Nighttime Lights - How are they useful?
Author: Jamila Nigmatulina
It is often harder than it seems to measure and trace how much productivity is increasing in a place. It becomes even harder in countries where national, let alone subnational statistics, are poor. In such countries, it is already difficult to tell where people live and how fast the population is growing. It is even harder to answer relevant policy questions regarding urban planning and transportation needs. Night Lights data can potentially help us find answers to these questions.
Image and Data processing by NOAA's National Geophysical Data Center.
New techniques of photographing places on earth with satellites reveal an opportunity to learn about economic activity and movements of people on our planet. People use light to produce, construct and live in the modern world. The patterns of light, as well as the movement in these patterns over time, can guide us as to where people choose to live and produce and how their choices change.
Night Lights data is introduced in the field of economics by Henderson et al (2012), and Nordhouse and Chen (2011). Both papers show that night-lights data can supplement measures of economic activity in countries where national statistics are poor. They also give insights on developments in GDP and population on subnational level. After publishing this work, night lights data was used extensively in other economic research papers by amongst others Michalopoulos and Papaioannou (2013), Michalopoulos and Papaioannou (2014), Harari (2015), Pinkovskiy and Sala-i-Martin (2016) and Pinkovskiy (2016).
However, researchers found mixed evidence on whether this data can be used as a proxy for subnational economic activity and, most importantly, for output growth over time. To test a relationship, researchers used data from countries where national statistics are more reliable and detailed. It turned out that nighttime lights correlated well with establishment density and population (available in grids of just 250 meters) in Sweden (Mellander et al, 2015), but not with wages. Nighttime lights also did not appear to be a good proxy for country growth over time, especially for population, GDP and capital stock (Addison and Steward, 2015) in a big sample of countries. Further, looking at regions over time in Brazil, India, the US and Western Europe, Bickenbach et al (2013) reported unstable elasticities. Our challenge today is to understand why this is the case, and whether the night lights conceal more information than they give. Nonetheless, there are still useful ways to apply the nighttime lights.
Data availability and sources
There are three usable datasets from NASA: Stable lights (“Average Visible, Stable Lights, & Cloud Free Coverages”), unfiltered lights and radiance calibrated lights.
Most papers work with the stable lights as they show the luminosity of cities and towns and are available for a period of 22 years, between 1992 and 2013. The unfiltered lights are available for the same period and can potentially give more values at low luminosity, which is helpful for studies in areas with low levels of development, but it is often hard to disentangle it from natural illumination of the soil.
Radiance calibrated lights give more variation at high luminosity levels, as they are not top coded. Which is the process of censoring all values above a certain threshold. Radiance calibrated lights are available for a number or years, namely 1996, 1999, 2000, 2002, 2004, 2005 and 2010.
Processing Night Lights: a practical example
In our research programme we use night lights data for several projects. Among them is a project where we look at the relative growth of lights in the hinterland compared to the national capital after democratisation in Sub-Saharan Africa. We use nighttime lights in two ways:
1) To create real, rather than administrative boundaries of towns and cities
2) To measure the amount of light emitted by the hinterland and how much they grow more over time
Processing stable nighttime lights can be done with ArcGIS software and is further explained in this excellent document by Matt Lowe from MIT, outlining the standard processing of lights in more detail.
Figure 1. Dar-es-Salaam city boundary created using nighttime lights form 2008-2012.
In our project, firstly, we use night lights data to create city boundaries. We identify contiguously lit areas in the raw raster dataset in the years 2008-2012. We assume that these areas are towns and cities and convert them to polygons with ArcGIS software. The software helps to combine light values from different years and restrict the light information to the coastline and country borders. See figure 1 for an example of such a city boundary for Dar-es-Salaam.
Secondly, to measure the amount of light emitted by each location, we simply add the pixels inside the boundaries per year. This captures both extensive and intensive changes of the luminosity in cities. Even though some recent literature showed unstable elasticities of lights relative to distinct economic variables, the change of lights can still give some signal of a combination of population growth, electrification and output.
