Tracking Economic Activity In Response To The Covid 19 Crisis Using Nighttime Lights The Case Of Morocco
Summary
A journal article, Tracking economic activity in response to the COVID-19 crisis using nighttime lights: The case of Morocco, published in Development Engineering, Vol. 6 (2021), by an author at the World Bank, in an EconStor copy with its terms-of-use cover page. The article examines whether high-frequency nighttime lights data from the VIIRS satellite sensor can provide near real-time tracking of the economic impacts of the COVID-19 crisis in Morocco. It reports a statistically significant correlation between quarterly movements in Morocco's nighttime light intensity and its real GDP. The abstract states that Morocco experienced a large drop in light intensity in March 2020, with larger declines in some regions and cities than others. The introduction reviews earlier nighttime-lights research, and the document ends with a reference list.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“Tracking Economic Activity In Response To The Covid 19 Crisis Using Nighttime Lights The Case Of Morocco.”
Original at doi.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/tracking-economic-activity-in-response-to-the-covid-19-crisis-using-ni-c03e7124f5/
, SHA-256 097139d9193508d5f1bddf7f42ed85971eea1d2561f61d510aab1647ac0faf97.Full text
Roberts, Mark
Article
Tracking economic activity in response to the COVID-19
crisis using nighttime lights: The case of Morocco
Development Engineering
Provided in Cooperation with:
Elsevier
Suggested Citation: Roberts, Mark (2021) : Tracking economic activity in response to the COVID-19
crisis using nighttime lights: The case of Morocco, Development Engineering, ISSN 2352-7285,
Elsevier, Amsterdam, Vol. 6, pp. 1-13,
https://doi.org/10.1016/j.deveng.2021.100067
This Version is available at:
https://hdl.handle.net/10419/242324
Standard-Nutzungsbedingungen: Terms of use:
Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Documents in EconStor may be saved and copied for your personal
Zwecken und zum Privatgebrauch gespeichert und kopiert werden. and scholarly purposes.
Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle You are not to copy documents for public or commercial purposes, to
Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich exhibit the documents publicly, to make them publicly available on the
machen, vertreiben oder anderweitig nutzen. internet, or to distribute or otherwise use the documents in public.
Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen If the documents have been made available under an Open Content
(insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, Licence (especially Creative Commons Licences), you may exercise
gelten abweichend von diesen Nutzungsbedingungen die in der dort further usage rights as specified in the indicated licence.
genannten Lizenz gewährten Nutzungsrechte.
https://creativecommons.org/licenses/by-nc-nd/4.0/
Development Engineering 6 (2021) 100067
Contents lists available at ScienceDirect
Development Engineering
journal homepage: www.elsevier.com/locate/deveng
Tracking economic activity in response to the COVID-19 crisis using
nighttime lights – The case of Morocco
Mark Roberts
Territorial Development Global Solutions Group; Urban, Disaster Risk, Resilience and Land Global Practice, The World Bank, Singapore Office, Singapore
A R T I C L E I N F O A B S T R A C T
JEL classification: Over the past decade, nighttime lights have become a widely used proxy for measuring economic activity. This
C32 paper examines the potential for high frequency nighttime lights data to provide “near real-time” tracking of the
C53 economic impacts of the COVID-19 crisis in Morocco. At the national level, there exists a statistically significant
O18
correlation between quarterly movements in Morocco’s overall nighttime light intensity and movements in its
R11
real GDP. This finding supports the cautious use of lights data to track the economic impacts of the COVID-19
Keywords:
crisis at higher temporal frequencies and at the subnational and city levels, for which GDP data are unavai
Morocco
COVID-19
lable. Relative to its pre-COVID-19 trend growth path of lights, Morocco experienced a large drop in the overall
Nighttime lights intensity of its lights in March 2020 following the country’s first COVID-19 case and the introduction of strict
Near real-time monitoring lockdown measures, from which it has subsequently struggled to recover. At the subnational and city levels,
Economic activity while all regions and cities examined shared in March’s national decline in nighttime light intensity, some
suffered much larger declines than others. Since then, the relative effects of the COVID-19 shock across regions
and cities appear to have largely persisted. Notwithstanding these findings, however, further research is required
to ascertain the exact causes of the observed changes in light intensity and to fully verify that the results are
driven by anthropogenic causes.
1. Introduction spread. Not only is the availability of near real-time data on economic
activity important at the national level, but also at the more local – i.e.,
Since it was first identified in Wuhan, China, in December 2019, the subnational regional and city levels – especially as many countries have
COVID-19 virus has swept the globe, resulting in not only a devastating moved away from national lockdowns to more geographically differ
loss of life,1 but also widespread economic crisis. In its October 2020 entiated strategies of disease containment.
World Economic Outlook, the IMF projected that global real GDP in Unfortunately, however, conventional measures of economic activity
2020 would fall by 4.4 percent (IMF, 2020a). This compares to its pro produced by national statistics offices are ill-suited to providing such
jection of a 3.4 percent increase made just prior to the global spread of (near) real-time monitoring of economic activity. Official data on such
COVID-19 (IMF, 2020b). The loss of incomes has resulted in a major activity, even at the national level, are often only available after a long
setback in the global fight against extreme poverty, with forecasts lag and at a relatively low temporal frequency. At best, GDP data tend to
indicating a rise of between 119 and 124 million in the number of people be reported at a quarterly frequency at the national level, and an annual
living on less than $1.90 per day in 2020 due to the pandemic (World frequency at the subnational level. For many developing countries, there
Bank, 2021). are no subnational GDP data at all, and, where they are available, they
Crucial to the formulation of effective policy responses to the crisis is only tend to be so for very broadly defined geographic regions (Roberts,
not just up-to-date information on the disease itself, but also “near real- forthcoming). In light of this, there has been growing research interest in
time” data on the associated economic impacts. Such data are important the development of proxy measures of economic activity that can be
for both the design of economic relief packages and in assessing the produced at a high temporal frequency on a (near) real-time basis,
possible economic trade-offs associated with lockdowns and other non- including for subnational regions and cities.2 Among the alternative
pharmaceutical interventions (NPIs) that aim to control the disease’s indicators that have been investigated are data derived from credit card
E-mail address: mroberts1@worldbank.org.
1
At the time of writing (June 2021), the worldwide death toll from COVID-19 is approaching four million.
2
Some of this interest pre-dates the COVID-19 crisis. The crisis has, however, given an extra impetus to such research.
https://doi.org/10.1016/j.deveng.2021.100067
Received 31 January 2021; Received in revised form 16 June 2021; Accepted 21 June 2021
Available online 26 June 2021
2352-7285/© 2021 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
M. Roberts Development Engineering 6 (2021) 100067
transactions and private sector payroll firms (Chetty et al., 2020), spanning 1992–2013, they suffer from several drawbacks that under
Google Trends search data (Woloszko, 2020), and data on outdoor air mine their ability to proxy economic activity (Gibson et al., 2021). Most
pollution (Liu et al., 2020; Masaki et al., 2020). In this paper, we notable among these drawbacks are those of top coding, “overglow”,
investigate the potential of a further unconventional measure of eco and the absence of onboard satellite intertemporal calibration of the
nomic activity – outdoor artificial lighting as detected by satellite sen data.
sors, commonly referred to as “nighttime lights” – to facilitate (near) Top coding refers to the fact that the DMSP-OLS sensors are unable to
real-time tracking of economic activity in response to the COVID-19 detect levels of nighttime luminosity above a certain threshold due to
crisis, focusing on the specific case of Morocco. sensor saturation. This leads to the cores of many cities, where much
To do this, we use monthly and quarterly measures of nighttime light economic activity is concentrated, being top coded with the result that
intensity derived from data collected by the Visible Infrared Imaging increases in their brightness cannot be detected.8 Meanwhile, overglow
Radiometer Suite (VIIRS) sensor onboard the Suomi National Polar- refers to the fact that the light emitted from a given point on the earth is
Orbiting Partnership (NPP) satellite. We first investigate whether, at recorded in the DMSP-OLS data as covering an area that extends, often
the national level, movements in light intensity are correlated with very far, beyond that point. This means that the light, and the economic
movements in real GDP using quarterly data.3 Having established the activity that it potentially measures, is spatially misattributed across
existence of a statistically significant correlation subject to suitable locations, which is a particular problem for subnational and city-level
cleaning of the lights data to remove background noise, we progress to analyses. According to Gibson (2020), both top-coding and overglow
examine the evolution of Morocco’s monthly lights following the result in mean-reverting errors in the DMSP-OLS lights data, implying
announcement of the country’s first confirmed case of COVID-19 on that it understates the true luminosity of brightly lit areas relative to less
March 2, 2020 at both the national and subnational (regional and city) brightly lit areas. Finally, the absence of onboard intertemporal cali
levels. In doing so, we focus on measures of nighttime lights that have bration of the DMSP-OLS data undermines their comparability over
been adjusted to control for “normal” seasonal variation associated with, time, which may help explain the finding by Goldblatt et al. (2020) that
for example, possible atmospheric effects and changes in vegetation.4 they perform poorly in predicting changes in economic activity.
