POVERTY & EQUITY NOTES M AY 2020 · NUMBER 21 How is the COVID-19 crisis affecting Nitrogen Dioxide emissions in Sub-Saharan Africa? Takaaki Masaki, Shohei Nakamura, and David Newhouse 1 Daily estimates of Nitrogen Dioxide (NO2)—one of a group of pollutant gases that derive from human activity, primarily combustion—can provide a useful, albeit “noisy”, real-time proxy measure of how COVID-19 has affected economic activity. To examine how the crisis affects NO2 levels in Sub-Saharan Africa (SSA), we calculated mean daily NO2 tropospheric estimates for 43 regions that contain either national capitals or large urban areas for the period July 2018 to April 2020. We found that emissions reductions in April were sizeable, larger in more developed areas, and large in many but not all of the countries that implemented national lockdowns. Overall, the data suggest that NO2 readings contain useful information to help track the reduction in motorized transport, and possibly overall economic activity, in urban, developed areas. Tracking changes in NO2 may be able to shed in 43 SSA subnational regions containing either light in close to real time on how COVID-19 national capitals or the largest cities in their reactions affect economic activity. NO2 is one of countries. It also examines how NO2 evolved in a group of gases called nitrogen oxides (NOx) March and April of 2020 in the aftermath of the produced as a byproduct of anthropogenic global spread of the virus, compared to all previous sources, primarily combustion, which is in turn months since July 2018. A reduction may provide a indicative of economic activity. Decreases in general indication of falling economic activity. atmospheric NO2 may proxy for how lockdown policies, the general fear of contracting Coronavirus, and the resulting economic slowdown Methodology is translating into a reduction in motorized transport and economic activity. 2 NO2 has the To examine how the crisis is impacting NO2 additional feature of lasting for a day or less. levels in SSA, we calculated mean daily NO2 tropospheric estimates for 43 regions that Decreases in NO2 in the aftermath of COVID-19 contain either national capitals or large urban have been well-documented in Hubei province, areas. 3 We derived the data covering the period where Wuhan is located, and Northwest Italy. from July 2018 to April 2020 from Google Earth NO2 readings, however, are also heavily influenced Engine’s L3 composite, which filters out pixels with by cloud cover, wind, and weather patterns, making cloud coverage greater than 75 percent. 4 The daily changes highly variable and noisy. This note Sentinel 5 Precursor satellite launched by the explores whether a reduction in NO2 since the European Space Agency in October 2017 provided onset of the Coronavirus in March 2020, is apparent 1 Key inputs for this note were provided by Javier Baez, Andres Chamorro, Keith Garrett, Varun Kshirgar, and Emmanuel Skoufias. 2 See Morris and Zhang 2019 and Montgomery and Holloway 2018. 3 See https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S5P_NRTI_L3_O3. 4 This dropped 344, or 1.5 percent, of the total observations of NO2. the original NO2 estimates. We remove outlier substantial in many areas, and these decreases values for NO2, defined as values of less than 20 were not observed in 2019 when emissions on percent or greater than 5 times the median value. 5 average rose 6 percent in March and 3 percent in April. To assess the magnitude of NO2 reductions in March and April 2020, we performed Reductions were generally larger in more multivariate regression analysis using the developed areas and barely noticeable in some specification employed by Baez, Kshirsagar and poorer areas. The large reductions in Skoufias (2020). Figure 1 displays the estimated Johannesburg, South Africa and Gaborone, changes in NO2 in March and April 2020 compared Botswana may not entirely be attributable to the to all previous months since July 2018, while Figure effects of the Coronavirus, since there were also 2 displays the results for a placebo test for March sizeable reductions in April 2019 (Figure 2). 7 and April 2019, for the region containing the capital However, NO2 reductions were much larger in both city in each country. Countries that announced places in 2020 than in 2019, and large declines in official lockdown policies are starred. 6 April occurred in several more developed urban areas such as Nairobi (Kenya), Kigali (Rwanda), Lagos (Nigeria), Accra (Ghana), and Dar-es-Salaam Results (Tanzania). Meanwhile, less developed areas such as Juba (South Sudan), Antananarivo (Madagascar), and Ouagadougou (Burkina Faso) saw no We highlight three key findings: noticeable fall in NO2 emissions. Emissions reductions in April were sizeable. The Emissions reductions were large in many but fall in 2020 was negative and statistically significant not all of the countries that implemented in 8 of the 43 regions in March and 11 of the national lockdown policies. Substantial regions in April. In contrast, a statistically significant reductions in April occurred in Kigali, rise in NO2 was only observed in 3 of the regions Johannesburg, Gaborone, Lagos, and Greater in both March and April. Taking a simple average Accra. The exceptions where