However the imagery most economists currently use has not been updated since 2013 and, according to Economics Professor John Gibson, it is highly flawed in the way it shows light on earth. The commonly-used images come from the Defence Meteorological Satellite Program (DMSP) which was designed to observe clouds for short-term weather forecasts rather than lights on earth.
鈥淚f you look at a DMSP image, it gives the less lit areas more light than they truly have 鈥� so it pulls them up to the average 鈥� and the most brightly lit parts, it pulls them down to the average,鈥� Professor Gibson says. 鈥淎nd that's why London doesn't look any brighter than the brightest part of Oxford, even though in reality parts of London are three times brighter.鈥�
This means any estimates of economic activity using these data are very prone to error.
Fortunately, a better source of night lights information exists. Monthly data from the Suomi satellite鈥檚 Visible Infrared Imaging Radiometer Suite (VIIRS) are at least 45 times more precise than the DMSP and, according to Professor Gibson鈥檚 study, are 80% better at predicting underlying economic activity. They are also much more up to date, having been collected from 2012 through to today.
Despite the far better performance of the VIIRS data, they remain almost unused by economists.
鈥淭his better source of data is kind of out of reach to many people because it hasn't been nicely processed in the way that the bad data has,鈥� Professor Gibson says. 鈥淪o the bad data have been made available in a form that almost anyone can download from 1992-2013 as annual composites.鈥�
In comparison, the VIIRS data are released monthly, so they include temporary sources of night light 鈥� such as fires, fishing boats, and sports arenas 鈥� which make them harder to interpret. Ordinarily it would require much time for analysts to combine 12 monthly images into one, showing lights for the whole year.
But in his study, published in the Oxford Bulletin of Economics and Statistics and supported by New Zealand鈥檚 Marsden Fund, Professor Gibson came up with a simpler way to do it.
鈥淓ven though we don't have the resources of the US government, we can leverage off what they've already done where they cleaned the VIIRS data for 2015 and made a nice annual composite. We can use that as a guide to get rid of all these random bits of noise in the data for the other years.
鈥淎nd cleaning them up in this way makes them far better predictors of on-the-ground activity, because it enables us to see the actual things that are permanently lit.鈥�
Professor Gibson says it is time for economists to step up and use the best night lights data available. 鈥淕uiding policy is important work, and we need to ensure our estimates are as accurate as possible,鈥� he says.
The paper Better Night Lights Data, For Longer is available at
Night lights data are images captured via satellite showing artificially lit areas on earth. Economists often use these to measure urban growth and local population changes in places where on-the-ground statistics are not reliable or not easily available.