We also examine the evolution of lights following the onset of COVID-19 By contrast, as stated earlier, we make use, in this paper, of lights
relative to estimated pre-COVID-19 trends, where these trends define data from the VIIRS satellite sensor. These are not subject to the top-
the potential counterfactual evolution of lights in the absence of COVID- coding problem, while the overglow problem is also much less severe
19. (Small, 2019). Because the data undergo onboard calibration, they are
Our paper builds on a well-established economics literature that uses also more comparable over time (Gibson et al., 2021). As such, the data
nighttime lights to proxy economic activity.5 This literature dates back appear, prima facie, better suited to detecting temporal changes in eco
to Henderson et al. (2012), who found that the growth of a country’s nomic activity than the DMSP-OLS data. Unlike the DMSP-OLS data,
nighttime light intensity provides a good proxy for its GDP growth over which are only available at an annual frequency, the VIIRS data are also
the long-term. They also found that fluctuations in light intensity at the publicly available at a monthly frequency with a short time lag, which is
national level can track annual fluctuations in economic growth. Since crucial for providing potential near real-time monitoring of the eco
Henderson et al., the use of nighttime lights to proxy economic activity nomic impacts of the COVID-19 crisis. In seeking to examine the po
has become commonplace within the economics literature, especially in tential of VIIRS lights to track the economic impacts of the crisis, this
the urban and spatial economics literatures where analysis is con paper is related to those of Elvidge et al. (2020) and Liu et al. (2020),
strained by the lack of official GDP data (see, for example, Bleakley and both of whom demonstrate a dimming and subsequent recovery of VIIRS
Lin, 2012; Storeygard, 2016; Jedwab et al., 2021). More generally, a lights across China associated with the COVID-19 pandemic, and Beyer
recent review uncovers more than 150 studies in economics that have et al. (2021), who look at the evolution of VIIRS lights for India in
used nighttime lights (Gibson et al., 2020). However, the use of lights to response to the pandemic.9
proxy for economic activity has not been without controversy. Chen and From a remote sensing perspective, Morocco is well-suited to the
Nordhaus (2011) are more downbeat in their findings than Henderson exploration of the potential of nighttime lights to provide high frequency
et al. (2012) on the ability of lights data to reliably proxy economic tracking of the economic impacts of the COVID-19 crisis. This is because,
activity, especially for more agricultural regions. Meanwhile, there is by virtue of its latitude, Morocco has very low cloud-coverage, which,
evidence that, while lights can predict longer-run variation in economic for any given month, results in more high-quality observations of lights.
activity over space, they perform much more poorly in predicting The potential study of the economic impacts of COVID-19 in Morocco is
changes in economic activity over time (Goldblatt et al., 2020). also interesting in and of itself given the lack of knowledge about these
With only a few exceptions, however, the economics literature that impacts. At the time of writing, there have been more than 523,000
uses nighttime lights measures to proxy levels and changes in economic COVID-19 cases in Morocco and in excess of 9200 deaths. Little, how
activity relies on data from Operational Linescan System (OLS) satellite ever, is known about the exact evolution of the economic impacts of the
sensors that were part of the now largely defunct US government De crisis on Morocco and how these have played-out across its regions and
fense Meteorological Satellite Program (DMSP).6,7 While these data cities.
have the advantage of having a relatively long annual time series The remainder of the paper is structured as follows. Section 2 de
scribes the VIIRS lights data and our procedures for cleaning and pro
cessing the data, in addition to the national real GDP data for Morocco.
3
Quarterly is the highest frequency at which national GDP data are available Section 3 describes our analysis of the correlation between changes in
for Morocco. nighttime light intensity and real GDP at the national level using
4
This still, however, leaves the possibility of distortion of the detection of
nighttime lights by atmospheric and other effects that may not follow a regular
seasonal pattern. We discuss this possibility further below.
5 8
This economics literature is pre-dated by several remote sensing studies In addition to the standard DMSP-OLS lights data, there is a radiance-
which investigate the potential of lights to proxy economic activity (see, for calibrated version of the data that was derived based on experiments with
example, Sutton et al., 2007). adjusting the gain settings of the satellite sensors. This version of the data,
6
The exceptions include Roberts (2018), Beyer et al. (2021), and Gibson which is available for seven years between 1996 and 2011, is not subject to top
et al. (2021). coding.
7 9
DMSP still has satellites orbiting and collecting nighttime lights data, but Although not focused on the COVID-19 pandemic, a further related paper is
the program will end once the current satellites reach their end of life (World Chodorow-Reich et al. (2020), who study the impacts of India’s demonetization
Bank, 2020). on real economic activity across districts using, inter alia, VIIRS data.
2
M. Roberts Development Engineering 6 (2021) 100067
quarterly data, which acts as a test of whether lights provide a possible (ii) the “cluster mask”; (iii) the “population mask”; and (iv) the “built-up
proxy for tracking changes in economic activity, both at a higher tem area” mask.15,16 The EOG mask zeroes out lights values for pixels that
poral frequency and at the subnational (regional and city) levels, in the EOG mask out in the creation of their annual lights products for 2015
absence of GDP data. Section 4 presents our analysis of the evolution of and 2016, which, as noted above, have been subjected to further
nighttime lights at the monthly frequency in response to the COVID-19 cleaning over and above their monthly composites.17 Meanwhile, the
crisis at the national, subnational regional, and city levels. Section 5 cluster mask excludes both outlier lights values and pixels that are
concludes by discussing caveats and possible future extensions to the identified as belonging to background noise clusters following the pro
analysis. cedure of Beyer et al. (2021). The population mask, by contrast, uses the
European Commission’s GHS-POP gridded population data set to
2. Data remove lights values that are associated with areas in which no popu
lation is estimated to have been present in 2015.18 Finally, the built-up
2.1. Nighttime lights data area mask removes lights values that are associated with areas in which
there was no detected built-up area in 2018. To derive this mask, we use
As discussed above, the nighttime lights data that we use are derived the built-up area layer from the 2018 MODIS land use – land cover
from observations collected by the VIIRS sensor onboard the Soumi-NPP product with a 1 km buffer to account for potential under-detection of
satellite, which is jointly operated by NASA and the US government’s built-up area.19 It should be noted that none of these masks is perfect –
National Oceanic and Atmospheric Administration (NOAA). In partic all are imperfect substitutes for the full, computationally highly inten
ular, we use the monthly composites of VIIRS lights produced by the sive, cleaning procedure that EOG has yet to apply to its series of
Earth Observation Group (EOG) at the Colorado School of Mines.10 monthly composites.
These composites cover virtually the entire globe at a resolution of 15 As can be seen from Fig. 1, which provides an illustration of levels of
arc seconds, which is equal to 460 m2 at the equator.11 This is much light intensity in March 2020 in the vicinity of Morocco’s most populous
higher than the 30 arc second resolution (approximately 1 km2 at the city, Casablanca, and surrounding secondary cities, the cluster mask is
equator) of the old DMSP-OLS lights data. For each pixel, the composites the least spatially extensive, in terms of pixels zeroed-out, followed by
report luminosity in nanowatts/cm2/sr as an average calculated over all the EOG mask, population mask and built-up area mask. Unlike the
cloud-free nights in a month. We make use of the monthly composites for other masks, the built-up area mask zeroes-out all light along the roads
April 2012–September 2020, where the April 2012 composite is the first that connect the cities in the map. Most notably, along the road corridor
that was produced following the launch of the Suomi-NPP satellite in that connects the cities of Casablanca, Deroua, Berrechid and Settat.
2011.12 As noted in Section 1, because of its latitude, Morocco has very Given that it is a priori unclear which of the masks is best – in terms of
low cloud coverage. As a result, virtually all pixels in the data for allowing changes in lights to proxy changes in economic activity – we
Morocco have – after dropping the data for June for each year for rea treat this as an empirical question. We settle this question by examining
sons that we explain below – at least two nights, and often many more, of the strength of the estimated correlations between the four different
cloud free coverage for all months between April 2012 and September cleaned lights data series that result from the application of the masks
2020.13 and GDP (see Section 3 below).20
In creating the monthly composites, EOG filter-out both lightning Even after the application of the above described masks, however,
and lunar illumination prior to averaging. However, unlike the annual distortions in the lights data for Morocco remain for June in each year.
composites, which are only currently available for 2015 and 2016, This is due to the summer solstice, which results in “stray-light”
which EOG also produce, they do not filter out lights from aurora, fires, contamination of the data. Such contamination arises from the Soumi-
boats, and other ephemeral lights. As such, the data are still subject to, NPP satellite being illuminated by sunlight while observing areas of
potentially considerable, background noise. To screen out this noise, we the Earth’s surface where the sun is under the horizon (Elvidge et al.,
undertake further cleaning of the monthly composites by applying 2017). Given the severity of this contamination for Morocco, we drop
various masks to the data.14 In particular, we experiment with four the June data for each year in our analysis below.21 It is important to
different background masks, which we refer to as: (i) the “EOG mask”;
15
In their work on China, Liu et al. (2020) instead mask-out pixels with a
radiance of less than 5 nW/cm2/sr, which they associate with vegetation, water,
10
This group was formerly based at NOAA. The group is the same as that snow, and rural areas.
16
which produced the DMSP-OLS annual composite lights data sets on which All four masks are constructed using global data sets, implying that they can
most research that has used nighttime lights has relied. also easily be constructed for countries other than Morocco.
