Testing Economics from the Celestial Observatory

celestial-header.png

Economics is a field where theories are easy to create, and very difficult to test. It’s impractical, if not impossible, to run a controlled experiment on a scale large enough to measure the true effect of policy on economies. Looking at historical data has problems as well, as economic statistics are often too coarse or unreliable to model or investigate macroeconomic theories. But a new paper from Computation Institute researchers presents an extraterrestrial solution -- using satellite images of Earth to measure economic activity on a global scale, and even provide early warnings about economic and humanitarian crises.

The method, dubbed “the Celestial Observatory” in a new paper by authors Eamon Duede and Victor Zhorin, capitalizes upon the correlation between the amount of light and economic activity in a particular region. Economists have previously used satellite images to test how well “light spillage” at night predicts economic indicators such as gross domestic product of countries and regions. Duede and Zhorin took this method to the next step, using it as a rich dataset for testing the veracity of an enduring economic theory.

The project itself came together serendipitously within the Computation Institute. Zhorin, a research scientist with the Center for Robust Decision Making on Climate and Energy Policy (RDCEP), was searching for data on carbon emissions when he chanced upon a public dataset of satellite imagery spanning from 1992 to 2013, provided by the Earth Observation Group, NOAA National Geophysical Data Center. Realizing the potential of the data for economic research, he partnered with Eamon Duede, executive director at Knowledge Lab, on finding new ways to use this information.

For starters, the duo decided to utilize the full resolution of the images to study global economic activity on a minute scale. Previous studies using nighttime light images had aggregated the results to the state or country level, then compared the brightness to similarly aggregated economic statistics. But using resources from the University of Chicago Research Computing Center, Duede and Zhorin measured the brightness at individual pixels -- in some cases with resolution as high as 500 meters, less than two football fields -- to instead study the distribution of brightness and economic activity within countries, states, and even cities.

The researchers also borrowed a method from astronomy to visualize and measure changes in economic activity over time. In the largest analysis, they used all 21 years of the dataset to compare changes in light intensity around the entire world between 2013 and 1992. By normalizing the changes to the global average, they could identify areas of rapid growth (red) or slower growth and decline (blue). The resulting map aligned with many well-established economic observations, such as the rapid growth of China, the benefits of European Union membership for Eastern European nations, and the decline of the Rust Belt in the United States.

Annual average mean-centered change, ConUS. (top) 1993-2006 period, (bottom) 2007-2013 period. Blue-coded areas correspond to net negative dynamics (decay) and red-colored areas to net positive dynamics (growth).

Annual average mean-centered change, ConUS. (top) 1993-2006 period, (bottom) 2007-2013 period. Blue-coded areas correspond to net negative dynamics (decay) and red-colored areas to net positive dynamics (growth).

But excitingly, their approach allowed for analysis at much smaller scales as well. Within China, for example, the map revealed rapid growth along coastal cities but much slower growth in Northern regions, where some economists theorize that government subsidies are propping up struggling industries. More recent maps using the highest-resolution images can even visualize the economic activity of individual roads or neighborhoods, and observe the motion of large cargo ships.

The authors then put this ultra-detailed assay to use on the difficult matter of testing macroeconomic theories. For this paper, they looked at the theory of convergence -- in short, that the economies of poor countries will grow faster than the economies of rich countries, reducing disparities over time. Globally, the satellite images revealed growing convergence after the end of the Cold War. However, the worldwide financial crisis that began in 2008 reversed this trend, with increasing divergence for three years after.

The paper also looks at convergence and divergence within individual countries to see how sensitive they were to the financial crisis and other economic shocks, including wars and major policy changes. Changes in growth trajectories tracked with civil war in countries such as Syria, or national borders in regions such as Ireland/Northern Ireland, suggesting that these images could also serve as a sensitive warning system for humanitarian crisis and policy outcomes.

Because satellite imagery continues to increase in resolution and frequency, this method will likely become even more useful in the future -- perhaps providing close to real-time and block-by-block feedback on economic changes. While the size of these derived datasets currently remain difficult to work with or publish, the promise of the method as a physical measure for studying economic theories and policy should drive increased interest, granting more economists membership to the celestial observatory.

This article was originally posted via The Computation Institute.