Eyes in the Sky: How Satellite Data Transforms Economic Modeling

The International Monetary Fund (IMF) recently unveiled its PortWatch platform, a groundbreaking application resulting from a collaborative effort with partners such as the World Bank, Oxford University, the United Nations, the World Trade Organization, and ESRI, which offers the ArcGIS geospatial data analysis software.

PortWatch harnesses the power of satellite imagery and satellite enabled Automatic Identification System (AIS) fused with economic databases and models to anticipate the impact of climatic events on ports, maritime routes and their corresponding spillover effect on trading partners. This innovative platform illustrates the transformative potential of satellite data for economists worldwide, spanning academia, financial institutions, and development organizations.

It comes as no surprise that economists would turn to satellite data as a source of input for economic research. Perennially driven by an ability to innovate in leveraging databases and building analytical models, economists often face persistent challenges caused by data gaps emanating from insufficient historical accuracy or difficulties in field observations to consolidate databases.

Herein lies the relevance and significance of satellite imagery, offering economists three distinct advantages. First, it provides them with an unprecedented access to data that would be too challenging or impossible to acquire. The increase in resolution from satellite images also helps economists build better proxies out of identifiable features, with additional granularity helping build proxy databases down to local or municipal economic areas. Finally, data is getting easier and easier to access and to integrate thanks to free and open platforms like Google Earth Engine or the Copernicus Open Access Hub.

The benefits extend further, as economists can now make use of the advantages brought by satellite data to improve the performance of their economic and econometric models. Satellite imagery can facilitate the collection of panel data at a low marginal cost, on a regular basis, and at a large scale—data that would be prohibitively expensive to measure accurately without satellite technology. In addition, satellite imagery can greatly improve data consistency and reliability, notably in regions where official statistics may be susceptible to manipulations or misreporting in times of war or corrupt bureaucracies, which in turn facilitates the creation of robust panel data and time series databases. Finally, the wider geographical coverage of a satellite image ensures observations are consistent and reliable across vast territories, allowing for comprehensive analyses and economic modeling.

Such perks have led economists to mobilize satellite imagery in several recent empirical works, from research in development economics to financial markets analysis.

For about a decade now, development economists have been using satellite-based night lights imagery of the Earth as an invaluable tool to gauge economic activity and map poverty in conflict zones or countries with faulty statistical systems. Using territorial light intensity as a proxy for economic activity has allowed them to more accurately map and understand the actual extent of GDP losses resulting from conflicts or monitor the impact of economic isolation, something the comparison between the North and South Korean night lights eloquently display.

However, with improvements in satellite imagery resolutions, economists are now transitioning to daytime satellite data that can more easily be used in conjunction with machine learning due to the increased number of identifiable features. Deploying such techniques allows for the identification of surface groups acting as proxies for economic activity. This shift has enabled precise predictions of economic activity at smaller regional levels over extended time horizons.

An example of applying this data to econometric models includes research that identified the impact of higher education institutions on innovation output in East and Western Germany, an analysis otherwise impossible due to lack of reliable data from cold war Eastern European economies. More broadly speaking, at smaller regional levels, these new data inputs facilitate the analysis of historical economic developments and the evaluation of public policy reforms.

In another recent publication, researchers used satellite imagery to monitor the number of containers in ports and accurately predict stock index returns in 27 out of 33 countries on a daily basis. The study concluded that investors using satellite imagery of seaports will on average receive annualized returns of 16 percent, using the predictive relationship between port container number and economic activity. To achieve these results, their model looked at the predictive ability of the number of containers in relation to the growth of industrial production and compared to traditional economic data such as industrial production and consumer price indices (CPI), satellite imagery offers greater frequency and is available in quasi real time, offering greater predictive power.

The new dimension of IMF’s PortWatch is that it offers more directly actionable insights that build on the combination of insights from satellite data, added to the econometric models of leading international financial and economic institutions. This marriage of satellite data with economic models could mark a paradigm shift in economic analysis. From closing data gaps to insightful predictions of economic trends, satellite data is emerging as a new tool to shape the future of economic modeling. VS Mathieu Luinaud is a manager in strategy consulting within the PwC Space Practice and a senior lecturer in Economics at Sciences Po Paris.