11 17
The composites cover the globe from 75 N latitude to 65 S. The EOG cleaning procedure is described in Elvidge et al. (2017). Both
12
Monthly VIIRS composites are made freely available by EOG with a lag of Beyer et al. (2021) and Gibson et al. (2021) apply similar masks to clean VIIRS
three months on their website (https://payneinstitute.mines.edu/eog-2/viirs/). monthly composites of background noise for India (Beyer et al.) and China,
We obtained access to the composites through EOG’s paid subscription service Indonesia, and South Africa (Gibson et al.).
18
which provides access with a maximum lag of around two weeks. GHS-POP can be downloaded from https://ghsl.jrc.ec.europa.eu/download.
13
As an illustration of this, in Section 4.3.2, we analyze the evolution of php?ds=pop.
19
nighttime lights for the fourteen Moroccan cities that had a 2015 population The MODIS (Moderate Resolution Imaging Spectroradiometer) product is
more than 200,000. For each city, the average share of pixels that have at least available from https://modis.gsfc.nasa.gov/data/dataprod/mod12.php. MODIS
two nights of cloud-free coverage, after dropping the data for June for each is an instrument aboard NASA’s Terra and Aqua satellites.
20
year, over all months between April 2012–September 2020 exceeds 98.7 Given that GHS-POP grids population based on distributing population
percent in all cases. The average share of pixels that have at least one night of counts for administrative units across the built-up area within those units, the
cloud-free coverage exceeds 99.5 percent for all cities. This suggests that our population mask is also, in effect, a type of built-up area mask. However, the
results are unlikely to be driven by variations in cloud coverage. built-up area layer according to which GHS-POP grids population differs from
14
We only apply this additional cleaning for our national and subnational the 2018 MODIS land use – land cover product. The differences in detection of
regional analysis of the lights data. For our city-level analysis, low level back built-up area explain the differences between the two masks.
21
ground noise, which arises largely outside of urban areas, is already de facto EOG do provide a “stray-light corrected” configuration of their monthly
masked-out through the definition of the cities. VIIRS composites in addition to the versions of the composites that we use.
However, the correction procedure is imperfect and affects the more general
quality of the data. The decision to use the non-corrected data and drop June
for each year was based on advice from EOG.
3
M. Roberts Development Engineering 6 (2021) 100067
Fig. 1. Nighttime light intensity in March 2020 in the vicinity of Casablanca and surrounding secondary cities based on the four different masks: (a) cluster mask; (b)
EOG mask; (c) population mask; and (d) built-up area mask.
note that there may also be other sources of distortion in the data in chained prices with a base year of 2007. The data are available only as
associated with, for example, atmospheric effects. To the extent that a seasonally adjusted series, where seasonal adjustment has been per
these follow a predictable seasonal pattern, our analysis below accounts formed by the Directorate of Statistics using the US Census Bureau’s X-
for them. However, this does not rule-out the possibility of distortions 12-ARIMA procedure (Findley et al., 1998).
arising from atmospheric effects that do not follow such a pattern. If the
timing and size of such distortions are, for example, correlated with the 3. Ability of nighttime lights to track economic activity
onset of the COVID-19 pandemic in Morocco then this will undermine
the attribution of changes in lights to causes related to the pandemic. To test whether or not changes in nighttime light intensity can
As our measure of aggregate light intensity, we use the “sum of potentially be used to proxy changes in economic activity at a high
lights” (SOL), which, for any given area, is derived by summing light temporal frequency and/or at spatial scales at which official data on
intensity values for all pixels in the cleaned data that fall within that such activity are unavailable, we run several time series regressions to
area. Hence, we calculate SOL values for all months (bar June for each examine the correlation between changes in light intensity and changes
year) between April 2012 and September 2020 for Morocco overall, its in real GDP at the national level for Morocco using quarterly data for the
Admin-1 subnational regions, and for all cities which had a 2015 pop period 2012 Q3 – 2020 Q1.22,23 These regressions are as follows:
ulation of at least 200,000. We also calculate quarterly SOL values for
ln(GDPt ) = α + βt + γQ + δ ln(SOLt ) + εt (1)
Morocco overall by averaging SOL values over the months in each
quarter. The quarterly SOL values are used to examine the correlation
between changes in lights and changes in real GDP.
22
2.2. Real GDP data We exclude the Q2 data for each year due to the straylight corruption of the
June lights data (see Section 2.1).
23
Regression models (1)–(3) are static models in the sense that they model a
Quarterly real GDP data, which is the highest frequency that is
contemporaneous relationship between ln (SOL) and ln (GDP). As stated by
available, covering the period 2012 Q3 – 2020 Q1, come from Morocco’s Wooldridge (2016), in time series analysis, “… a static model is postulated
Directorate of Statistics. These are national level data that are measured when a change in z [the independent variable] at time t is believed to have an
immediate effect on y [the dependent variable].” (emphasis added). This im
plies that, even though the regressions are specified in log levels, the rela
tionship between ln(SOL) and ln(GDP) is identified by the changes in the two
variables over time.
4
M. Roberts Development Engineering 6 (2021) 100067
ln(GDPt ) = α + βY + γQ + δ ln(SOLt ) + εt (2) cluster mask is slightly lower at 0.262.27 Nevertheless, it remains highly
statistically significant. At 0.35, the R2 is likewise very slightly lower.28
ln(GDP)D,t = α + β ln(SOL)D,t + εt (3) Nevertheless, an R2 of 0.35–0.37 still implies that fluctuations in
nighttime light intensity are unable to “explain” almost two-thirds of the
In equation (1), the natural log of Morocco’s real GDP in quarter t is
quarterly variation in real GDP, meaning that the relationship is rather
regressed on its SOL for the same quarter while controlling for both a
noisy. This is evident from Fig. 2, which provides a visualization of the
linear time trend and quarter effects as captured by the set of quarterly
column [3b] results based on the EOG and cluster masks from Table 1.
dummies, Q. We also estimate equation (1) without the quarterly
Hence, for both masks, the magnitude of the change in lights is much
dummies. In this case, the regression tests whether quarterly changes in
closer to the magnitude of the change in real GDP for some fluctuations
SOL are correlated with quarterly deviations of GDP from its growth
than for others. Consistent with this, if we focus on the data for the
path. When the quarterly dummies are included, the regression tests
period 2014 Q1 – 2020 Q1 rather than for 2012 Q3 – 2020 Q1 then the
whether changes in SOL can predict deviations in GDP around its growth
estimated elasticity of GDP with respect to SOL increases to 0.35–0.36,
path over and above regular quarterly fluctuations. Meanwhile, in
depending on the mask used. For the most recent fluctuations, visual
equation (2), the linear time trend from equation (1) is replaced by a full
inspection of Fig. 2 suggests that the elasticity may even be close to
set of year effects, Y.24 As with equation (1), we estimate (2) both
unity, especially for results based on the cluster mask.
excluding and including the quarter effects, Q. Finally, in equation (3),
Despite this, it is also apparent from Fig. 2 that, throughout the
we regress de-trended (ln) GDP on de-trended (ln) SOL where de-
sample-period, the timing of changes in nighttime light intensity line-up
trending is performed separately for both the ln (GDP) and ln (SOL)
reasonably with the timing of changes in real GDP, especially for the
series. This is done by first separately regressing each of ln (GDP) and ln
EOG mask. For the EOG mask, a switch from a below to an above trend
(SOL) on either a linear time trend or a full-set of year effects. The re
level of real GDP or vice versa is matched by a corresponding switch in
siduals from these separate regressions then act as the de-trended ln
SOL from below to above trend or vice versa in 10 out of 23 cases (i.e. in
(GDP) and ln (SOL) time series.25 In total, therefore, we estimate six
43.5 percent of cases). For the remaining 13 cases, the switch in SOL
regression specifications – both equations (1) and (2) each with and
occurs within plus or minus one quarter of the switch in GDP. Mean
without the quarter effects, and equation (3) based on de-trending using
while, for the cluster mask, a switch in SOL occurs in the same quarter as
a linear time trend and year effects. These six specifications are esti
a switch in real GDP in seven out of 23 cases (30.4 percent) and within
mated for SOL calculated both on the basis of the “raw” lights data
one quarter for the remaining 16 cases. This implies that the cleaned
provided by EOG and for each of the four SOL series that have been
lights data are perhaps better used to indicate likely changes in the di
cleaned using the EOG, cluster, population and built-up area masks. In
rection of real economic activity than to provide precise quantitative
all cases, regression is by least squares with Newey-West standard errors,
estimates, based on the overall estimated elasticity, of the change in real
assuming a first-order serial correlation process for the error term
GDP for any given fluctuation. Although, even when it comes to
(Newey and West, 1987).26
detecting the timing of changes in real GDP, significant care still needs to
Table 1 reports the regression results. As can be seen, there is no
be exercised given that the lights and GDP data do not line-up
statistically significant relationship between lights and GDP for any of
perfectly.29
the six regression specifications when using the “raw” lights data to
calculate SOL. The fit of the de-trended regressions in columns [3a] and
4. COVID-19 and the evolution of Morocco’s nighttime lights
[3b], as indicated by their R2 values, is furthermore very low. However,
once we mask out background noise from the lights data, we do obtain
4.1. Methodology for estimating the “impact” of the COVID-19 crisis on
statistically significant relationships for most of the specifications, along
lights
with better fits for the de-trended regressions. While this is the case for
the results based on all four masks, it is the EOG mask followed by the
Having established that the lights data can, with reasonable caution,
cluster mask which generates the strongest relationships, judged based
be used to indicate changes in real economic activity, we now turn to
both on statistical significance and the fits of the de-trended regressions.
examine the evolution of Morocco’s lights in response to the COVID-19
If we focus on the column [3b] results then the estimated elasticity of
crisis. In doing so, we first focus on the monthly evolution at the national
GDP with respect to SOL is 0.295 when using the EOG mask, which is
level before turning to examine the monthly evolution at the subnational
significant at the 1 percent level, with an R2 of 0.37. This implies that
regional and city levels. In all three cases, our methodology involves
changes in the intensity of Morocco’s lights around their trend can
“explain” just over one-third of the variation in GDP around its trend.