Issues: overglow, gas flares, and the aurora and zero lights
There are, however, some common issues with night lights data, namely overglow, gas flares and poor reading of low luminosity.
The first, overglow, is an issue for larger cities that emit a high intensity of light. The problem is less prominent in Africa, although for larger metropolises, an approximately 1-2km fringe of extra light is emitted. One solution is to use stricter thresholds, above the luminosity values of 10 or 20 for larger cities (but small cities may drop if a stricter threshold is used). This source outlines a possible solution to the overglow problem.
Not all lights emissions stand for pure economic activity in towns. Some areas in Nigeria, Angola, Cameroon, Chad, Democratic Republic of Congo, Gabon, Republic of Congo and Sudan (out of Sub-Saharan Africa countries), emit high luminosity because of gas flares. To make both urban boundaries and luminosity proxy better we must exclude these areas. NOAA provides gas flare boundaries to which we clip our city boundaries and we can use this boundaries to exclude pixels.
In Nigeria, the whole Niger Delta is contaminated, so it is impossible to distinguish between the gas flares and the activity of each city in this area for a sizeable number of cities. Therefore, only the city boundaries in the gas flared area can be backed out with a stricter luminosity threshold (of 30).
Many zero valued city-years appear in Africa, because NOAA filters the values below 5 due to their similarity to the background illumination of the earth. Thus, we have a censoring of lights at the value of around 5 for a large number of cities in Africa. As a warning to researchers studying the continent, in our research we find that the process of places shifting from unlit to lit over time overpowers the changes of the intensity of lights in the places that are already lit.
Economists are still working to extract the right signal from looking at the satellite photographs. We must get a better understanding of why the changes in lights in regions and countries over time do not perfectly predict the changes of their total economic activity and population. It may be that we have to wait for new sensors, such as Visible Infrared Imaging Radiometer Suite to get more precise readings of lights from the places on earth.
Nevertheless, the satellite photographs will give us a consistent monitor of where populations choose to move and where economies grow or decline. Such detailed knowledge will guide researchers and policy makers as to what helps development and what policies we need to promote it.
Addison, Douglas M., and Benjamin P. Stewart. Nighttime lights revisited: the use of nighttime lights data as a proxy for economic variables. No. 7496. The World Bank, 2015.
Bickenbach, F., Bode, E., Nunnenkamp, P., & Söder, M. (2016). Night lights and regional GDP. Review of World Economics, 152(2), 425-447.
Chen, Xi, and Nordhaus, William D. (2010). The value of luminosity data as a proxy for economic statistics. NBER working papers: 16317.
Country Boundaries: GADM (http://gadm.org/)
Harari, Mariaflavia. "Cities in bad shape: Urban geometry in india." Job market paper, Massachusetts Institute of Technology (2015).
Henderson, V., Storeygard, A., & Weil, D. N. (2011). Measuring Economic Growth from Outer Space. American Economic Review, 102 (2012 ), 994 – 1028.
Mellander, C., Lobo, J., Stolarick, K., & Matheson, Z. (2015). Night-time light data: A good proxy measure for economic activity?. PloS one, 10(10), e0139779.
Michalopoulos Stelios Papaioannou Elias, “Pre-Colonial Ethnic Institutions and Contemporary African Development,” Econometrica, 81 (2013), 113 – 152.
Michalopoulos Stelios Papaioannou Elias, “National Institutions and Sub-National Development in Africa,” Quarterly Journal of Economics, 129 (2014) 151 – 213.
Nighttime lights: Image and Data processing by NOAA's National Geophysical Data Center. DMSP data collected by the US Air Force Weather Agency.
Pinkovskiy, M., & Sala-i-Martin, X. (2016). Lights, camera,.income! illuminating the national accounts-household surveys debate. Quarterly Journal of Economics, 131(2), 579–631.
Pinkovskiy Maxim L., “Growth Discontinuities at Borders” forthcoming, Journal of Economic Growth