highlights the effects of the pandemic across various segments of the popu- lation. Fifth, pandemic relief measures such as electricity subsidies administered in Ghana played a key role in raising consumption. While the subsidies were important to sustain- ing households’ consumption from falling below pre-pandemic levels, they also induced an unintended consequence: excess consumption among households. Finally, contrary to earlier criticisms that the design of the subsidy program in Ghana was regressive, our back-of-the-envelope calculations suggest that the subsidy amounts to low-income (life- line) customers were about 67% higher than total subsidies received by high-income (non- lifeline) households (customers). This was largely driven by the extension of the subsidy for a longer duration for low-income households. Overall, the findings from this paper highlight the effects of the pandemic on house- holds and firms, and the role of relief measures such as subsidies in mitigating the effect of the pandemic on end-users. 17 References Abay, K., Berhane, G., Hoddinott, J., and Tafere, K. (2020). Covid-19 and food security in ethiopia: do social protection programs protect? Abbas, J., Mubeen, R., Iorember, P. T., Raza, S., and Mamirkulova, G. (2021). Exploring the impact of covid-19 on tourism: transformational potential and implications for a sustainable recovery of the travel and leisure industry. Current Research in Behavioral Sciences, 2:100033. Adams-Prassl, A., Boneva, T., Golin, M., and Rauh, C. (2020). Inequality in the impact of the coronavirus shock: Evidence from real time surveys. Journal of Public economics, 189:104245. Albanesi, S. and Kim, J. (2021). Effects of the covid-19 recession on the us labor market: Occupation, family, and gender. Journal of Economic Perspectives, 35(3):3–24. Amare, M., Abay, K. A., Tiberti, L., and Chamberlin, J. (2021). Covid-19 and food security: Panel data evidence from nigeria. Food Policy, 101:102099. Basurto, M. P., Dupas, P., and Robinson, J. (2020). Decentralization and efficiency of subsidy targeting: Evidence from chiefs in rural malawi. Journal of public economics, 185:104047. Berkouwer, S. B., Biscaye, P. E., Puller, S., and Wolfram, C. D. (2022). Disbursing emer- gency relief through utilities: Evidence from ghana. Journal of Development Economics, 156:102826. Cicala, S. (2020). Powering work from home. Technical report, National Bureau of Eco- nomic Research. Cong, L. W., Yang, X., and Zhang, X. (2021). Smes amidst the pandemic and reopening: Digital edge and transformation. The SC Johnson College of Business Applied Economics and Policy Working Paper Series No. Cortes, G. M. and Forsythe, E. (2020). Heterogeneous labor market impacts of the covid-19 pandemic. ILR Review, page 00197939221076856. Davis, L. W. (2014). The economic cost of global fuel subsidies. American Economic Review, 104(5):581–85. Gentilini, U., Almenfi, M., Orton, I., et al. (2020). Social protection and jobs responses to covid-19: a real-time review of country measures. Hahn, R. W. and Metcalfe, R. D. (2021). Efficiency and equity impacts of energy subsidies. American Economic Review, 111(5):1658–88. Irwin, N. B., McCoy, S. J., and McDonough, I. K. (2021). Water in the time of corona (virus): The effect of stay-at-home orders on water demand in the desert. Journal of Environmental Economics and Management, 109:102491. Isaksson, A.-S. and Kotsadam, A. (2018). Chinese aid and local corruption. Journal of Public Economics, 159:146–159. Klein, T. (2022). Agree to disagree? predictions of us nonfarm payroll changes between 2008 and 2020 and the impact of the covid19 labor shock. Journal of Economic Behavior & Organization, 194:264–286. Knutsen, C. H., Kotsadam, A., Olsen, E. H., and Wig, T. (2017). Mining and local corrup- tion in africa. American Journal of Political Science, 61(2):320–334. 18 Komives, K., Foster, V., Halpern, J., Wodon, Q., et al. (2005). Water, electricity, and the poor: Who benefits from utility subsidies? World Bank Publications. Mulder, N. (2020). The impact of the covid-19 pandemic on the tourism sector in latin america and the caribbean, and options for a sustainable and resilient recovery. Saltane, V., Ereshchenko, V., Hovhannisyan, S., and Mensah, J. T. (2022). Responses of the electricity sector in 120 economies to the covid-19 pandemic. Schmidt, E., Dorosh, P., and Gilbert, R. (2021). Impacts of covid-19 induced income and rice price shocks on household welfare in papua new guinea: Household model esti- mates. Agricultural Economics, 52(3):391–406. Younger, S. et al. (2016). The impact of reforming energy subsidies, cash transfers, and taxes on inequality and poverty in ghana and tanzania. Technical report, Tulane Uni- versity, Department of Economics. Zhang, R., Li, H., and Khanna, N. (2021). Environmental justice and the covid-19 pan- demic: Evidence from new york state. Journal of environmental economics and management, 110:102554. 19 7 Figures Figure 1: COVID-19 and Electricity Consumption in Rwanda (a) All (b) Residential (c) Non-Residential (d) Small and Medium Industries The figure shows the changes in electricity consumption in the respective months following the start of the COVID lockdown in Rwanda (March 2020). The estimates are difference-in-difference estimates and their corresponding 95% confidence intervals are computed using the expression: DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k , with parameter estimates from equation (1). 20 Figure 2: COVID-19 and Electricity Consumption in the Non-Residential Sector in Rwanda (a) Commercial (b) Hotels (c) Health Centers (d) Public works The figure shows the changes in electricity consumption in the respective months following the start of the COVID lockdown in Rwanda (March 2020). The estimates are difference-in-difference estimates and their corresponding 95% confidence intervals are computed using the expression: DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k , with parameter estimates from equation (1). 21 Figure 3: Rural-Urban differences in Effects of COVID on Residential Electricity Consump- tion in Rwanda The figure shows the changes in electricity consumption in the respective months following the start of the COVID lockdown in Rwanda (March 2020). The estimates are difference-in-difference estimates and their corresponding 95% confidence intervals are computed using the expression: DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k , with parameter estimates from equation (1). 22 Figure 4: COVID-19 and Electricity Consumption in Ghana (a) All (b) Residential (c) Non-Residential (d) Heavy Industries The figure shows the changes in electricity consumption in the respective months following the start of the COVID lockdown in Ghana (April 2020). The estimates are difference-in-difference estimates and their corresponding 95% confidence intervals are computed using the expression: DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k , with parameter estimates from equation (1). 23 Figure 5: COVID-19 Lockdowns and Electricity Consumption in Ghana (a) All (b) Residential (c) Non-Residential (d) Heavy Industries The figure shows the changes in electricity consumption in the respective months following the start of the COVID lockdown in Ghana (April 2020). The estimates are difference-in-difference estimates and their corresponding 95% confidence intervals are computed using the expression: DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k , with parameter estimates from equation (1). In each panel, the estimates for lockdown and non-lockdown areas are obtained from separate regressions. 