The estimated elasticity of GDP with respect to SOL when using the
27
Our estimated elasticity of 0.262–0.295 is remarkably close to that obtained
by Henderson et al. (2012), which they estimate using annual GDP and lights
data for a global panel of countries. This is despite Henderson et al.‘s analysis
24
The control for a time trend/year effects helps avoid spurious correlation being based on the old DMSP-OLS data as opposed to the newer VIIRS data.
28
between nighttime lights and GDP. Table A1 in Annex 1 presents results based on the EOG and cluster masks,
25
It will be noted from Table 1 that estimation of equation (3) produces from the estimation of equations (1)–(3) for the industry, services, and agri
identical point estimates of the coefficient on ln(SOLt ) as estimation of equation cultural sectors. These regressions suggest that fluctuations in light intensity are
(1) when excluding the quarter effects. The difference lies in the associated most highly correlated with fluctuations in industry GDP. This is notably the
standard errors (t-values), which are smaller (larger) when we estimate equa case in the estimation of equation (2) and of equation (3) when de-trending is
tion (3). based on year effects. Interestingly, fluctuations in light intensity also hold
26
This is based on the rejection of the hypothesis that the error terms in the some predictive ability for fluctuations in agricultural GDP when estimating
regressions are serially uncorrelated using both Durbin’s alternative test for equation (1) and equation (3) when de-trending is based on a linear time trend.
29
serial correlation and the Breusch-Godfrey test for serial correlation. In general, One might suspect the relationship between fluctuations in real GDP and
results are very similar when we instead estimate using standard OLS or instead SOL of being asymmetric between economic expansions and contractions. For
assume a second-order serial correlation process. Note that Newey-West stan example, Henderson et al. (2012) hypothesize “the possibility that because
dard errors are also robust to heteroskedasticity. some lights growth reflects the installation of new capacity, lights are nonde
creasing, so that economic downturns will not be reflected in lights.” However,
they find no evidence in support of such an asymmetric relationship in their
own work, and, from Fig. 2, there likewise seems no evidence to support the
existence of an asymmetric relationship in quarterly data for Morocco.
5
M. Roberts Development Engineering 6 (2021) 100067
Table 1
Relationship between nighttime lights and GDP, quarterly data.
Time trend Time trend + quarter effects Year effects Year + quarter effects De-trended (time trend) De-trended (year effects)
[1a] [1b] [2a] [2b] [3a] [3b]
“Raw” nighttime lights data
ln(SOL) 0.025 0.020 0.083 0.006 0.025 0.083
(0.89) (1.39) (1.19) (0.68) (0.91) (1.50)
R2 0.95 0.99 0.93 0.99 0.03 0.11
EOG mask
ln(SOL) 0.129** 0.088*** 0.295*** 0.051** 0.129** 0.295***
(2.26) (3.86) (4.43) (2.25) (2.31) (5.56)
2
R 0.96 0.99 0.95 0.99 0.13 0.37
Cluster mask
ln (SOL) 0.100** 0.06*** 0.262*** 0.035 0.100** 0.262***
(2.11) (3.11) (4.23) (1.54) (2.16) (5.30)
R2 0.96 0.99 0.95 0.99 0.11 0.35
Population mask
ln (SOL) 0.109* 0.089*** 0.250*** 0.051** 0.109* 0.250***
(1.91) (3.96) (3.26) (2.52) (1.96) (4.09)
2
R 0.96 0.99 0.94 0.99 0.10 0.29
Built-up area mask
ln (SOL) 0.082 0.087*** 0.197 0.052*** 0.082 0.197*
(1.17) (3.65) (1.55) (3.81) (1.20) (1.94)
R2 0.95 0.99 0.93 0.99 0.06 0.18
Notes: ***, ** and * indicate significance at the 1, 5 and 10 percent levels respectively. t-values are reported in parentheses and are based on Newey-West standard
errors, assuming a first-order serial correlation process. Estimated constant terms in the regressions not reported.
Fig. 2. a. Relationship between nighttime lights and GDP, EOG mask. Fig. 2b. Relationship between nighttime lights and GDP, cluster mask.
estimating a multivariate regression that models the evolution of SOL as atmospheric and other sensor related effects, and COVID is a dummy
following an exponential time trend while controlling for systematic variable which equals one if a month falls in the post-COVID-19 onset
differences in light intensity between months that are allowed to differ period and zero otherwise.30 Given that Morocco’s first case of COVID-
between the pre- and post-COVID-19 onset periods. From this model, 19 was announced on March 2, 2020, we define the post-COVID-19
SOL values adjusted for both seasonal effects – in particular, adjusted for onset period as consisting of all months from March 2020 onwards.
systematic monthly variations in the lights data that may be due, for For pre-COVID-19 months, the estimated residuals from this regression,
example, to seasonal atmospheric effects and changes in vegetation that ε̂i,m , give the seasonally adjusted SOL values relative to the pre-COVID-
are unrelated to economic activity – and the pre-crisis trend are derived, 19 SOL trend growth path, the estimated slope of which is itself given by
and changes in these values calculated. β. Meanwhile, ̂
̂ δ provides estimates of seasonally adjusted SOL values
More formally, using monthly SOL data for the period April relative to the pre-crisis trend for post-COVID-19 onset months. As with
2012–September 2020 (excluding June data for each year), we estimate the regressions in Section 3, we estimate equation (4) using least squares
the following regression: with Newey-West standard errors to control for both first-order serial
( )
ln SOLi,m = α + βm + γMonth + δ(COVID.Month) + εi,m (4) correlation and heteroskedasticity.31
where SOLi,m is a geographic area i’s SOL in month m, m is a time-trend
variable, Month are a full set of monthly dummies that help control for 30
The monthly dummies also help to control for any seasonal variation in
seasonal effects, including those associated with possible seasonal
cloud coverage, although, as discussed in Section 2.1, cloud-free coverage is
high in Morocco, suggesting that this is not a major concern when analyzing
lights data for the country.
31
As with the regressions in Section 3, the use of Newey-West standard errors
is justified by the rejection of the hypothesis that the error terms are serially
uncorrelated using both Durbin’s alternative test for serial correlation and the
Breusch-Godfrey test for serial correlation.
6
M. Roberts Development Engineering 6 (2021) 100067
A corollary of our approach to examining the evolution of lights any “true” on-the-ground change in luminosity. Unfortunately, howev
following the onset of COVID-19 is that we effectively use the pre- er, the global nature of the COVID-19 pandemic makes such a research
COVID-19 trend growth path of an area’s lights – be the area Morocco design unfeasible. This means that caution must be exercised in inter
overall or one of its subnational regions or cities – as defining the preting the estimates of “impact” that we present as causal.
counterfactual relative to which changes in seasonally-adjusted light
intensity are measured. A conceptual illustration of this is provided by 4.2. National-level results
Fig. 3. In this figure, ̂ε Feb. is the (hypothetical) residual from estimating
equation (4) for February 2020 – the last pre-COVID-19 onset month – The monthly national level results based on the EOG mask which
and corresponds to the (ln) percentage points by which area i’s generates the strongest relationship between lights and real GDP in
seasonally adjusted SOL deviated from its pre-COVID-19 growth trend in quarterly data (see Section 3) are presented in Table 2 and Fig. 4.32 We
that month. Meanwhile, δ ̂March corresponds to the estimated coefficient highlight three key findings. The first is that, following the detection of
on the interaction dummy variable (COVID.Month) for March 2020. This the first COVID-19 case and the imposition of strict non-pharmaceutical
is an estimate of the (ln) percentage points by which area i’s seasonally interventions (NPIs) to control the disease’s spread, Morocco witnessed
adjusted SOL deviated from the pre-COVID-19 growth trend in the a sharp fall in the intensity of its nighttime lights. Fig. 5 shows the sharp
month in which Morocco’s first COVID-19 case was announced. Given increase in the numbers of confirmed COVID-19 cases and deaths that
this, ̂
δ March − ε̂Feb. then provides a measure of the “impact” of COVID-19 occurred in Morocco in March 2020 following the detection of the first
on area i’s (seasonally adjusted) light intensity, which is equal to the (ln) case, while Fig. 6 shows the sharp increase in the stringency of Moroc
percentage point change in SOL between the two months relative to the co’s NPIs, as captured by the NPI stringency index constructed by Hale
estimated pre-COVID-19 trend growth path of nighttime light intensity. et al. (2020), that was implemented in reaction to this. Accompanying
In addition to presenting estimates of changes in seasonally adjusted this, Morocco witnessed a steep month-on-month decline of 10.9 per
SOL relative to the pre-COVID-19 estimated trend growth path, we also centage points in its overall light intensity (i.e. seasonally adjusted SOL)
present estimates of year-on-year (y-o-y) changes in seasonally adjusted relative to the pre-crisis trend (Table 2; Fig. 4).33 As a result, whereas in
SOL relative to this same growth path. For March 2020, such an estimate February 2020, prior to the first case, the intensity of Morocco’s lights
had been 3.0 percentage points above its pre-COVID-19 trend, by March,
corresponds to δ ̂March − ε̂March,2019 where ̂ε March,2019 is the residual for
the intensity was 7.9 percentage points below its pre-COVID-19 trend
March 2019 from estimating equation (4).
