But the explosion in big data and machine learning is providing us with tools that are helping us improve our understanding of these complex systems: data with the granularity needed to trace the identity of many components and the techniques needed to make more advanced predictions based on this data. For more than a decade now, the field of economic complexity has pioneered the use of machine learning and network science techniques to answer questions of economic growth, regional diversification, and inequality. While those questions are not new, the use of machine learning expands the toolbox methods by allowing researchers to avoid the need to aggregate data into broad groups. This helps account for the non-fungibility of physical and human capital, since surgeons cannot be replaced by pianists, nor are harvesting machines interchangeable with rollercoasters. These methods can capture an economics where structure is not limited by coarse concepts such as agriculture, manufacturing, and services: but where we can literally distinguish apples from oranges, and oranges from fertilisers, packaging, and farming equipment.
The data-rich approaches of economic complexity are increasingly embraced by policymakers around the world, because of their ability to explain changes in regional diversification patterns and explain international variations in inclusive green growth. These approaches are useful to those looking to understand the economic opportunities of narrowly defined regions and industries. Not what is good in general, but what is specific to the city or region of concern for a policymaker.
This can be intellectually refreshing, especially for the policymakers who have been told countless times to focus on good governance, education, and institutions, without too many specifics of what these broad concepts mean, or how they apply to their distinct conundrums. However, the realpolitik of their work requires estimates of the economic potential of specific regions and industries. Maps of the competitive landscape and their chances of success in any particular industry, research activity, or innovation. As a result, they look for estimates that take into consideration information such as a region’s specific portfolio of industries, export products, and patents — the kind of information that economic complexity methods can tackle well. These are not abstracted categorisations, but granular indices that give way to an augmented form of policymaking, where data-rich platforms empower economic analysts in their quest to fulfil specific economic objectives.
Consider the case ofData Mexico, the official economic data distribution and visualisation platform of Mexico’s secretary of the economy. The government of Mexico has been using Data Mexico for the last two years to train their embassies and regional governments: it allows analysts to quickly retrieve visualisations about thousands of municipalities, industries, and occupations, and provides them with the information they need to promote commercial relationships. Do you need to, say, meet with shoe manufacturers from Vietnam? Data Mexico has detailed information about the shoe industry in Mexico and about current commercial relationships between Mexico and Vietnam, and can inform your decision. In this way, by pairing up a data platform with professional analysts, the secretary of the economy of Mexico is demonstrating a new model of economic intelligence support that radiates from the central government to regional units and diplomatic missions.
Today, the field of economic complexity is growing fast as young scholars continue to enter the area, developing and introducing new methods and ideas. These advances involve both empirical work — such as more nuanced measures of complexity that are better at predicting growth, inequality, and emissions — and theoretical work involving deep connections with traditional economic growth literature. But more importantly, the field is helping bring together expertise from a range of disciplines, such as economic geography, computer science, and physics. This is how we can best understand the economy: not in theory, but in its natural complexity.
*Director, Center for Collective Learning, ANITI, University of Toulouse.
Hidalgo, César A., ‘Economic complexity theory and applications’, Nature Reviews Physics 3 (2021): 92, https://doi.org/10.1038/s42254-020-00275-1.
Hidalgo, C., Why information grows: The evolution of order, from atoms to economies, Basic Books (2015).
Conflict of Interest statement:
César Hidalgo is a co-founder and current CIO of Datawheel, a company developing economic intelligence platforms including Data Mexico, Data USA, and The Observatory of Economic Complexity.