24 Figure 6: Electricity Subsidies and Consumption in Ghana The figure shows the changes in electricity consumption in the respective months following the start of the COVID lockdown in Ghana (April 2020). The estimates are difference-in-difference estimates and their cor- responding 95% confidence intervals are computed using the expression: DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k , with parameter estimates from equation (1). In each panel, the estimates for lifeline and non-lifeline cus- tomers are obtained from separate regressions. 25 Figure 7: Amount of Electricity Subsidies to Ghanaian Households during the COVID-19 Pandemic The figure shows the amount in Ghana Cedis (GHS) of subsidies received by the median household. Aver- age exchange rate during the period 1 USD: 5.5 GHS. 26 7.1 Tables Table 1: COVID-19 and Electricity Consumption in Rwanda log of Consumption All Residential Non-Residential Small& Medium Industries (1) (2) (3) (4) (5) (6) (7) (8) I (Mar-Dec 2020) 0.026*** 0.024*** 0.046*** 0.044*** -0.090*** -0.095*** 0.072*** 0.070** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.027) (0.027) I (Mar-Dec 2019) 0.016*** 0.016*** 0.015*** 0.014*** 0.033*** 0.033*** 0.050** 0.057** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.025) (0.024) I (Mar-Dec 2018) -0.002** -0.003*** -0.001 -0.002*** 0.007*** 0.003* 0.131*** 0.129*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.034) (0.034) DID 0.019*** 0.017*** 0.039*** 0.038*** -0.110*** -0.113*** -0.018 -0.023 (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.032) (0.032) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.858 0.858 0.849 0.849 0.856 0.856 0.719 0.720 Mean dep. Var 27.552 27.552 20.426 20.426 67.526 67.526 256.741 256.741 Obs 21352142 21352142 17097530 17097530 3212773 3212773 13345 13345 Notes: Dependent variable is the log of monthly electricity consumption. I (Mar-Dec 2020) is an indicator variable equal 1 for the period March 1- December 31 2020. Likewise, I (Mar-Dec 2019) and I (Mar-Dec 2018) are indicator for the same period March 1- December 31 in 2019 and 2018 respectively. The reference group is the period Jan-Feb in 2018, 2019, and 2020. DID estimates are derived from the expression DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k . Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 27 Table 2: COVID-19 and Electricity Consumption in Rwanda log of Consumption All Residential Non-Residential Small & Medium Industries (1) (2) (3) (4) (5) (6) (7) (8) DID (Month> March=0) 0.017*** 0.019*** 0.032*** 0.036*** -0.074*** -0.072*** -0.130*** -0.146*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.038) (0.041) DID (Month> March=1) -0.017*** -0.017*** 0.022*** 0.022*** -0.279*** -0.279*** -0.240*** -0.239*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.058) (0.058) DID (Month> March=2) -0.004*** 0.001 0.029*** 0.033*** -0.177*** -0.172*** -0.052 -0.043 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.042) (0.044) DID (Month> March=3) 0.003*** 0.002** 0.027*** 0.026*** -0.130*** -0.131*** 0.029 0.033 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.042) (0.042) DID (Month> March=4) 0.010*** 0.007*** 0.031*** 0.027*** -0.112*** -0.114*** 0.031 0.04 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.043) (0.043) DID (Month> March=5) 0.020*** 0.030*** 0.039*** 0.049*** -0.101*** -0.091*** 0.03 0.041 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.043) (0.048) DID (Month> March=6) 0.044*** 0.041*** 0.063*** 0.061*** -0.084*** -0.085*** 0.066 0.071* (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.044) (0.043) DID (Month> March=7) 0.054*** 0.052*** 0.071*** 0.068*** -0.056*** -0.057*** 0.038 0.049 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.044) (0.045) DID (Month> March=8) 0.037*** 0.027*** 0.048*** 0.038*** -0.049*** -0.060*** 0.02 0.019 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.046) (0.048) DID (Month> March=9) 0.024 0.022 0.031 0.028 -0.062 -0.062 -0.022 0 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.047) (0.048) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes Mean dep. Var 27.552 27.552 20.426 20.426 67.526 67.526 256.741 256.741 Obs 21352142 21352142 17097530 17097530 3212773 3212773 13345 13345 Notes: Dependent variable is the log of monthly electricity consumption. DID (Month> April= ω ) represents the DID estimates for the ω months since the start of COVID lockdown (April, 2020). The DID estimates are derived from the expression DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k . Climate Ctrls include the monthly average temperature and total precipitation . Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 28 Table 3: COVID-19 and Electricity Consumption in Non-Residential Sector in Rwanda log of Consumption Commercial Hotels Health Centers Public Works (1) (2) (3) (4) (5) (6) (7) (8) DID (Month> March=0) -0.075*** -0.073*** -0.067*** -0.069*** 0.079*** 0.086*** 0.007 0.003 (0.002) (0.002) (0.027) (0.029) (0.028) (0.029) (0.035) (0.036) DID (Month> March=1) -0.280*** -0.280*** -0.270*** -0.269*** 0.004 0.004 0.013 0.013 (0.003) (0.003) (0.043) (0.043) (0.029) (0.028) (0.054) (0.054) DID (Month> March=2) -0.178*** -0.173*** -0.179*** -0.177*** 0.02 0.03 -0.077* -0.076 (0.003) (0.003) (0.039) (0.039) (0.031) (0.032) (0.045) (0.048) DID (Month> March=3) -0.131*** -0.132*** -0.146*** -0.146*** 0.075*** 0.074** -0.063 -0.061 (0.003) (0.003) (0.04) (0.04) (0.032) (0.032) (0.041) (0.041) DID (Month> March=4) -0.112*** -0.115*** -0.122*** -0.121*** 0.061** 0.054* -0.043 -0.04 (0.003) (0.003) (0.041) (0.042) (0.03) (0.03) (0.045) (0.045) DID (Month> March=5) -0.101*** -0.091*** -0.108*** -0.105*** 0.048 0.068** -0.034 -0.033 (0.003) (0.003) (0.039) (0.039) (0.031) (0.033) (0.042) (0.052) DID (Month> March=6) -0.084*** -0.086*** -0.123*** -0.122*** 0.049* 0.043 -0.021 -0.019 (0.003) (0.003) (0.041) (0.041) (0.027) (0.027) (0.049) (0.05) DID (Month> March=7) -0.056*** -0.057*** -0.087** -0.085** -0.018 -0.024 -0.001 0.003 (0.003) (0.003) (0.039) (0.04) (0.031) (0.031) (0.053) (0.054) DID (Month> March=8) -0.050*** -0.060*** -0.110*** -0.112*** 0.095*** 0.072*** 0.023 0.024 (0.003) (0.003) (0.043) (0.047) (0.028) (0.029) (0.057) (0.054) DID (Month> March=9) -0.062 -0.063 -0.083 -0.08 0.098 0.094 -0.004 0.001 (0.003) (0.004) (0.043) (0.044) (0.027) (0.028) (0.06) (0.062) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes Mean dep. Var 65.244 65.244 353.152 353.152 359.381 359.381 398.729 398.729 Obs 3188121 3188121 9014 9014 12167 12167 3471 3471 Notes: Dependent variable is the log of monthly electricity consumption. DID (Month> April= ω ) represents the DID estimates for the ω months since the start of COVID lockdown (April, 2020). The DID estimates are derived from the expression DID ω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k . Climate Ctrls include the monthly average temperature and total precipitation . Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 29 Table 4: COVID-19 and Electricity Consumption in Rwanda: Urban vs Rural Residential Urban Rural (1) (2) (3) (4) I (Mar-Dec 2020) 0.070*** 0.068*** 0.035*** 0.033*** (0.001) (0.001) (0.001) (0.001) I (Mar-Dec 2019) 0.030*** 0.029*** 0.007*** 0.006*** (0.001) (0.001) (0.001) (0.001) I (Mar-Dec 2018) 0.003*** 0.002 -0.003*** -0.004*** (0.001) (0.001) (0.001) (0.001) DID 0.054*** 0.053*** 0.033*** 0.032*** (0.001) (0.001) (0.001) (0.001) Climate Ctrls No Yes No Yes Customer FE Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes R-squared 0.821 0.821 0.793 0.793 Mean dep. Var 37.415 37.415 11.670 11.670 Obs 5815537 5815537 11279386 11279386 Notes: Dependent variable is the log of monthly electricity consumption. I (Mar-Dec 2020) is an indicator variable equal 1 for the period March 1- December 31 2020. Likewise, I (Mar-Dec 2019) and I (Mar-Dec 2018) are indicator for the same period March 1- December 31 in 2019 and 2018 respectively. The reference group is the period Jan-Feb in 2018, 2019, and 2020. DID estimates are derived from the expression DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k . Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 30 Table 5: COVID-19 and Electricity Consumption in Rwanda: Urban vs Rural Residential Urban Rural (1) (2) (3) (4) DID (Month> March=0) 0.047*** 0.052*** 0.025*** 0.027*** (0.001) (0.001) (0.001) (0.001) DID (Month> March=1) 0.034*** 0.032*** 0.017*** 0.017*** (0.002) (0.002) (0.001) (0.001) DID (Month> March=2) 0.049*** 0.063*** 0.019*** 0.022*** (0.002) (0.002) (0.001) (0.001) DID (Month> March=3) 0.046*** 0.043*** 0.017*** 0.017*** (0.002) (0.002) (0.001) (0.001) DID (Month> March=4) 0.043*** 0.034*** 0.025*** 0.023*** (0.002) (0.002) (0.001) (0.001) DID (Month> March=5) 0.058*** 0.084*** 0.031*** 0.038*** (0.002) (0.002) (0.001) (0.001) DID (Month> March=6) 0.087*** 0.083*** 0.053*** 0.051*** (0.002) (0.002) (0.001) (0.001) DID (Month> March=7) 0.090*** 0.083*** 0.063*** 0.062*** (0.002) (0.002) (0.001) (0.001) DID (Month> March=8) 0.056*** 0.026*** 0.047*** 0.039*** (0.002) (0.002) (0.001) (0.001) DID (Month> March=9) 0.028 0.021 0.034 0.034 (0.002) (0.002) (0.001) (0.001) Climate Ctrls No Yes No Yes Customer FE Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes R-squared 0.821 0.821 0.793 0.793 Mean dep. Var 37.415 37.415 11.670 11.670 Obs 5815537 5815537 11279386 11279386 Notes: Dependent variable is the log of monthly electricity consumption. DID (Month> April= ω ) represents the DID estimates for the ω months since the start of COVID lockdown (April, 2020). The DID estimates are derived from the expression DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k . Climate Ctrls include the monthly average temperature and total precipitation . Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 31 Table 6: COVID-19 and Electricity Consumption in Ghana: Average Effects log of Consumption All Residential Non-Residential Heavy Industries (1) (2) (3) (4) (5) (6) (7) (8) I (Apr-Dec 2020) 0.161*** 0.212*** 0.176*** 0.228*** 0.028*** 0.071*** -0.159*** -0.126*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.015) (0.014) I (Apr-Dec 2019) -0.017*** 0.036*** -0.016*** 0.037*** -0.020*** 0.026*** -0.041*** -0.005 (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.014) (0.013) I (Apr-Dec 2018) -0.047*** -0.009*** -0.049*** -0.010*** -0.035*** 0.001 -0.035** -0.008 (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.015) (0.015) DID 0.145*** 0.189*** 0.160*** 0.205*** 0.021*** 0.058*** -0.155*** -0.127*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.018) (0.018) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.709 0.710 0.685 0.686 0.774 0.774 0.858 0.858 Mean dep. Var 383.1 383.115 123.781 123.781 419.210 419.210 152470.284 152470.284 Obs 41958788 41958777 37299880 37299869 4602780 4602780 52317 52317 Notes: Dependent variable is the log of monthly electricity consumption. I (Apr-Dec 2020) is an indicator variable equal 1 for the period April 1- December 31 2020. Likewise, I (Apr-Dec 2019) and I (Apr-Dec 2018) are indicator for the same period April 1- December 31 in 2019 and 2018 respectively. The reference group is the period Jan-Mar in 2018, 2019, and 2020. DID estimates are derived from the expression DID ω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k . Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 32 Table 7: COVID-19 and Electricity Consumption in Ghana: Dynamic Effects log of Consumption All Residential Non-Residential Heavy Industries (1) (2) (3) (4) (5) (6) (7) (8) DID (Month> April=0) 0.430*** 0.431*** 0.457*** 0.459*** 0.195*** 0.195*** -0.361*** -0.357*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.021) (0.022) DID (Month> April=1) 0.173*** 0.180*** 0.196*** 0.203*** -0.019*** -0.016*** -0.159*** -0.159*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.022) (0.022) DID (Month> April=2) 0.162*** 0.142*** 0.180*** 0.158*** 0.009*** -0.002 -0.087*** -0.052* (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.024) (0.028) DID (Month> April=3) 0.141*** 0.152*** 0.155*** 0.166*** 0.030*** 0.038*** -0.097*** -0.123*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.004) (0.022) (0.025) DID (Month> April=4) 0.138*** 0.098*** 0.147*** 0.104*** 0.062*** 0.047*** -0.107*** -0.089*** (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.024) (0.026) DID (Month> April=5) 0.196*** 0.166*** 0.209*** 0.177*** 0.086*** 0.074*** -0.071*** -0.054** (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.024) (0.025) DID (Month> April=6) 0.130*** 0.098*** 0.142*** 0.107*** 0.034*** 0.017*** -0.090*** -0.065*** (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.023) (0.024) DID (Month> April=7) 0.183*** 0.176*** 0.198*** 0.190*** 0.062*** 0.062*** -0.064*** -0.071*** (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.022) (0.024) DID (Month> April=8) 0.278*** 0.278*** 0.439*** 0.440*** 0.111*** 0.114*** -0.044* -0.057** (0.010) (0.010) (0.022) (0.022) (0.011) (0.011) (0.023) (0.025) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.711 0.711 0.687 0.687 0.775 0.775 0.859 0.859 Mean dep. Var 383.115 383.115 123.781 123.781 419.210 419.210 152470.284 152470.284 Obs 41958788 41958777 37299880 37299869 4602780 4602780 52317 52317 Notes: Dependent variable is the log of monthly electricity consumption. DID (Month> April= ω ) represents the DID estimates for the ω months since the start of COVID lockdown (April, 2020). The DID estimates are derived from the expression DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k . Climate Ctrls include the monthly average temperature and total precipitation . Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 33 Table 8: COVID-19 and Electricity Consumption in Lockdown Communities: Ghana log of Consumption All Residential Non-Residential Heavy Industries (1) (2) (3) (4) (5) (6) (7) (8) DID (Month> April=0) 0.404*** 0.406*** 0.452*** 0.454*** 0.118*** 0.118*** -0.409*** -0.410*** (0.002) (0.002) (0.002) (0.002) (0.006) (0.006) (0.022) (0.023) DID (Month> April=1) 0.149*** 0.160*** 0.173*** 0.185*** 0.015*** 0.020*** -0.180*** -0.181*** (0.002) (0.002) (0.002) (0.002) (0.006) (0.006) (0.023) (0.023) DID (Month> April=2) 0.173*** 0.146*** 0.193*** 0.164*** 0.053*** 0.038*** -0.100*** -0.099*** (0.002) (0.002) (0.002) (0.002) (0.006) (0.006) (0.027) (0.032) DID (Month> April=3) 0.142*** 0.174*** 0.155*** 0.191*** 0.072*** 0.088*** -0.112*** -0.106*** (0.002) (0.003) (0.003) (0.003) (0.007) (0.007) (0.025) (0.028) DID (Month> April=4) 0.113*** 0.068*** 0.128*** 0.078*** 0.035*** 0.009 -0.138*** -0.122*** (0.003) (0.003) (0.003) (0.003) (0.007) (0.007) (0.026) (0.028) DID (Month> April=5) 0.196*** 0.167*** 0.213*** 0.181*** 0.104*** 0.088*** -0.096*** -0.085*** (0.003) (0.003) (0.003) (0.003) (0.007) (0.007) (0.027) (0.028) DID (Month> April=6) 0.119*** 0.073*** 0.133*** 0.082*** 0.044*** 0.019*** -0.112*** -0.106*** (0.003) (0.003) (0.003) (0.003) (0.007) (0.007) (0.025) (0.027) DID (Month> April=7) 0.150*** 0.154*** 0.168*** 0.174*** 0.048*** 0.049*** -0.090*** -0.079*** (0.003) (0.003) (0.003) (0.003) (0.007) (0.007) (0.025) (0.026) DID (Month> April=8) 0.159*** 0.175*** 0.433*** 0.452*** 0.046*** 0.053*** -0.071*** -0.060** (0.013) (0.013) (0.063) (0.063) (0.014) (0.014) (0.026) (0.028) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.697 0.697 0.624 0.624 0.795 0.795 0.869 0.869 Mean dep. Var 1032.762 1032.762 224.075 224.075 763.197 763.197 143144.950 143144.950 Obs 8635711 8635711 7327395 7327395 1268209 1268209 38885 38885 Notes: Dependent variable is the log of monthly electricity consumption. DID (Month> April= ω ) represents the DID estimates for the ω months since the start of COVID lockdown (April, 2020). The DID estimates are derived from the expression DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k . Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 34 Table 9: COVID-19 and Electricity Consumption in Non-Lockdown Communities: Ghana log of Consumption All Residential Non-Residential Heavy Industries (1) (2) (3) (4) (5) (6) (7) (8) DID (Month> April=0) 0.438*** 0.439*** 0.460*** 0.462*** 0.224*** 0.225*** -0.229*** -0.224*** (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.052) (0.052) DID (Month> April=1) 0.179*** 0.186*** 0.201*** 0.208*** -0.027*** -0.025*** -0.098* -0.100** (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.051) (0.051) DID (Month> April=2) 0.161*** 0.142*** 0.178*** 0.158*** -0.001 -0.010*** -0.049 0.003 (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.050) (0.054) DID (Month> April=3) 0.142*** 0.149*** 0.155*** 0.163*** 0.019*** 0.025*** -0.054 -0.104** (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.047) (0.051) DID (Month> April=4) 0.145*** 0.106*** 0.153*** 0.111*** 0.076*** 0.062*** -0.021 -0.003 (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.052) (0.057) DID (Month> April=5) 0.198*** 0.169*** 0.211*** 0.179*** 0.086*** 0.075*** 0.001 0.014 (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.051) (0.054) DID (Month> April=6) 0.135*** 0.105*** 0.146*** 0.114*** 0.035*** 0.022*** -0.026 0.018 (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.050) (0.053) DID (Month> April=7) 0.191*** 0.182*** 0.204*** 0.194*** 0.068*** 0.067*** 0.010 -0.012 (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.050) (0.052) DID (Month> April=8) 0.372*** 0.369*** 0.473*** 0.471*** 0.200*** 0.202*** 0.030 -0.005 (0.015) (0.015) (0.025) (0.025) (0.018) (0.018) (0.051) (0.053) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.691 0.691 0.676 0.676 0.753 0.753 0.836 0.836 Mean dep. Var 214.740 214.740 99.257 99.257 288.256 288.256 179514.430 179514.430 Obs 33321054 33321054 29970907 29970907 3334122 3334122 13427 13427 Notes: Dependent variable is the log of monthly electricity consumption. DID (Month> April= ω ) represents the DID estimates for the ω months since the start of COVID lockdown (April, 2020). The DID estimates are derived from the expression DIDω = β ˆω − 1 2019 β ˆω 2020 2 k=2018 k . Climate Ctrls include the monthly average temperature and total precipitation . Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 35 Table 10: Electricity Consumption by Subsidy Rates in Ghana Residential Lifeline Customers (100% subsidy) Non-Lifeline Customers (50% subsidy) (1) (2) (3) (4) I (Apr-Dec 2020) 0.403*** 0.441*** (0.001) (0.001) I (Apr-Dec 2019) -0.096*** -0.053*** (0.001) (0.001) I (Apr-Dec 2018) -0.085*** -0.053*** (0.001) (0.001) I (Apr-Jun 2020) 0.106*** 0.135*** (0.001) (0.001) I (Apr-Jun 2019) 0.024*** 0.055*** (0.001) (0.001) I (Apr-Jun 2018) -0.032*** -0.003*** (0.001) (0.001) I (Jul-Dec 2020) -0.077*** 0.004*** (0.001) (0.001) I (Jul-Dec 2019) 0.051*** 0.130*** (0.001) (0.001) I (Jul-Dec 2018) -0.016*** 0.037*** (0.001) (0.001) DID (Apr-Dec) 0.408*** 0.494*** (0.001) (0.001) DID (Apr-Jun) 0.078*** 0.105*** (0.001) (0.001) DID (Jun-Dec) -0.110*** -0.043*** (0.001) (0.001) Climate Ctrls No Yes No Yes Customer FE Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes R-squared 0.542 0.543 0.525 0.526 Mean dep. Var 45.144 45.145 175.043 175.044 Obs 14595224 14595225 19028901 19028902 Notes: Dependent variable is the log of monthly electricity consumption. The DID estimates are derived from the expression: ˆω − 1 2019 β ˆω DIDω = β 2020 2 k=2018 k . Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 36 Table 11: Effects of Electricity Subsidies on Consumption in Ghana Residential Dep. var: Log of electricity consumption (1) (2) (3) (4) 1(non-lifeline)× postsubsidy -0.162*** -0.165*** -0.173*** -0.173*** (0.014) (0.014) (0.009) (0.010) Customer type Trends No Yes Yes No Climate Ctrls No Yes No No Customer FE Yes Yes Yes Yes Month FE Yes Yes No No District-Month FE No No Yes Yes R-squared 0.812 0.812 0.808 0.802 Mean dep. Var 45.144 45.145 175.043 175.044 Obs 9140330 9140330 9140330 9140330 Notes: Dependent variable is the log of monthly electricity consumption. Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district-customer level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 37 38 A ONLINE APPENDIX A.1 Tables Table A1: COVID-19 and Electricity Consumption in Rwanda log of Consumption All Residential Non-Residential Small& Medium Industries (1) (2) (3) (4) (5) (6) (7) (8) I(2020) x I (Month> March=0) 0.028*** 0.027*** 0.040*** 0.039*** -0.039*** -0.042*** -0.044 -0.063* (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.030) (0.033) I(2020) x I (Month> March=1) -0.022*** -0.028*** 0.019*** 0.013*** -0.290*** -0.299*** -0.146** -0.172*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.062) (0.064) I(2020) x I (Month> March=2) 0.009*** 0.007*** 0.037*** 0.036*** -0.140*** -0.140*** 0.087** 0.102*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.037) (0.039) I(2020) x I (Month> March=3) 0.007*** 0.005*** 0.030*** 0.028*** -0.112*** -0.112*** 0.130*** 0.156*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.034) (0.038) I(2020) x I (Month> March=4) 0.034*** 0.032*** 0.053*** 0.051*** -0.065*** -0.065*** 0.131*** 0.164*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.035) (0.040) I(2020) x I (Month> March=5) 0.047*** 0.063*** 0.065*** 0.081*** -0.058*** -0.041*** 0.135*** 0.154*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.035) (0.050) I(2020) x I (Month> March=6) 0.026*** 0.030*** 0.045*** 0.049*** -0.091*** -0.086*** 0.132*** 0.145*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.036) (0.038) I(2020) x I (Month> March=7) 0.051*** 0.042*** 0.066*** 0.057*** -0.038*** -0.046*** 0.104*** 0.106*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.036) (0.037) I(2020) x I (Month> March=8) 0.026*** 0.010*** 0.040*** 0.023*** -0.059*** -0.074*** 0.066* 0.062 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.035) (0.042) I(2020) x I (Month> March=9) 0.054*** 0.045*** 0.065*** 0.055*** -0.033*** -0.042*** 0.074** 0.084** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.034) (0.035) I(2019) x I (Month> March=0) 0.019*** 0.016*** 0.013*** 0.010*** 0.048*** 0.045*** 0.086*** 0.085*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.033) (0.033) I(2019) x I (Month> March=1) 0.007*** 0.004*** 0.007*** 0.005*** 0.010*** 0.007*** 0.132*** 0.120*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.037) (0.037) I(2019) x I (Month> March=2) 0.027*** 0.018*** 0.022*** 0.014*** 0.053*** 0.045*** 0.128*** 0.127*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.037) (0.039) I(2019) x I (Month> March=3) 0.015*** 0.011*** 0.014*** 0.010*** 0.027*** 0.025*** 0.085** 0.100*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.037) (0.039) I(2019) x