(Table 2; Fig. 4(a)). This month-on-month decline in lights was the
The application of the above methodology to derive estimates of
largest that Morocco had experienced since December 2013.34 An even
“impact” for any given geographic area is subject to at least one
larger decline in light intensity, relative to the pre-crisis trend, is evident
important caveat. This is that, ideally, we would like to be able to
on a year-on-year basis – between March 2019 and March 2020,
compare the evolution of lights before and after the onset of COVID-19
seasonally adjusted SOL fell by 13.4 percentage points relative to this
with the evolution in a set of comparison locations that did not experi
trend. This is the fifth largest year-on-year decline in seasonally adjusted
ence COVID-19. This would then facilitate the application of a
SOL during the entire time-series of monthly lights.
difference-in-differences estimator with the evolution of lights in the
The second key finding is that Morocco’s lights showed some signs of
comparison locations defining the counterfactual against which the
recovery in May 2020 as COVID-19 cases and deaths both leveled-off.
impact of COVID-19 is estimated. Given a suitable set of comparison
Hence, following March’s large drop, Morocco experienced a further,
locations, the application of such an estimator would help to control for
much smaller, drop in its overall (seasonally adjusted) light intensity in
omitted variables that are common to both the COVID-19 affected and
April 2020 of 3 percentage points, relative to the pre-COVID-19 trend
comparison locations, but which may be correlated with both the in
(Table 2). This left Morocco’s lights 13.9 percentage points dimmer in
tensity of nighttime lights and COVID-19. This could potentially include
April 2020 than they were in February 2020, the final pre-crisis month
(Fig. 4). May, however, then saw a partial recovery in the intensity of
Morocco’s seasonally adjusted lights as the numbers of confirmed
COVID-19 cases and deaths stabilized (Fig. 5). Relative to the pre-crisis
trend, seasonally adjusted light intensity increased by 6.3 percentage
points between April and May 2020 (Table 2). On a year-on-year basis,
however, May’s change in nighttime light intensity remained negative.
In our final key finding, however, the partial May recovery was short-
lived, and, by September 2020, Morocco’s lights were as dim as they
were in March following the initial impact of the COVID-19 shock.
Hence, although we lack useable June lights data due to the stray-light
problem, Fig. 4 shows that, by July, the May uptick in seasonally
adjusted light intensity had been more than wiped-out. Relative to the
pre-crisis trend, light intensity had thus reverted to just below its April
32
Results based on the application of the cluster, population, and built-up area
masks, which are similar to the results based on the EOG mask, are available on
request. The full results from the estimation of equation (4) based on the EOG
Fig. 3. Conceptual illustration of estimation of effects of COVID-19 on night
mask can be found in Table A2, Annex 2.
time light intensity. 33
Assuming an elasticity of 0.30, this corresponds to a fall in real GDP of 3.3
percentage points, although, as discussed in Section 3, any precise estimate of
non-seasonal and unaccounted for atmospheric effects that may lead to the change in real GDP derived from the lights data needs to be treated with
distorted nighttime lights measurements. In the absence of such a caution.
research design, and, in particular, given our simpler time series design, 34
The fall was also the fourth largest since data records began. The three
we cannot completely rule-out the possibility of a large atmospheric largest month-on-month declines in seasonally adjusted light intensity relative
“shock” that could have coincided with the onset of the COVID-19 crisis to the pre-COVID-19 trend were in December 2013 (− 14.70 percentage points),
to cause a decline in measured nighttime light intensity in the absence of March 2013 (− 14.28 percentage points), and November 2012 (− 14.07 per
centage points). These declines all took place during the Arab Spring period.
7
M. Roberts Development Engineering 6 (2021) 100067
Table 2
Nighttime lights statistics.
Difference from trend (%)
Feb. 2020 Mar. 2020 Apr. May 2020 Jul. Aug. Sep.
2020 2020 2020 2020
3.00 − 7.88** − 10.92*** − 4.58** − 11.55*** − 7.03** − 7.34***
Change relative to trend (percentage points)
Feb. → Mar. Mar. → Apr. Apr. → May May → Jul. Jul. → Aug. Aug. → Sep.
− 10.88 − 3.03 6.34 − 6.97 4.52 − 0.32
Year-on-year change relative to trend (percentage points)
Feb. 2020 Mar. 2020 Apr. May 2020 Jul. Aug. 2020 Sep. 2020
2020 2020
1.47 − 13.44 − 12.00 − 7.05 − 6.49 2.94 − 6.25
Notes: ***, ** and * indicate significance at the 1, 5 and 10 percent levels respectively. Results based on EOG mask, which omits areas excluded in 2015 and 2016
annual composites. Trend refers to the pre-COVID-19 trend of seasonally adjusted SOL, as estimated using monthly data for Apr. 2012–Feb. 2020. No results reported
for June 2020 due to stray-light contamination of lights data.
Fig. 4. Changes in seasonally adjusted NTL intensity.
Fig. 5. Confirmed COVID-19 cases and deaths.
Fig. 6. NPI stringency index.
level. This coincided with an upturn in the growth in the number of
confirmed COVID-19 cases (Fig. 5). In August, seasonally adjusted light the pre-crisis trend (Table 2).
intensity increased, but, by September, Morocco’s seasonally adjusted One concern with the above findings, that we have already alluded
light intensity remained, relative to the pre-crisis trend, 10.3 percentage to, is that they may be the spurious result of atmospheric effects that just
points below its level in the final pre-COVID-19 month of February happened to bias the measurement of lights downwards in March 2020
2020. This is where it stood in March 2020 following the initial onset of at the same time as the onset of COVID-19 in Morocco and the imple
the shock, implying the absence of any overall recovery with light in mentation of NPIs or, equivalently, the disappearance of a temporary
tensity standing a statistically significant 7.3 percentage points below upwards bias in the measurement of lights in February 2020 due to
8
M. Roberts Development Engineering 6 (2021) 100067
atmospheric effects. Some support for the latter possibility seems to be change in a region’s seasonally adjusted light intensity between
provided by Fig. 4(a), which shows that there was a large upward February and September and the change in its seasonally adjusted light
movement in the intensity of Morocco’s lights between January and intensity between February and March. Consistent with this, the two
February 2020. Indeed, the magnitude of this upwards movement (+9.8 regions, Rabat – Salé – Kénitra and Tanger – Tetouan – Al Hoceima, that
percentage points relative to the pre-COVID-19 trend) was almost as appear to have been initially hardest hit by the crisis remained the
large as the subsequent downwards movement (- 10.9 percentage hardest hit in September. Conversely, Souss – Massa and Casablanca –
points) that occurred in March and coincided with the onset of COVID- Settat, which were the two regions that appear to have been initially
19. However, this would not explain the large year-on-year decline in least hard hit, were also the only two to have recovered to levels of light
the intensity of lights between March 2019 and March 2020. Also, the intensity, relative to the pre-COVID-19 trend, that were above their pre-
disappearance of a temporary February 2020 distortion cannot explain crisis levels in February (Fig. 7).
why light intensity would have remained significantly below its pre- More generally, assuming that they are not driven by unaccounted
COVID-19 growth path for the remainder of the sample-period, which for atmospheric distortions, our results allow us to distinguish between
runs up until September 2020. Still, the noisiness of the seasonally three groups of regions in Morocco – the hardest hit, the hard-hit and the
adjusted SOL time series relative to the pre-COVID-19 growth path recovered. In addition to Rabat – Salé – Kénitra and Tanger – Tetouan –
which is evident in Fig. 4(a) means that we cannot unequivocally rule- Al Hoceima, the hardest hit group also includes Fès – Meknès. Compared
out the possibility that the results may be driven by unaccounted for to their pre-crisis trends, these areas had September 2020 levels of light
(non-seasonal) atmospheric effects. intensity between 19.2 and 26.0 percentage points lower than they were
in February 2020. This corresponds to a 5.8–7.8 percentage points drop
in real GDP given an elasticity of 0.30 with the caveat, again, that
4.3. Subnational results
estimated real GDP changes need to be treated with caution. The hard-
hit group meanwhile comprises Drâa – Tafilalet, Marrakech – Safi,
4.3.1. Admin-1 regions35
Oriental – RIF and Béni Mellal – Khénifra. This group suffered light in
Table 3 and Fig. 7 present our subnational results which cover nine of
tensity (estimated real GDP) declines between February and September
Morocco’s Admin-1 level regions, as defined by their current bound
2020 in the range of 5.0–12.0 (1.5–3.6) percentage points. Finally, Souss
aries, which date back to 2015.36 The first finding to emphasize is that
– Massa and Casablanca – Settat are the recovered group, with
while all regions shared in March 2020’s national lights decline, some
September 2020 light intensity (estimated real GDP) levels that were 1.2
suffered more than others. Thus, out of the nine regions analyzed, Fès –
(0.4) and 5.7 (1.7) percentage points respectively above their February
Meknès experienced the largest month-on-month decline – 16.5 per
2020 levels, relative to pre-crisis trends (Table 3, Fig. 7). However,
centage points – in the intensity of its seasonally adjusted lights in March
whether Souss – Massa and Casablanca – Settat have retained their
relative to its pre-COVID-19 trend growth path. Tanger – Tetouan - Al
recovered status as Morocco’s COVID-19 crisis has further evolved is an
Hoceima and Oriental - RIF also experienced declines greater than 15
open question.