NO2 fell only slightly across the 43 capital regions, emissions fell despite the adoption of lockdowns are Greater approximately 5 percent in March 2020 and 11 Monrovia (Liberia), Antananarivo (Madagascar), percent in April. This is significantly less than both Freetown (Sierra Leone), and Harare (Zimbabwe). In the 20 percent reduction observed in Hubei these places, the announced lockdown may have province in March, and the larger reductions led to little change in the use of motorized observed in Northwest province of Italy in both transport, there was little motorized transport prior March and April. Nonetheless, the fall in April was 5 More formally, we estimate the following equation: to heteroscedasticity and serial correlation when estimating ln 2 = + 1 ℎ + 2 + 3 + 3 + standard errors. As a placebo, we replace ℎ as dummies 4 ln where NO2d represents the mean NO2 reading, for March and April 2019 to evaluate significant NO2 reduction over the region, for day d as measured by the number of days in previous year. since January 1. Month represents a vector of two dummy 6 Partial lockdowns, such as those implemented in Kenya, are variables indicating March and April 2020 (post-COVID crisis), not counted. so 1 contains the two main coefficients of interest and the 7 In Johannesburg, the difference between the 2020 and 2019 remaining variables are controls. Sin and Cosine represent estimates is approximately 0.66, which remains large and vectors of the first 3 powers of the sine and cosine of the day, statistically significant at the 5 percent level. In Gaborone, the which are included to control for natural oscillations in NO2. analogous difference is approximately 0.33, which is not Day represents 6 dummy variables indicating the day of the statistically significant. week. The log of the number of days since July 1, 2018 is also included to control for a smooth time trend. The regression is estimated with Newey-West standard errors, which are robust May 2020 · Number 21 2 to the crisis, or the area of interest includes a large In some cases, NO2 reductions may be result from portion of rural areas that masked reductions in a decline in the volume of traffic, for example, while urban areas. 8 others may stem from the termination of factory production. Monthly data on night-time light Overall, the data suggest that NO2 readings luminosity, when it becomes available, is also worth collected by Sentinel 5P contain useful information investigating as an alternative indicator of to help track the reduction in motorized transport, economic activity. More detailed analysis can also and possibly overall economic activity, in urban shed light on the effect of adopting lockdown areas due to the Coronavirus, particularly in more policies on emissions. developed areas. Finally, it is worth emphasizing that the data do not currently exist to draw any conclusion Caveats and Next Steps for Research about the relationship between reduced NO2 and a decline in economic activity or welfare. Satellite-based NO2 readings do not extend before Several unanswered questions call for further summer 2018 and are therefore insufficient to research. First, it remains unclear how exogenous estimate the economic effect of a crisis of this natural factors like wind and weather may influence magnitude. The planned collection of extensive daily NO2 readings, and explicitly controlling for phone survey data in the coming months will these factors may yield more precise estimate of provide a valuable opportunity to link these and correlation of COVID-19 with NO2. Second, it other high-frequency indicators to standard would be useful to better understand, if possible, economic indicators and better track the pace of how different types of economic activities are the recovery in SSA and beyond. contributing to the reduction in NO2. Figure 1: Estimated change in log NO2 for SSA capital regions for March and April 2020, relative to baseline of Jan 1 to Feb 29  March  April Note: This figure plots estimated changes in NO2 in March and April 2020, relative to the period from July 2018 to February 2020, as described in footnote 6. 8 No shapefile could be obtained for Harare, Zimbabwe, so the entire country was used. May 2020 · Number 21 3 Figure 2: Placebo test of estimated change in log NO2 for SSA capital regions for March and April 2019, relative to baseline of Jan 1 to Feb 28  March  April Note: This figure plots estimated changes in NO2 in March and April 2019, relative to all other months from July 2018 to April 2020, as described in footnote 6. ABOUT THE AUTHORS Takaaki Masaki is an Economist in the World Bank’s Poverty and Equity Global Practice. tmasaki@worldbank.org Shohei Nakamura is an Economist in the World Bank’s Poverty and Equity Global Practice. snakamura2@worldbank.org David Newhouse is a Senior Economist in the World Bank’s Poverty and Equity Global Practice. dnewhouse@worldbank.org This note series is intended to summarize good practices and key policy findings on Poverty-related topics. The views expressed in the notes are those of the authors and do not necessarily reflect those of the World Bank, its board or its member countries. Copies of these notes are available on www.worldbank.org/poverty May 2020 · Number 21 4