I (Month> March=4) 0.042*** 0.045*** 0.040*** 0.042*** 0.067*** 0.071*** 0.076** 0.093** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.035) (0.038) I(2019) x I (Month> March=5) 0.027*** 0.038*** 0.025*** 0.036*** 0.051*** 0.062*** 0.042 0.050 (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.033) (0.038) I(2019) x I (Month> March=6) -0.013*** -0.008*** -0.014*** -0.009*** 0.002 0.007*** 0.024 0.025 (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.033) (0.034) I(2019) x I (Month> March=7) 0.006*** -0.011*** 0.005*** -0.013*** 0.027*** 0.008*** -0.016 -0.034 (0.001) (0.001) (0.001) (0.001) (0.002) (0.003) (0.035) (0.049) I(2019) x I (Month> March=8) -0.001 -0.012*** -0.000 -0.011*** 0.004* -0.007*** -0.062 -0.068 (0.001) (0.001) (0.001) (0.001) (0.002) (0.003) (0.038) (0.043) I(2019) x I (Month> March=9) 0.034*** 0.027*** 0.034*** 0.028*** 0.042*** 0.033*** -0.012 -0.037 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.039) (0.043) I(2018) x I (Month> March=0) 0.005*** -0.001 0.002*** -0.003*** 0.022*** 0.016*** 0.087** 0.080** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.038) (0.040) I(2018) x I (Month> March=1) -0.017*** -0.028*** -0.012*** -0.023*** -0.033*** -0.047*** 0.056 0.014 (0.001) (0.001) (0.001) (0.001) (0.002) (0.003) (0.045) (0.055) I(2018) x I (Month> March=2) -0.002*** -0.006*** -0.006*** -0.010*** 0.022*** 0.019*** 0.150*** 0.163*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.047) (0.049) I(2018) x I (Month> March=3) -0.006*** -0.005*** -0.007*** -0.007*** 0.010*** 0.013*** 0.117** 0.146*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.003) (0.051) (0.054) I(2018) x I (Month> March=4) 0.007*** 0.006*** 0.005*** 0.005*** 0.026*** 0.028*** 0.125** 0.156*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.051) (0.054) I(2018) x I (Month> March=5) 0.026*** 0.028*** 0.027*** 0.029*** 0.035*** 0.038*** 0.168*** 0.176*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.046) (0.046) I(2018) x I (Month> March=6) -0.023*** -0.014*** -0.022*** -0.013*** -0.016*** -0.008*** 0.106** 0.122** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.045) (0.048) I(2018) x I (Month> March=7) -0.013*** -0.007*** -0.014*** -0.008*** 0.008*** 0.013*** 0.148*** 0.147*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.046) (0.047) I(2018) x I (Month> March=8) -0.021*** -0.021*** -0.018*** -0.017*** -0.023*** -0.022*** 0.154*** 0.154*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.046) (0.046) I(2018) x I (Month> March=9) 0.027*** 0.019*** 0.034*** 0.026*** 0.015*** 0.008*** 0.204*** 0.205*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.048) (0.049) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.858 0.858 0.849 0.849 0.856 0.856 0.722 0.722 Mean dep. Var 27.552 27.552 20.426 20.426 67.526 67.526 256.741 256.741 Obs 21352142 21352142 17097530 17097530 3212773 3212773 13345 13345 Notes: Dependent variable is the log of monthly electricity consumption. Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level; ∗∗ Significant at 5 percent level; ∗∗∗ Significant at 1 percent level 39 Table A2: COVID-19 and Electricity Consumption in Non-Residential Sector in Rwanda log of Consumption Commercial Hotels Health Centers Public Works (1) (2) (3) (4) (5) (6) (7) (8) I(2020) x I (Month> March=0) -0.039*** -0.042*** -0.038* -0.041* 0.079*** 0.077*** 0.016 0.011 (0.002) (0.002) (0.023) (0.025) (0.020) (0.021) (0.032) (0.034) I(2020) x I (Month> March=1) -0.292*** -0.300*** -0.265*** -0.269*** -0.000 -0.013 -0.050 -0.056 (0.003) (0.003) (0.040) (0.042) (0.025) (0.026) (0.043) (0.050) I(2020) x I (Month> March=2) -0.141*** -0.141*** -0.173*** -0.171*** 0.031 0.029 -0.041 -0.037 (0.003) (0.003) (0.037) (0.038) (0.026) (0.027) (0.035) (0.032) I(2020) x I (Month> March=3) -0.113*** -0.113*** -0.128*** -0.125*** 0.047* 0.043 -0.013 -0.006 (0.003) (0.003) (0.033) (0.036) (0.027) (0.028) (0.035) (0.036) I(2020) x I (Month> March=4) -0.066*** -0.065*** -0.085*** -0.081** 0.060** 0.054** 0.004 0.013 (0.003) (0.003) (0.032) (0.036) (0.026) (0.027) (0.035) (0.036) I(2020) x I (Month> March=5) -0.058*** -0.041*** -0.080*** -0.075** 0.052** 0.085*** 0.026 0.029 (0.003) (0.003) (0.030) (0.033) (0.026) (0.032) (0.030) (0.043) I(2020) x I (Month> March=6) -0.092*** -0.087*** -0.110*** -0.107*** 0.080*** 0.087*** 0.019 0.022 (0.003) (0.003) (0.032) (0.032) (0.023) (0.024) (0.035) (0.038) I(2020) x I (Month> March=7) -0.039*** -0.047*** -0.041 -0.042 0.014 -0.004 -0.008 -0.006 (0.003) (0.003) (0.030) (0.033) (0.026) (0.027) (0.037) (0.038) I(2020) x I (Month> March=8) -0.059*** -0.075*** -0.052 -0.055 0.097*** 0.064** 0.037 0.038 (0.003) (0.003) (0.035) (0.041) (0.022) (0.025) (0.034) (0.043) I(2020) x I (Month> March=9) -0.034*** -0.042*** -0.027 -0.028 0.093*** 0.073*** -0.011 -0.008 (0.003) (0.003) (0.035) (0.036) (0.022) (0.023) (0.044) (0.046) I(2019) x I (Month> March=0) 0.049*** 0.046*** 0.021 0.020 -0.015 -0.021 0.021 0.021 (0.002) (0.002) (0.019) (0.019) (0.025) (0.025) (0.022) (0.021) I(2019) x I (Month> March=1) 0.010*** 0.007*** 0.017 0.015 0.005 -0.000 -0.050 -0.053 (0.002) (0.002) (0.021) (0.022) (0.022) (0.022) (0.039) (0.041) I(2019) x I (Month> March=2) 0.053*** 0.045*** -0.008 -0.009 0.019 0.003 0.019 0.021 (0.002) (0.002) (0.020) (0.022) (0.019) (0.020) (0.033) (0.034) I(2019) x I (Month> March=3) 0.027*** 0.025*** 0.034 0.035 -0.025 -0.034 0.036 0.041 (0.002) (0.002) (0.023) (0.024) (0.023) (0.023) (0.026) (0.028) I(2019) x I (Month> March=4) 0.067*** 0.071*** 0.060** 0.063** 0.004 0.007 0.027 0.030 (0.002) (0.002) (0.025) (0.026) (0.021) (0.022) (0.036) (0.039) I(2019) x I (Month> March=5) 0.051*** 0.062*** 0.050** 0.053* -0.042* -0.021 0.067* 0.067 (0.002) (0.002) (0.025) (0.028) (0.025) (0.026) (0.037) (0.042) I(2019) x I (Month> March=6) 0.002 0.007*** 0.022 0.023 -0.016 -0.007 0.033 0.033 (0.002) (0.002) (0.026) (0.026) (0.022) (0.023) (0.034) (0.036) I(2019) x I (Month> March=7) 0.027*** 0.008*** 0.056** 0.051 0.038* 0.001 -0.005 -0.007 (0.002) (0.003) (0.027) (0.035) (0.023) (0.028) (0.034) (0.049) I(2019) x I (Month> March=8) 0.004* -0.007*** 0.055* 0.051* -0.018 -0.041 -0.031 -0.030 (0.002) (0.003) (0.028) (0.031) (0.024) (0.026) (0.051) (0.053) I(2019) x I (Month> March=9) 0.042*** 0.033*** 0.076*** 0.072** -0.026 -0.041* -0.034 -0.040 (0.003) (0.003) (0.028) (0.032) (0.022) (0.024) (0.043) (0.050) I(2018) x I (Month> March=0) 0.022*** 0.016*** 0.037 0.035 0.015 0.003 -0.002 -0.004 (0.002) (0.002) (0.025) (0.025) (0.027) (0.029) (0.021) (0.023) I(2018) x I (Month> March=1) -0.033*** -0.047*** -0.007 -0.014 -0.013 -0.035 -0.076* -0.085 (0.002) (0.003) (0.026) (0.035) (0.024) (0.027) (0.043) (0.054) I(2018) x I (Month> March=2) 0.022*** 0.019*** 0.020 0.021 0.002 -0.005 0.053 0.057 (0.002) (0.002) (0.027) (0.027) (0.024) (0.025) (0.050) (0.053) I(2018) x I (Month> March=3) 0.010*** 0.013*** 0.003 0.006 -0.031 -0.029 0.063* 0.071** (0.002) (0.003) (0.034) (0.036) (0.028) (0.030) (0.035) (0.032) I(2018) x I (Month> March=4) 0.026*** 0.028*** 0.014 0.018 -0.006 -0.006 0.068* 0.076** (0.003) (0.003) (0.042) (0.043) (0.023) (0.025) (0.038) (0.036) I(2018) x I (Month> March=5) 0.035*** 0.038*** 0.006 0.008 0.050** 0.055** 0.055 0.057 (0.003) (0.003) (0.039) (0.040) (0.023) (0.024) (0.050) (0.047) I(2018) x I (Month> March=6) -0.017*** -0.008*** 0.003 0.007 0.077*** 0.096*** 0.046 0.048 (0.003) (0.003) (0.039) (0.043) (0.024) (0.026) (0.056) (0.051) I(2018) x I (Month> March=7) 0.008*** 0.013*** 0.035 0.035 0.026 0.039 -0.009 -0.010 (0.003) (0.003) (0.041) (0.043) (0.027) (0.027) (0.070) (0.069) I(2018) x I (Month> March=8) -0.023*** -0.023*** 0.062* 0.062* 0.023 0.026 0.059 0.059 (0.003) (0.003) (0.037) (0.037) (0.023) (0.023) (0.059) (0.058) I(2018) x I (Month> March=9) 0.015*** 0.008*** 0.035 0.033 0.016 -0.000 0.019 0.021 (0.003) (0.003) (0.040) (0.040) (0.026) (0.027) (0.065) (0.072) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.853 0.853 0.815 0.815 0.801 0.801 0.673 0.673 Mean dep. Var 65.244 65.244 40 353.152 353.152 359.381 359.381 398.729 398.729 Obs 3188121 3188121 9014 9014 12167 12167 3471 3471 Notes: Dependent variable is the log of monthly electricity consumption. Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level; ∗∗ Significant at 5 percent level; ∗∗∗ Significant at 1 percent level Table A3: COVID-19 and Electricity Consumption in Rwanda: Urban vs Rural Residential Urban Rural (1) (2) (3) (4) I(2020) x I (Month> March=0) 0.051*** 0.043*** 0.035*** 0.034*** (0.001) (0.001) (0.001) (0.001) I(2020) x I (Month> March=1) 0.038*** 0.017*** 0.010*** 0.005*** (0.001) (0.002) (0.001) (0.001) I(2020) x I (Month> March=2) 0.054*** 0.048*** 0.029*** 0.029*** (0.001) (0.001) (0.001) (0.001) I(2020) x I (Month> March=3) 0.044*** 0.039*** 0.024*** 0.024*** (0.001) (0.001) (0.001) (0.001) I(2020) x I (Month> March=4) 0.070*** 0.071*** 0.045*** 0.044*** (0.001) (0.001) (0.001) (0.001) I(2020) x I (Month> March=5) 0.091*** 0.139*** 0.053*** 0.065*** (0.001) (0.002) (0.001) (0.001) I(2020) x I (Month> March=6) 0.079*** 0.093*** 0.029*** 0.033*** (0.001) (0.001) (0.001) (0.001) I(2020) x I (Month> March=7) 0.102*** 0.077*** 0.049*** 0.044*** (0.001) (0.002) (0.001) (0.001) I(2020) x I (Month> March=8) 0.073*** 0.027*** 0.024*** 0.013*** (0.001) (0.002) (0.001) (0.001) I(2020) x I (Month> March=9) 0.099*** 0.074*** 0.049*** 0.043*** (0.001) (0.002) (0.001) (0.001) I(2019) x I (Month> March=0) 0.011*** 0.001 0.015*** 0.013*** (0.001) (0.001) (0.001) (0.001) I(2019) x I (Month> March=1) 0.019*** 0.009*** 0.001 -0.001 (0.001) (0.001) (0.001) (0.001) I(2019) x I (Month> March=2) 0.020*** -0.005*** 0.023*** 0.018*** (0.001) (0.001) (0.001) (0.001) I(2019) x I (Month> March=3) 0.011*** 0.006*** 0.015*** 0.013*** (0.001) (0.001) (0.001) (0.001) I(2019) x I (Month> March=4) 0.050*** 0.065*** 0.034*** 0.036*** (0.001) (0.001) (0.001) (0.001) I(2019) x I (Month> March=5) 0.038*** 0.073*** 0.018*** 0.025*** (0.001) (0.002) (0.001) (0.001) I(2019) x I (Month> March=6) 0.003** 0.019*** -0.023*** -0.020*** (0.001) (0.001) (0.001) (0.001) I(2019) x I (Month> March=7) 0.027*** -0.023*** -0.006*** -0.019*** (0.001) (0.002) (0.001) (0.001) I(2019) x I (Month> March=8) 0.038*** 0.008*** -0.020*** -0.028*** (0.001) (0.002) (0.001) (0.001) I(2019) x I (Month> March=9) 0.081*** 0.066*** 0.011*** 0.004*** (0.001) (0.002) (0.001) (0.001) I(2018) x I (Month> March=0) -0.003** -0.018*** 0.005*** 0.001 (0.001) (0.001) (0.001) (0.001) I(2018) x I (Month> March=1) -0.010*** -0.040*** -0.014*** -0.023*** (0.001) (0.002) (0.001) (0.001) I(2018) x I (Month> March=2) -0.011*** -0.023*** -0.003*** -0.005*** (0.001) (0.001) (0.001) (0.001) I(2018) x I (Month> March=3) -0.016*** -0.014*** -0.002* -0.001 (0.001) (0.002) (0.001) (0.001) I(2018) x I (Month> March=4) 0.004*** 0.008*** 0.005*** 0.006*** (0.001) (0.002) (0.001) (0.001) I(2018) x I (Month> March=5) 0.028*** 0.037*** 0.026*** 0.028*** (0.001) (0.002) (0.001) (0.001) I(2018) x I (Month> March=6) -0.019*** 0.000 -0.024*** -0.017*** (0.002) (0.002) (0.001) (0.001) I(2018) x I (Month> March=7) -0.003** 0.012*** -0.021*** -0.017*** (0.002) (0.002) (0.001) (0.001) I(2018) x I (Month> March=8) -0.004** -0.006*** -0.026*** -0.025*** (0.002) (0.002) (0.001) (0.001) I(2018) x I (Month> March=9) 0.060*** 0.039*** 0.019*** 0.014*** (0.002) (0.002) (0.001) (0.001) Climate Ctrls No Yes No Yes Customer FE Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes R-squared 0.821 0.821 0.793 0.793 Mean dep. Var 37.415 37.415 11.670 11.670 Obs 5815537 5815537 11279386 11279386 41 Notes: Dependent variable is the log of monthly electricity consumption. Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level Table A4: COVID-19 and Electricity Consumption in Ghana log of Consumption All Residential Non-Residential Heavy Industries (1) (2) (3) (4) (5) (6) (7) (8) I(2020) x I (Month> April=0) 0.416*** 0.434*** 0.444*** 0.463*** 0.175*** 0.183*** -0.377*** -0.386*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.017) (0.017) I(2020) x I (Month> April=1) 0.167*** 0.196*** 0.188*** 0.219*** -0.016*** 0.001 -0.157*** -0.196*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.017) (0.023) I(2020) x I (Month> April=2) 0.101*** 0.119*** 0.119*** 0.137*** -0.055*** -0.031*** -0.194*** -0.268*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.005) (0.020) (0.043) I(2020) x I (Month> April=3) 0.107*** 0.100*** 0.122*** 0.112*** -0.017*** 0.002 -0.186*** -0.268*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.006) (0.019) (0.053) I(2020) x I (Month> April=4) 0.094*** 0.018*** 0.104*** 0.020*** 0.012*** -0.003 -0.199*** -0.220*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.005) (0.020) (0.046) I(2020) x I (Month> April=5) 0.113*** 0.084*** 0.125*** 0.092*** 0.013*** 0.019*** -0.167*** -0.219*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.005) (0.019) (0.045) I(2020) x I (Month> April=6) 0.141*** 0.125*** 0.151*** 0.132*** 0.060*** 0.066*** -0.069*** -0.115*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.004) (0.019) (0.036) I(2020) x I (Month> April=7) 0.160*** 0.150*** 0.170*** 0.159*** 0.068*** 0.067*** -0.022 -0.031 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.019) (0.021) I(2020) x I (Month> April=8) 0.128*** 0.115*** 0.165*** 0.153*** 0.047*** 0.044*** -0.051*** -0.052** (0.008) (0.008) (0.019) (0.019) (0.008) (0.008) (0.019) (0.020) I(2019) x I (Month> April=0) 0.031*** 0.047*** 0.032*** 0.049*** 0.022*** 0.029*** -0.037** -0.047*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.016) (0.016) I(2019) x I (Month> April=1) 0.021*** 0.041*** 0.021*** 0.041*** 0.021*** 0.035*** 0.011 -0.023 (0.001) (0.001) (0.001) (0.001) (0.002) (0.003) (0.017) (0.021) I(2019) x I (Month> April=2) -0.067*** -0.037*** -0.065*** -0.035*** -0.080*** -0.048*** -0.118*** -0.219*** (0.001) (0.002) (0.001) (0.002) (0.002) (0.006) (0.019) (0.052) I(2019) x I (Month> April=3) -0.007*** -0.028*** -0.004*** -0.029*** -0.026*** -0.016*** -0.071*** -0.129*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.005) (0.018) (0.043) I(2019) x I (Month> April=4) -0.037*** -0.083*** -0.035*** -0.087*** -0.051*** -0.053*** -0.102*** -0.141*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.005) (0.018) (0.042) I(2019) x I (Month> April=5) -0.061*** -0.063*** -0.061*** -0.065*** -0.062*** -0.044*** -0.096*** -0.169*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.005) (0.019) (0.047) I(2019) x I (Month> April=6) 0.032*** 0.063*** 0.032*** 0.063*** 0.030*** 0.067*** 0.011 -0.096* (0.001) (0.002) (0.001) (0.002) (0.003) (0.006) (0.019) (0.055) I(2019) x I (Month> April=7) -0.048*** -0.059*** -0.051*** -0.063*** -0.024*** -0.026*** 0.035* 0.031 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.018) (0.019) I(2019) x I (Month> April=8) -0.009 -0.027*** -0.080*** -0.101*** -0.001 -0.009 -0.006 0.006 (0.007) (0.007) (0.015) (0.015) (0.008) (0.008) (0.019) (0.020) I(2018) x I (Month> April=0) -0.059*** -0.042*** -0.058*** -0.040*** -0.063*** -0.055*** 0.005 -0.010 (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.016) (0.017) I(2018) x I (Month> April=1) -0.033*** -0.008*** -0.035*** -0.009*** -0.017*** -0.000 -0.008 -0.050* (0.001) (0.001) (0.001) (0.001) (0.002) (0.003) (0.021) (0.028) I(2018) x I (Month> April=2) -0.054*** -0.008*** -0.055*** -0.007*** -0.047*** -0.010** -0.095*** -0.213*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.005) (0.020) (0.057) I(2018) x I (Month> April=3) -0.062*** -0.076*** -0.061*** -0.078*** -0.067*** -0.058*** -0.107*** -0.160*** (0.001) (0.001) (0.001) (0.002) (0.003) (0.005) (0.021) (0.043) I(2018) x I (Month> April=4) -0.051*** -0.077*** -0.051*** -0.081*** -0.050*** -0.046*** -0.084*** -0.122*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.005) (0.020) (0.040) I(2018) x I (Month> April=5) -0.104*** -0.101*** -0.106*** -0.106*** -0.084*** -0.067*** -0.097*** -0.161*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.005) (0.020) (0.042) I(2018) x I (Month> April=6) -0.011*** -0.009*** -0.015*** -0.014*** 0.022*** 0.030*** 0.031 -0.003 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.019) (0.026) I(2018) x I (Month> April=7) 0.001 0.006*** -0.003*** 0.002* 0.035*** 0.036*** 0.050** 0.051** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.020) (0.020) I(2018) x I (Month> April=8) -0.292*** -0.299*** -0.468*** -0.475*** -0.126*** -0.131*** -0.008 0.006 (0.010) (0.010) (0.017) (0.017) (0.013) (0.013) (0.018) (0.019) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.711 0.711 0.687 0.687 0.775 0.775 0.859 0.859 Mean dep. Var 383.115 383.115 123.781 123.781 419.210 419.210 152470.284 152470.284 Obs 41958788 41958777 37299880 37299869 4602780 4602780 52317 52317 Notes: Dependent variable is the log of monthly electricity consumption. Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 42 Table A5: COVID-19 and Electricity Consumption in Lockdown Communities Ghana log of Consumption All Residential Non-Residential Heavy Industries (1) (2) (3) (4) (5) (6) (7) (8) I(2020) x I (Month> April=0) 0.414*** 0.440*** 0.464*** 0.493*** 0.118*** 0.131*** -0.414*** -0.420*** (0.002) (0.002) (0.002) (0.002) (0.006) (0.006) (0.018) (0.018) I(2020) x I (Month> April=1) 0.160*** 0.221*** 0.182*** 0.250*** 0.034*** 0.066*** -0.168*** -0.165*** (0.002) (0.002) (0.002) (0.003) (0.005) (0.007) (0.018) (0.024) I(2020) x I (Month> April=2) 0.138*** 0.219*** 0.159*** 0.252*** 0.018*** 0.058*** -0.203*** -0.172*** (0.002) (0.004) (0.002) (0.004) (0.005) (0.010) (0.023) (0.046) I(2020) x I (Month> April=3) 0.128*** 0.192*** 0.143*** 0.218*** 0.046*** 0.076*** -0.207*** -0.156*** (0.002) (0.005) (0.002) (0.005) (0.006) (0.012) (0.021) (0.056) I(2020) x I (Month> April=4) 0.105*** 0.067*** 0.121*** 0.082*** 0.024*** -0.002 -0.237*** -0.173*** (0.002) (0.004) (0.002) (0.004) (0.006) (0.011) (0.023) (0.050) I(2020) x I (Month> April=5) 0.152*** 0.182*** 0.168*** 0.204*** 0.069*** 0.080*** -0.193*** -0.142*** (0.002) (0.004) (0.002) (0.004) (0.006) (0.011) (0.022) (0.049) I(2020) x I (Month> April=6) 0.159*** 0.187*** 0.170*** 0.203*** 0.098*** 0.110*** -0.083*** -0.048 (0.002) (0.003) (0.002) (0.003) (0.006) (0.009) (0.022) (0.039) I(2020) x I (Month> April=7) 0.156*** 0.161*** 0.169*** 0.176*** 0.082*** 0.083*** -0.038* -0.025 (0.002) (0.002) (0.002) (0.002) (0.006) (0.006) (0.021) (0.024) I(2020) x I (Month> April=8) 0.157*** 0.152*** 0.356*** 0.352*** 0.073*** 0.069*** -0.075*** -0.065*** (0.011) (0.011) (0.061) (0.061) (0.011) (0.011) (0.022) (0.023) I(2019) x I (Month> April=0) 0.050*** 0.071*** 0.055*** 0.078*** 0.026*** 0.038*** -0.012 -0.018 (0.002) (0.002) (0.002) (0.002) (0.004) (0.004) (0.016) (0.016) I(2019) x I (Month> April=1) 0.043*** 0.090*** 0.043*** 0.096*** 0.041*** 0.065*** 0.024 0.029 (0.002) (0.002) (0.002) (0.002) (0.004) (0.006) (0.017) (0.022) I(2019) x I (Month> April=2) -0.041*** 0.063*** -0.037*** 0.080*** -0.057*** -0.005 -0.114*** -0.078 (0.002) (0.004) (0.002) (0.005) (0.005) (0.011) (0.018) (0.054) I(2019) x I (Month> April=3) 0.018*** 0.050*** 0.023*** 0.062*** -0.009* 0.005 -0.079*** -0.029 (0.002) (0.004) (0.002) (0.004) (0.005) (0.011) (0.018) (0.046) I(2019) x I (Month> April=4) 0.005*** 0.003 0.011*** 0.012*** -0.022*** -0.028*** -0.120*** -0.066 (0.002) (0.004) (0.002) (0.004) (0.005) (0.010) (0.019) (0.045) I(2019) x I (Month> April=5) -0.030*** 0.032*** -0.027*** 0.045*** -0.044*** -0.015 -0.091*** -0.046 (0.002) (0.004) (0.002) (0.004) (0.005) (0.011) (0.020) (0.049) I(2019) x I (Month> April=6) 0.050*** 0.175*** 0.052*** 0.194*** 0.037*** 0.100*** 0.025 0.067 (0.002) (0.005) (0.002) (0.006) (0.005) (0.013) (0.019) (0.058) I(2019) x I (Month> April=7) 0.014*** 0.011*** 0.013*** 0.010*** 0.024*** 0.022*** 0.056*** 0.065*** (0.002) (0.002) (0.002) (0.002) (0.006) (0.006) (0.018) (0.019) I(2019) x I (Month> April=8) -0.019** -0.043*** -0.140*** -0.168*** 0.007 -0.006 0.006 0.009 (0.009) (0.009) (0.019) (0.019) (0.011) (0.011) (0.021) (0.022) I(2018) x I (Month> April=0) -0.030*** -0.003* -0.031*** -0.001 -0.027*** -0.012*** 0.002 -0.003 (0.002) (0.002) (0.002) (0.002) (0.004) (0.004) (0.017) (0.017) I(2018) x I (Month> April=1) -0.020*** 0.032*** -0.024*** 0.035*** -0.002 0.025*** -0.001 0.004 (0.002) (0.002) (0.002) (0.002) (0.004) (0.006) (0.022) (0.030) I(2018) x I (Month> April=2) -0.029*** 0.084*** -0.032*** 0.095*** -0.014*** 0.045*** -0.092*** -0.068 (0.002) (0.004) (0.002) (0.004) (0.004) (0.011) (0.021) (0.062) I(2018) x I (Month> April=3) -0.047*** -0.014*** -0.047*** -0.008** -0.042*** -0.028*** -0.111*** -0.072 (0.002) (0.003) (0.002) (0.004) (0.005) (0.009) (0.023) (0.045) I(2018) x I (Month> April=4) -0.021*** -0.004 -0.025*** -0.002 0.000 0.006 -0.079*** -0.037 (0.002) (0.003) (0.002) (0.004) (0.005) (0.009) (0.021) (0.041) I(2018) x I (Month> April=5) -0.059*** -0.002 -0.064*** 0.001 -0.027*** -0.000 -0.102*** -0.066 (0.002) (0.004) (0.002) (0.004) (0.005) (0.009) (0.022) (0.045) I(2018) x I (Month> April=6) 0.029*** 0.053*** 0.022*** 0.049*** 0.070*** 0.081*** 0.033 0.049* (0.002) (0.002) (0.002) (0.003) (0.005) (0.006) (0.020) (0.028) I(2018) x I (Month> April=7) -0.001 0.002 -0.010*** -0.006*** 0.044*** 0.046*** 0.048** 0.043** (0.002) (0.002) (0.002) (0.002) (0.005) (0.005) (0.021) (0.020) I(2018) x I (Month> April=8) 0.015 -0.002 -0.013 -0.032* 0.047*** 0.038*** -0.014 -0.019 (0.010) (0.010) (0.018) (0.018) (0.013) (0.013) (0.020) (0.021) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.697 0.697 0.624 0.624 0.795 0.795 0.869 0.869 Mean dep. Var 1032.762 1032.762 224.075 224.075 763.197 763.197 143144.950 143144.950 Obs 8635711 8635711 43 7327395 7327395 1268209 1268209 38885 38885 Notes: Dependent variable is the log of monthly electricity consumption. Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level Table A6: COVID-19 and Electricity Consumption in Non-Lockdown Communities in Ghana log of Consumption All Residential Non-Residential Heavy Industries I(2020) x I (Month> April=0) 0.417*** 0.434*** 0.440*** 0.458*** 0.195*** 0.201*** -0.275*** -0.288*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.042) (0.042) I(2020) x I (Month> April=1) 0.169*** 0.192*** 0.190*** 0.215*** -0.032*** -0.019*** -0.127*** -0.204*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.004) (0.042) (0.050) I(2020) x I (Month> April=2) 0.093*** 0.100*** 0.111*** 0.117*** -0.077*** -0.061*** -0.167*** -0.327*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.006) (0.043) (0.078) I(2020) x I (Month> April=3) 0.103*** 0.082*** 0.118*** 0.094*** -0.036*** -0.025*** -0.126*** -0.325*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.007) (0.039) (0.095) I(2020) x I (Month> April=4) 0.092*** 0.008*** 0.101*** 0.009*** 0.008** -0.008 -0.094** -0.198** (0.001) (0.002) (0.001) (0.002) (0.003) (0.006) (0.041) (0.087) I(2020) x I (Month> April=5) 0.105*** 0.064*** 0.117*** 0.072*** -0.004 -0.003 -0.095** -0.240*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.006) (0.037) (0.079) I(2020) x I (Month> April=6) 0.137*** 0.113*** 0.147*** 0.120*** 0.048*** 0.050*** -0.029 -0.142** (0.001) (0.001) (0.001) (0.002) (0.003) (0.005) (0.038) (0.065) I(2020) x I (Month> April=7) 0.160*** 0.148*** 0.171*** 0.157*** 0.063*** 0.062*** 0.025 -0.005 (0.001) (0.001) (0.001) (0.001) (0.003) (0.004) (0.039) (0.044) I(2020) x I (Month> April=8) 0.108*** 0.095*** 0.130*** 0.116*** 0.038*** 0.035*** 0.018 0.004 (0.011) (0.011) (0.019) (0.019) (0.012) (0.012) (0.040) (0.041) I(2019) x I (Month> April=0) 0.025*** 0.041*** 0.026*** 0.043*** 0.021*** 0.026*** -0.108** -0.119*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.003) (0.043) (0.043) I(2019) x I (Month> April=1) 0.016*** 0.030*** 0.016*** 0.031*** 0.014*** 0.024*** -0.028 -0.094** (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.044) (0.048) I(2019) x I (Month> April=2) -0.073*** -0.057*** -0.072*** -0.056*** -0.089*** -0.066*** -0.133** -0.339*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.006) (0.055) (0.094) I(2019) x I (Month> April=3) -0.013*** -0.045*** -0.011*** -0.047*** -0.032*** -0.027*** -0.050 -0.211** (0.001) (0.002) (0.001) (0.002) (0.003) (0.006) (0.048) (0.083) I(2019) x I (Month> April=4) -0.048*** -0.102*** -0.046*** -0.107*** -0.061*** -0.066*** -0.049 -0.173** (0.001) (0.002) (0.001) (0.002) (0.003) (0.006) (0.043) (0.079) I(2019) x I (Month> April=5) -0.069*** -0.083*** -0.069*** -0.086*** -0.069*** -0.058*** -0.110** -0.279*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.006) (0.046) (0.087) I(2019) x I (Month> April=6) 0.027*** 0.042*** 0.027*** 0.041*** 0.027*** 0.052*** -0.028 -0.251*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.007) (0.049) (0.096) I(2019) x I (Month> April=7) -0.063*** -0.076*** -0.066*** -0.079*** -0.041*** -0.043*** -0.027 -0.044 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.047) (0.048) I(2019) x I (Month> April=8) 0.043*** 0.027** 0.054** 0.036 0.004 -0.002 -0.039 -0.022 (0.011) (0.011) (0.024) (0.024) (0.012) (0.012) (0.045) (0.046) I(2018) x I (Month> April=0) -0.068*** -0.053*** -0.066*** -0.050*** -0.080*** -0.074*** 0.014 -0.010 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) (0.041) (0.042) I(2018) x I (Month> April=1) -0.037*** -0.016*** -0.039*** -0.017*** -0.024*** -0.011*** -0.031 -0.113* (0.001) (0.001) (0.001) (0.001) (0.003) (0.004) (0.054) (0.064) I(2018) x I (Month> April=2) -0.062*** -0.027*** -0.062*** -0.026*** -0.062*** -0.035*** -0.103** -0.322*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.006) (0.052) (0.096) I(2018) x I (Month> April=3) -0.066*** -0.089*** -0.065*** -0.091*** -0.079*** -0.074*** -0.094* -0.231*** (0.001) (0.002) (0.001) (0.002) (0.003) (0.005) (0.052) (0.086) I(2018) x I (Month> April=4) -0.060*** -0.095*** -0.058*** -0.098*** -0.074*** -0.074*** -0.098* -0.218*** (0.001) (0.002) (0.001) (0.002) (0.004) (0.006) (0.052) (0.082) I(2018) x I (Month> April=5) -0.118*** -0.126*** -0.119*** -0.129*** -0.111*** -0.100*** -0.083* -0.230*** (0.001) (0.002) (0.001) (0.002) (0.004) (0.006) (0.047) (0.080) I(2018) x I (Month> April=6) -0.024*** -0.026*** -0.026*** -0.029*** -0.001 0.005 0.022 -0.069 (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.045) (0.056) I(2018) x I (Month> April=7) 0.002* 0.007*** -0.001 0.004*** 0.032*** 0.033*** 0.057 0.058 (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.051) (0.051) I(2018) x I (Month> April=8) -0.570*** -0.575*** -0.741*** -0.746*** -0.328*** -0.332*** 0.014 0.039 (0.017) (0.017) (0.024) (0.024) (0.024) (0.024) (0.040) (0.042) Climate Ctrls No Yes No Yes No Yes No Yes Customer FE Yes Yes Yes Yes Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.691 0.691 0.676 0.676 0.753 0.753 0.836 0.836 Mean dep. Var 214.740 214.740 99.3 99.257 288.256 288.256 179514.430 179514.430 Obs 33321054 33321054 29970907 29970907 3334122 3334122 13427 13427 44 Notes: Dependent variable is the log of monthly electricity consumption. Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level Table A7: Effects of Electricity Subsidies on Consumption in Ghana Lifeline Customers (100% subsidy) Non-Lifeline Customers (50% subsidy) (1) (2) (3) (4) I(2020) x I (Month> April=0) 0.597*** 0.627*** 0.316*** 0.328*** (0.002) (0.002) (0.001) (0.001) I(2020) x I (Month> April=1) 0.358*** 0.366*** 0.053*** 0.095*** (0.001) (0.002) (0.001) (0.001) I(2020) x I (Month> April=2) 0.310*** 0.221*** -0.039*** 0.036*** (0.002) (0.003) (0.001) (0.002) I(2020) x I (Month> April=3) 0.348*** 0.172*** -0.063*** 0.017*** (0.002) (0.003) (0.001) (0.002) I(2020) x I (Month> April=4) 0.355*** 0.089*** -0.102*** -0.081*** (0.002) (0.003) (0.001) (0.002) I(2020) x I (Month> April=5) 0.388*** 0.196*** -0.089*** -0.034*** (0.002) (0.003) (0.001) (0.002) I(2020) x I (Month> April=6) 0.419*** 0.294*** -0.066*** -0.025*** (0.002) (0.002) (0.001) (0.002) I(2020) x I (Month> April=7) 0.466*** 0.417*** -0.066*** -0.056*** (0.002) (0.002) (0.001) (0.001) I(2020) x I (Month> April=8) 0.454*** 0.416*** -0.018 -0.015 (0.033) (0.033) (0.024) (0.024) I(2019) x I (Month> April=0) 0.002* 0.034*** 0.055*** 0.064*** (0.001) (0.001) (0.001) (0.001) I(2019) x I (Month> April=1) -0.020*** -0.028*** 0.050*** 0.085*** (0.001) (0.002) (0.001) (0.001) I(2019) x I (Month> April=2) -0.112*** -0.205*** -0.035*** 0.060*** (0.001) (0.003) (0.001) (0.002) I(2019) x I (Month> April=3) -0.060*** -0.239*** 0.029*** 0.089*** (0.002) (0.003) (0.001) (0.002) I(2019) x I (Month> April=4) -0.100*** -0.310*** 0.009*** 0.048*** (0.002) (0.003) (0.001) (0.002) I(2019) x I (Month> April=5) -0.159*** -0.304*** 0.009*** 0.081*** (0.002) (0.003) (0.001) (0.002) I(2019) x I (Month> April=6) -0.092*** -0.210*** 0.128*** 0.237*** (0.002) (0.004) (0.001) (0.003) I(2019) x I (Month> April=7) -0.223*** -0.262*** 0.078*** 0.081*** (0.002) (0.002) (0.001) (0.001) I(2019) x I (Month> April=8) -0.171*** -0.205*** -0.051*** -0.063*** (0.032) (0.032) (0.019) (0.019) I(2018) x I (Month> April=0) -0.085*** -0.059*** -0.041*** -0.027*** (0.001) (0.001) (0.001) (0.001) I(2018) x I (Month> April=1) -0.061*** -0.059*** -0.016*** 0.025*** (0.002) (0.002) (0.001) (0.001) I(2018) x I (Month> April=2) -0.080*** -0.120*** -0.039*** 0.056*** (0.002) (0.003) (0.001) (0.002) I(2018) x I (Month> April=3) -0.084*** -0.223*** -0.044*** 0.007*** (0.002) (0.003) (0.001) (0.002) I(2018) x I (Month> April=4) -0.081*** -0.242*** -0.034*** 0.010*** (0.002) (0.003) (0.001) (0.002) I(2018) x I (Month> April=5) -0.149*** -0.263*** -0.077*** -0.013*** (0.002) (0.003) (0.001) (0.002) I(2018) x I (Month> April=6) -0.072*** -0.121*** 0.027*** 0.058*** (0.002) (0.002) (0.001) (0.002) I(2018) x I (Month> April=7) -0.068*** -0.051*** 0.050*** 0.050*** (0.002) (0.002) (0.001) (0.001) I(2018) x I (Month> April=8) -0.487*** -0.483*** -0.642*** -0.652*** (0.042) (0.042) (0.024) (0.024) Climate Ctrls No Yes No Yes Customer FE Yes Yes Yes Yes District X Year FE Yes Yes Yes Yes R-squared 0.544 0.544 0.528 0.528 Mean dep. Var 45.144 45.144 175.043 175.043 Obs 14595224 14595224 19028901 19028901 Notes: Dependent variable is the log of monthly electricity consumption. Climate Ctrls include the monthly average temperature and total precipitation. Standard errors clustered at district level in parenthesis. 45 ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level