percentage points, while a further three (Rabat – Salé – Kénitra, Drâa-
Tafilalet, and Marrakech – Safi) experienced declines more than 10
4.3.2. Cities
percentage points. The only region to escape a large fall in the intensity
Table 4 reports results for the 14 Moroccan cities whose population
of its lights in March was Souss – Massa (- 0.5 percentage points).37
exceeded 200,000 in 2015, where we define cities as “urban centers”
Given an assumed elasticity of 0.30, a 15 (10) percentage point decline
following the European Commission’s degree of urbanization method
in light intensity is equivalent to a 4.5 (3.0) percentage point drop in real
ology (Dijkstra and Poelman, 2014; Dijkstra et al., 2021). According to
GDP, although, as noted above, any attempt to derive precise quanti
this definition, a city is composed of a contiguous set of 1 km2 pixels,
tative estimates of changes in real GDP needs to be treated with caution.
each of which has a population density of at least 1500 people per km2
In addition to the heterogeneous movements of nighttime lights
and whose collective population is at least 50,000.38 As might be ex
following the onset of Morocco’s COVID-19 crisis in March 2020,
pected, the results largely mirror those for Morocco’s Admin-1 level
Table 3 and Fig. 7 also reveal that the effects of the shock appear to have
regions. Thus, the cities that suffered the largest (smallest) declines in
been highly persistent across regions. Thus, while light intensity levels
their seasonally adjusted nighttime lights in March 2020 relative to their
relative to pre-crisis trends have fluctuated at the regional level since
pre-COVID-19 trends tended to be those which fall under the Admin-1
March, there is a strong correlation between the initial “impact” of the
level regions whose seasonally adjusted lights also declined by the
shock in March and the overall “impact” as observed in September 2020.
most (least) in March 2020. And, as at the Admin-1 level region, the
This is reflected in an estimated correlation coefficient of 0.69 and an
“impacts” of the COVID-19 shock have been highly persistent across
estimated Spearman’s rank correlation coefficient of 0.75 between the
cities. This is again indicated by the high estimated correlation and
Spearman’s rank correlation coefficients (0.80 and 0.82 respectively)
35 when comparing the change in seasonally adjusted lights between
Again, we focus on results based on the EOG mask. Results are, however,
February and September 2020 with the corresponding change between
similar regardless of the choice of mask.
36
Our analysis excludes three sparsely populated regions that were not fully
February and March 2020 across cities. Indeed, these estimated corre
covered by the lights data provided by EOG under their subscription service. lation coefficients are even higher than the corresponding coefficients
However, an analysis of December 2019 lights data, where we have complete that we reported for the Admin-1 region level, indicating even stronger
coverage of all regions including these, shows that they accounted for only 4.6 persistence at the city than at the regional level.
percent of total SOL for that month (calculated using pixel values equal to or Notwithstanding the overall similarity of the results, however, we
greater than one), which is consistent with their low overall density of eco also observe some important differences in results across cities which
nomic activity. Furthermore, results from Gibson et al. (2021), based on data belong to the same Admin-1 level region. This is the case, for example,
for Indonesia, China, and South Africa, suggest that VIIRS lights data are less of
a reliable proxy for economic activity for low-density regions.
37
Similar to national level results, most Admin-1 regions experienced a
38
month-on-month increase in light intensity in February 2020, just prior to the Population is measured using the European Commission’s GHS-POP popu
onset of COVID-19. Again, this raises some concern that the declines estimated lation grid, which is the same grid as that on which the population mask is
for March 2020 may be the result of the disappearance of a temporary February based. The shapes of the 14 Moroccan cities were extracted from Urban Center
2020 distortion of lights that resulted in an upwards measurement bias. Database UCDB R2019A (Florczyk et al., 2015). Results for cities were derived
Although this seems unlikely to fully account for the results presented, we can, without applying a background noise mask given that background noise tends
again, not unequivocally rule it out. to be associated with non-urban areas.
9
M. Roberts Development Engineering 6 (2021) 100067
Table 3
Changes in light intensity relative to pre-COVID-19 trend (percentage points), Admin-1 regions.
Admin-1 unit Change from previous month, percentage points (Change from February 2020, percentage points)
March April May July August Sept.
2020 2020 2020 2020 2020 2020
Rabat – Salé – Kénitra − 11.95 − 3.11 3.58 − 15.60 11.23 − 10.14
(-11.95) (-15.06) (-11.48) (-27.08) (-15.85) (-25.99)
Tanger – Tetouan - Al Hoceima − 15.33 − 3.76 7.13 − 6.98 3.40 − 6.52
(-15.33) (-19.10) (-11.96) (-18.95) (-15.55) (-22.07)
Fès – Meknès − 16.49 − 11.79 13.47 − 4.47 1.36 − 1.27
(-16.49) (-28.28) (-14.82) (-19.28) (-17.93) (-19.20)
Béni Mellal – Khénifra − 7.42 − 9.97 8.37 − 2.28 0.90 − 1.57
(-7.42) (-17.39) (-9.02) (-11.29) (-10.39) (-11.96)
Oriental – RIF − 15.28 − 2.65 12.41 − 7.82 3.78 3.18
(-15.28) (-17.94) (-5.53) (-13.35) (-9.57) (-6.39)
Marrakech – Safi − 11.62 − 6.92 11.76 − 4.08 3.16 1.69
(-11.62) (-18.54) (-6.78) (-10.86) (-7.70) (-6.01)
Drâa – Tafilalet − 11.65 7.26 1.20 − 6.10 − 0.87 5.21
(-11.65) (-4.40) (-3.19) (-9.29) (-10.16) (-4.95)
Casablanca – Settat − 6.49 4.90 − 3.72 − 4.78 8.59 2.67
(-6.49) (-1.58) (-5.30) (-10.08) (-1.49) (1.18)
Souss – Massa − 0.53 − 7.88 13.66 − 8.18 − 4.23 12.75
(-0.53) (-8.31) (5.35) (-2.83) (-7.06) (5.70)
Mean ¡10.75 ¡3.77 7.54 ¡6.70 3.04 0.67
(-10.75) (-14.51) (-6.97) (-13.67) (-10.63) (-9.97)
Median ¡11.65 ¡3.76 8.37 ¡6.10 3.16 1.69
(-11.65) (-17.39) (-6.78) (-11.29) (-10.16) (-6.39)
Notes: Admin-1 area boundaries conform to current regional boundaries based on the shapefile downloaded from: https://www.arcgis.com/home/item.html?id=21b
cbcaa915c433ba7c7850bafeede7b. No results reported for June 2020 due to stray-light contamination of the lights data.
Fig. 7. Changes in light intensity relative to pre-COVID-19 trend (percentage points), Admin-1 areas.
for the cities of Tangier and Tetouan, both of which are in the region of shows, masks out roads between cities. In the case of China, Liu et al.
Tanger – Tetouan – Al Hoceima. Hence, the decline in seasonally (2020) report that the dimming of lights associated with the pandemic
adjusted light intensity, relative to the pre-COVID-19 trend, that Tangier mainly occurred in the commercial centers of cities. Although further
suffered between February and March 2020 was 23.8 percentage points, work matching nighttime lights data with land use maps would be
which was nearly double the 12 percentage points decline that Tetouan required to establish this, it seems plausible that a similar story may hold
suffered. Furthermore, by September 2020, Tangier’s decline in light for Morocco.
intensity from February 2020 had further increased to 37.3 percentage
points, while Tetouan’s decline remained roughly the same as in March 5. Conclusion
2020 at 13.7 percentage points.
Comparing Table 4 with Tables 2 and 3, we can further see that the This paper has explored the potential for high frequency nighttime
declines in seasonally adjusted lights, relative to pre-COVID-19 trends, lights data to provide “near real-time” tracking of the economic impacts
are frequently larger at the city level than either at the Admin-1 region of the COVID-19 crisis for the specific case of Morocco. At the national
or national levels. This suggests that the negative economic “impacts” of level, there exists a strong correlation between quarterly movements in
the COVID-19 crisis in Morocco have been most heavily felt in cities Morocco’s overall light intensity and movements in its real GDP, which
subject again to the same, previously expressed caveat, that we cannot supports the use of lights to track the economic impacts of the COVID-19
entirely rule-out the possibility that our results are driven by unac crisis at both higher temporal frequencies and at the subnational level,
counted for (non-seasonal) atmospheric effects that distort the mea for which GDP data are unavailable. Consistent with large economic
surement of lights. It also suggests that our results at the national and impacts of the crisis, Morocco experienced a large drop, relative to the
Admin-1 region levels are not driven by the dimming of lights along, pre-COVID-19 crisis trend, in the overall intensity of its lights in March
for example, the transport routes that connect cities. This conclusion is 2020, from which it has subsequently struggled to recover. At the sub
corroborated by the fact that the results that we obtain at both the na national level, while all regions shared in March’s national decline in
tional and Admin-1 region levels are qualitatively similar when, instead nighttime light intensity, Rabat – Salé – Kénitra, Tanger – Tetouan – Al
of the EOG mask, we use the built-up area mask, which, as Fig. 1(d) Hoceima, and Fès – Meknès experienced much larger declines than
10
M. Roberts Development Engineering 6 (2021) 100067
Table 4
Changes in light intensity relative to pre-COVID-19 trend (percentage points), cities with 2015 population > 200,000.
Change from previous month, percentage points (Change from February 2020, percentage points)
City Admin-1 unit March April May July August Sept.
2020 2020 2020 2020 2020 2020
Tangier Tanger – Tetouan - Al Hoceima − 23.76 − 6.53 10.33 − 2.86 − 3.32 − 11.18
(-23.76) (-30.29) (-19.96) (-22.82) (-26.14) (-37.32)
Rabat Rabat – Salé – Kénitra − 12.27 − 2.93 2.04 − 21.30 16.88 − 14.07
(-12.27) (-15.20) (-13.16) (-34.46) (-17.58) (-31.65)
Fez Fès – Meknès − 23.23 − 14.94 10.31 − 4.14 4.68 − 0.52
(-23.23) (-38.16) (-27.85) (-31.99) (-27.31) (-27.83)
Meknes Fès – Meknès − 17.67 − 13.27 9.54 2.20 − 0.45 − 6.84
(-17.67) (-30.93) (-21.39) (-19.18) (-19.63) (-26.48)
Kénitra Rabat – Salé – Kénitra − 17.64 1.87 4.82 − 15.27 16.27 − 11.70
(-17.64) (-15.77) (-10.95) (-26.22) (-9.95) (-21.64)
Khouribga Béni Mellal – Khénifra − 14.87 − 4.21 7.92 2.61 3.01 − 11.03
(-14.87) (-19.08) (-11.16) (-8.56) (-5.55) (-16.58)
Marrakesh Marrakech – Safi − 12.45 − 17.05 12.92 − 6.76 3.81 2.95
(-12.45) (-29.50) (-16.58) (-23.33) (-19.52) (-16.57)
Tetouan Tanger – Tetouan - Al Hoceima − 12.03 − 5.91 − 2.33 − 10.36 28.39 − 11.46
(-12.03) (-17.94) (-20.27) (-30.63) (-2.24) (-13.70)
Oujda Oriental – RIF − 15.82 − 22.43 31.14 1.85 − 7.12 6.72
(-15.82) (-38.24) (-7.10) (-5.25) (-12.38) (-5.65)
El Jadida Casablanca – Settat − 13.51 3.15 − 4.31 3.36 3.59 7.14
(-13.51) (-10.36) (-14.67) (-11.30) (-7.71) (-0.57)
Nador Oriental – RIF − 10.47 − 4.50 0.59 4.44 6.57 4.81
(-10.47) (-14.97) (-14.39) (-9.94) (-3.38) (1.43)
Casablanca Casablanca – Settat − 9.70 11.08 − 4.62 − 9.39 10.65 3.62
(-9.70) (1.38) (-3.24) (-12.63) (-1.98) (1.64)
Safi Marrakech – Safi − 7.72 − 7.15 4.21 12.75 2.81 − 2.45
(-7.72) (-14.86) (-10.66) (2.10) (4.91) (2.46)
Agadir Souss – Massa − 1.27 − 3.59 10.21 − 8.50 − 5.02 17.31
(-1.27) (-4.86) (5.35) (-3.15) (-8.17) (9.14)
Mean ¡13.74 ¡6.17 6.63 ¡3.67 5.77 ¡1.91
(-13.74) (-19.91) (-13.29) (-16.96) (-11.19) (-13.10)
Median ¡12.98 ¡5.20 6.37 ¡3.50 3.70 ¡1.49
(-12.98) (-16.86) (-13.77) (-15.91) (-9.06) (-15.13)
Notes: Each city is defined as an “urban center” following the European Commission’s degree of urbanization methodology (Dijkstra and Poelman, 2014; Dijkstra
et al., 2021). Table is restricted to urban centers with a population of more than 200,000 in 2015.
others. Since then, the relative effects of the COVID-19 shock across only imperfectly aligned with the timing of changes in the direction of
regions appear to have largely persisted. Similar, and, indeed, stronger, real GDP. Second, at both the national and sub-national levels, we have
persistence is also evident for our sample of Moroccan cities. Moreover, shown that the large decline in nighttime light intensity that occurred in
the effects of the COVID-19 crisis on lights appear to have been, in many March 2020 at the onset of the COVID-19 crisis in Morocco was preceded
cases, stronger at the city than either at the subnational regional or by an almost equally large increase in light intensity in February 2020.
national levels, which is consistent with cities bearing much of the brunt This raises the concern that our results may be driven by temporary
of the crisis. measurement error driven by unaccounted for atmospheric distortions
Our analysis can potentially be extended in several directions. Thus, that might have biased upwards the measurement of light intensity in
while we have analyzed the evolution of nighttime lights, as a proxy February. While the fact that there were also year-on-year declines in
measure of economic activity, at both the national and sub-national light intensity between March 2019 and March 2020 provides some
(regional and city) levels for Morocco, the methods showcased in this reassurance, further research is required to definitively establish that the
paper can equally be applied to other Middle East and North African estimated changes in light intensity reported are driven by anthropo
(MENA) countries, taking advantage of the region’s low cloud-cover genic causes. Third, and finally, we have seen that Morocco is particu
which facilitates frequent high-quality observation of lights. A second larly affected by the distortion of its June lights data due to the stray-
natural extension would be to investigate the underlying determinants light phenomenon, which led to us dropping data for this month.
of the heterogenous nighttime lights reactions of Morocco’s different Taken together, these caveats imply that the results presented in this
regions and cities to the COVID-19 crisis, which might include, for paper should be taken as suggestive of the evolution of economic activity
example, differences in their economic and demographic structures, as following the onset of the COVID-19 crisis rather than as providing the
well as differences in the spread of the disease itself and the associated final word on the magnitude of this evolution, where much more caution
policy reactions. A final potential extension would be to analyze the is needed. Ideally, one would like to triangulate the results presented in
evolution of lights in response to the crisis within Morocco’s cities in the this paper with those obtained from other high frequency, potentially
spirit of Liu et al.‘s work for China to see if the negative impacts of the near real-time, proxy measures of economic activity for which the
COVID-19 crisis are associated with particular geographic areas – e.g., possible measurement errors are orthogonal to those in the lights data.
commercial centers – within cities.
Finally, there are several crucial caveats that should be kept in mind Declaration of competing interest
when considering the results in this paper. First, while we have
demonstrated that quarterly changes in nighttime light intensity are The authors declare that they have no known competing financial
significantly correlated with quarterly movements in real GDP for interests or personal relationships that could have appeared to influence
Morocco, we have also found that this relationship contains significant the work reported in this paper.
noise and that the timing of changes in direction in light intensity are
11
M. Roberts Development Engineering 6 (2021) 100067
Acknowledgments anonymous referees for their helpful comments on an earlier draft of this
paper. The findings, interpretations, and conclusions expressed in this
The author thanks Hogeun Park and New Light Technologies – and paper are entirely those of the author. They do not necessarily represent
especially Ran Goldblatt – for their processing of the nighttime lights the views of the International Bank for Reconstruction and Develop
data. He also thanks Jesko Hentschel, Carole Megevand, Jaafar Sadok ment/World Bank and its affiliated organizations, or those of the Exec
Friaa, Larbi Jaïdi, Mounssif Aderkaoui, Amjad Khan, Augustin Maria, utive Directors of the World Bank or the governments they represent.
Chaymae Belouali, Somik Lall, Robert Beyer, Amina Coulibaly and three
Annex 1. Relationship between nighttime light intensity and sectoral GDP
Table A1
Relationship between nighttime lights and GDP generated by different sectors, quarterly data (2012 Q3 – 2020 Q1).
Time trend Time trend + quarter effects Year effects Year + quarter effects De-trended (time trend) De-trended (year effects)
[1a] [1b] [2a] [2b] [3a] [3b]
EOG mask
Industry
ln (SOL) 0.143 0.051 0.453*** 0.173** 0.143 0.453***
(1.49) (0.90) (3.91) (2.58) (1.52) (4.28)
R2 0.86 0.95 0.90 0.98 0.07 0.49
Services
ln (SOL) − 0.057 − 0.078* 0.141*** − 0.015 − 0.057 0.141***
(-1.07) (-2.05) (3.02) (-0.52) (-1.11) (3.78)
2
R 0.97 0.98 0.97 0.99 0.05 0.23
Agriculture
ln (SOL) 0.460** 0.533** 0.026 0.159 0.460*** 0.026
(2.11) (2.81) (0.21) (0.96) (2.15) (0.26)
R2 0.43 0.48 0.92 0.93 0.12 0.00
Cluster mask
Industry
ln (SOL) 0.143* 0.063 0.413*** 0.149* 0.143* 0.413***
(1.77) (1.11) (3.58) (2.00) (1.82) (4.48)
R2 0.86 0.95 0.90 0.97 0.10 0.48
Services
ln (SOL) − 0.013 − 0.03 0.137*** − 0.002 − 0.013 0.137***
(-0.39) (-1.04) (3.63) (-0.06) (-0.40) (4.55)
R2 0.97 0.98 0.97 0.99 0.00 0.252
Agriculture
ln (SOL) 0.335* 0.398** − 0.052 0.032 0.335* − 0.05
(1.72) (2.18) (-0.40) (0.19) (1.77) (-0.50)
2
R 0.40 0.46 0.92 0.93 0.09 0.01
Notes: ***, ** and * indicate significance at the 1, 5 and 10 percent levels respectively. t-values are reported in parentheses and are based on Newey-West standard
errors, assuming a first-order serial correlation process. Estimated constant terms in the regressions not reported.
Annex 2. Full results from estimation of equation (4) – EOG mask
Table A2
Full results from estimation of equation (4).
Dependent variable: ln(SOLi,m )
EOG mask
Time trend 0.005***
(16.02)
COVID £
March − 0.079**
(-2.52)
April − 0.109***
(-4.75)
May − 0.046**
(-2.48)
July − 0.115***
(-4.10)
August − 0.070**
(-2.49)
Sept. − 0.073***
(-3.06)
(continued on next page)
12
M. Roberts Development Engineering 6 (2021) 100067
Table A2 (continued )
Dependent variable: ln(SOLi,m )
EOG mask
Constant 13.411***
(578.63)
R2 0.868
F 7.73***
n 93
Notes: ***, ** and * indicate significance at
the 1, 5 and 10 percent levels respectively. t-
values are reported in parentheses and are
based on Newey-West standard errors,
assuming a first-order serial correlation pro
cess. F(12, 75) refers to an F-test of the joint
significance of all explanatory variables with
12 and 75 degrees of freedom. Regressions
also include month dummies (estimated co-
efficients not reported) but exclude June data
due to stray-light contamination of the data.
EOG mask excludes areas excluded in 2015
and 2016 annual VIIRS composites.
References Henderson, J.V., Storeygard, A., Weil, D.N., 2012. Measuring economic growth from
outer space. Am. Econ. Rev. 102 (2), 994–1028.
IMF, 2020a. World Economic Outlook: A Long and Difficult Ascent. October 2020.
Beyer, R.C.M., Franco-Bedoya, S., Galdo, V., 2021. Examining the Economic Impact of
International Monetary Fund, Washington, D.C.
COVID-19 in India through Daily Electricity Consumption and Nighttime Light
IMF, 2020b. World Economic Outlook: Tentative Stabilization, Sluggish Recovery? January
Intensity, vol. 140. World Development (April).
2020. International Monetary Fund, Washington, D.C.
Bleakley, H., Lin, J., 2012. Portage and path dependence. Q. J. Econ. 127 (2), 587–644.
Jedwab, R., Pereira, D., Roberts, M., 2021. Cities of workers, children or seniors? Stylized
Chen, X., Nordhaus, W., 2011. Using luminosity data as a proxy for economic statistics.
facts and possible implications for growth in a global sample of cities. Reg. Sci.
Proc. Natl. Acad. Sci. Unit. States Am. 108 (21), 8589–8594.
Urban Econ. https://doi.org/10.1016/j.regsciurbeco.2020.103610.
Chetty, R., Friedman, J., Hendren, N., Stepner, M., The Opportunity Insights Team, 2020.
Liu, Q., Sha, D., Liu, W., Houser, P., Zhang, L., Hou, R., Lan, H., Flynn, C., Lu, M., Hu, T.,
The Economic Impacts of COVID-19: Evidence from a New Public Database Built
Yang, C., 2020. Spatiotemporal patterns of COVID-19 impact on human activities
Using Private Sector Data. NBER Working Paper No. 27431, Boston, M.A.
and environment in mainland China using nighttime light and air quality data. Rem.
Chodorow-Reich, G., Gopinath, G., Mishra, P., Naraynan, A., 2020. Cash and the
Sens. 12 (10), 1576.
economy: evidence from India’s demonetization. Q. J. Econ. 135 (1), 57–103.
Masaki, T., Nakamura, S., Newhouse, D., 2020. How Is the COVID-19 Crisis Affecting
Elvidge, C.D., Baugh, K., Zhizhin, M., Hsu, F.-C., Ghosh, T., 2017. VIIRS night-time lights.
Nitrogen Dioxide Emissions in Sub-saharan Africa? World Bank Poverty & Equity
Int. J. Rem. Sens. 38 (1), 5860–5879.
Note Number 21 The World Bank, Washington, D.C.
Elvidge, C.D., Ghosh, T., Hsu, F.-C., Zhizhin, M., Bazilian, M., 2020. The dimming of
Newey, W.K., West, K.D., 1987. A simple, positive semi-definite heteroskedasticity and
lights in China during the COVID-19 pandemic. Rem. Sens. 12 (17).
autocorrelation consistent covariance matrix. Econometrica 55 (3), 703–708.
Findley, D.F., Monsell, B.C., Bell, W.R., Otto, M.C., Chen, B.-C., 1998. New capabilities
Roberts, M. (forthcoming). “Shining a light on urban and spatial development” in Ijjasz-
and methods of the X-12-ARIMA seasonal-adjustment program. J. Bus. Econ. Stat. 16
Vasquez, E.J. and A. Jha (eds.), Disruptive Technologies in Sustainable
(2), 127–152.
Development. Washington, D.C.: The World Bank.
Florczyk, A., Corbane, C., Schiavina, M., Pesaresi, M., Maffenini, L., Melchiorri, M.,
Roberts, M., 2018. The many dimensions of urbanization and the productivity of cities in
Politis, P., Sabo, F., Freire, S., Ehrlich, D., Kemper, T., Tommasi, P., Airaghi, D.,
Latin America and the caribbean. In: Ferreyra, M.M., Roberts, M. (Eds.), Raising the
Zanchetta, L., 2019. GHS Urban Centre Database 2015, Multitemporal and
Bar for Productive Cities in Latin America and the Caribbean. The World Bank,
Multidimensional Attributes, R2019A. European Commission, Joint Research Centre
Washington, D.C.
(JRC). https://data.jrc.ec.europa.eu/dataset/53473144-b88c-44bc-b4a3-458
Small, C., 2019. Multisensor characterization of urban morphology and network
3ed1f547e.
structure. Rem. Sens. 11 (18), 2162.
Gibson, J., 2020. Better night lights data, for longer. Oxf. Bull. Econ. Stat. https://doi.
Storeygard, A., 2016. Farther on down the road: transport costs, trade and urban growth.
org/10.1111/obes.12417.
Rev. Econ. Stud. 83 (3), 1263–1295.
Gibson, J., Olivia, S., Boe-Gibson, G., 2020. Night lights in economics: sources and uses.
Sutton, P.C., Elvidge, C.D., Ghosh, T., 2007. Estimation of gross domestic product at sub-
J. Econ. Surv. 34 (5), 955–980.
national scales using nighttime satellite imagery. Int. J. Ecol. Econ. Stat. 8 (S07),
Gibson, J., Olivia, S., Boe-Gibson, G., Li, C., 2021. Which night lights data should we use
5–21.
in economics, and where? J. Dev. Econ. https://doi.org/10.1016/j.
Woloszko, N., 2020. A Weekly Tracker of Activity Based on Machine Learning and
jdeveco.2020.102602.
Google Trends. OECD Publishing, Paris. OECD Economics Department Working
Goldblatt, R., Heilmann, K., Vaizman, Y., 2020. Can medium-resolution satellite imagery
Papers 1634.
measure economic activity at small geographies? Evidence from lansat in Vietnam.
World Bank, 2020. Open Nightlights Tutorial. https://worldbank.github.io/OpenNight
World Bank Econ. Rev. 34 (3), 635–653.
Lights/welcome.html.
Hale, T., Angrist, N., Cameron-Blake, E., Hallas, L., Kira, B., Majumdar, S., Petherick, A.,
World Bank, 2021. Global Economic Prospects: January 2021. The World Bank,
Phillips, T., Tatlow, H., Webster, S., 2020. Oxford COVID-19 Government Response
Washington, D.C.
Tracker. Blavatnik School of Government, University of Oxford, Oxford.
13
File and source
- File
- tracking-economic-activity-in-response-to-the-covid-19-crisis-using-nighttime-lights-the-case-of-morocco.pdf
- Size
- 4,105,030 bytes
- SHA-256
- 097139d9193508d5f1bddf7f42ed85971eea1d2561f61d510aab1647ac0faf97
- Original
- doi.org