Under the Hood — The Computational Engine of Economic Development
César A. Hidalgo
For decades, economists have been demanding non-aggregate theories of economic growth and development. Perhaps Wassily Leontief said it first when he emphasized that a true understanding of the economy needs to look “under the hood” of economic aggregates. But the voice of Leontief, although no longer present, is still prevalent in his writings and that of others. Contemporaries of Leontief, such as Simon Kuznets, and more recent economists and scholars, such as Robert Lucas (1988), Esther Duflo and Abhijit Banerjee (2005) — and, of course, yours truly (Hausmann and Hidalgo 2011) — have also called for an understanding of the process of economic growth and development that avoids aggregation. But why?
The demand for non-aggregate theories of economic growth is easy to understand after considering the limitations of aggregation. Of course, we all know that — while useful to some extent — totals and averages provide only a coarse representation of complex systems, such as economies. But the limitations of our aggregative approaches transcend the abuse of aggregates because they also come from an unfortunate choice of units and language. Economics, being a discipline obsessed with prices, has pushed aggregations based on the language of commerce, translating everything into units of dollars, pesos, or pounds. Certainly, there is merit in the use of prices as a trick to facilitate aggregation, but prices are very much “over the hood” of economic systems. Under the hood, economies are made of people, objects and the ability of people to create objects, all of which can be powerfully described using the language of information and computation. Here, I will describe how we can use the language of computation and information to describe economic systems, and also, to obtain insights that are hard to come by using monetary descriptions of the economy.
The economy is made of people, networks of people and the things that people make. People and networks of people accumulate knowledge and knowhow, both individually and collectively, and they use that knowledge and knowhow to produce a variety of products that, in turn, augments people’s capacity to produce new products (Hidalgo 2015).
A traditional interpretation of products as physical capital would tell you that products are past production and would abstract products numerically based on a product’s cost or commercial value. Under the hood, however, products are made of order — or information. To understand this idea, imagine that you have just won a new Bugatti Veyron, a car worth roughly $2.5 million. Now imagine that you crash that Bugatti against a wall, escaping unharmed but totaling the car. Of course, the value of the Bugatti evaporated when you crashed it against the wall because this was not stored in its atoms, but rather in the way in which these atoms were arranged. And that physical order is information.
Under-the-hood products are made of information, which is better measured in bits than in dollars or euros. This means that the actions we use to make products are acts of computation. Of course, we often overlook the computational nature of economic activities, but making a sandwich, sorting socks, building a house, or writing a book; are acts of computation, because they are activities that involve rearranging the state of the world. No matter whether the rearrangement involves modifying synapses in your brain or sorting a pile of bricks, these rearrangements are technically acts of computation, as you are using energy to produce order or information. This tells us that the knowledge and knowhow that we accumulate as both individuals and as a society are nothing but the software that powers our economy’s computational capacity, and that the economy is nothing but a manifestation of the co-evolution of information and computation. Products are made of information, which we can measure in bits, and people execute computation, which we can measure in flops (floating-point operations per second).
But how can we measure the bits and flops of an economy?
One trick is to characterize the economy by focusing on the mix of products that economies make (Hidalgo and Hausmann 2009; Hausmann and Hidalgo 2014). The mix of products that an economy makes gives us an indication of its ability to produce order and, hence, it is a proxy of its collective computational capacity. Also, by looking at data on which countries (or regions) make which products, we can gauge the relative computational capacities of each and explain international differences in income.
As a first approximation consider the bilateral trade between Chile and Korea. Chile has a positive trade surplus with Korea, since it exports $4.86 billion to Korea and imports only $2.3 billion. Yet, when we look under the hood of these aggregates (Fig. 1 and 2), we realize that Chile exports atoms to Korea while it imports the way in which atoms are arranged. So while Chile has a positive balance of trade with Korea, it has a negative balance of information (and computation). Korea is a more sophisticated computer than Chile, as it has been able to integrate the global economy by selling information embodied in matter. Chile, on the other hand, sells the extraction of local rocks.
A statistical validation of the economic relevance of the language of information and computation involves using the ability of countries to make products to explain their incomes and future economic growth. Because of political, linguistic (Ronen et al. 2014) and geographic barriers, each country, city or region acts partly as a separate computer and expresses its ability to produce information in the mix of products that it makes.
Late in the last decade, I developed a mathematical technique that can be used to characterize an economy’s ability to produce products. This measure of economic complexity, which makes use of information about the diversity of countries and the ubiquity of products, explains a substantial fraction of a country’s level of income, but it also explains future economic growth. This is because countries that have a capacity to produce products (i.e., to compute information) that exceeds what would be expected given their current level of income tend to grow faster than those that don’t have that excess computational capacity. China and India, for instance, are countries that have a computational capacity comparable with that of countries ten times richer than they are — and are therefore doomed to grow.
Thus, by looking under the hood of economies — that is, by focusing on both the information embodied in products and countries’ ability to make products — we get a description of the economy that helps explain cross-country differences in income and economic growth. This is an approach that also brings our descriptions of the economy closer to the descriptions of other systems of organized complexity (Weaver 1948), since the language of information and computation is not only useful to describe the economy, but also other complex systems, such as the biological cell.
Yet, the value of a description centered on information and computation does not only lie in its ability to bring economics closer to the natural sciences; it also helps us value different aspects of the economy in the right way. We are all familiar with the cliché of the castaway holding a briefcase full of money on a desert island. Of course, money is useless for the castaway because there is nothing for him to buy. But, just as objects are a more fundamental form of economic value than currency, the ability to create objects is a more fundamental form of economic value than the objects themselves. It is the ability to make, computation, that determines the capacity of economic systems and what we find once we lift Leontief’s proverbial hood.
Banerjee, Abhijit, and Esther Duflo. “Growth Theory through the Lens of Development Economics.” Handbook of Economic Growth, Volume 1A, edited by Philippe Aghion and Steven Durlauf. Amsterdam: Elsevier, 2005: 473–552.
Hausmann, Ricardo, and César A. Hidalgo. “The Network Structure of Economic Output.” Journal of Economic Growth (16) 4: 309–342, 2011.
Hausmann, Ricardo, and César A. Hidalgo. The Atlas of Economic Complexity: Mapping Paths to Prosperity. Cambridge, MA: MIT Press, 2014.
Hidalgo, César A. (2015). Why Information Grows: The Evolution of Order, from Atoms to Economies. New York: Basic Books, 2015.
Hidalgo, César A., and Ricardo Hausmann. “The Building Blocks of Economic Complexity.” Proceedings of the National Academy of Sciences (106) 26: 10570–10575, 2009.
Leontief, Wassily. “The Structure of the U.S. Economy.” Scientific American (212) 4: 25–35, 1965.
Lucas, Robert E. “On the Mechanics of Economic Development.” Journal of Monetary Economics (22) 1: 3–42, 1988.
Ronen, Shahar, Bruno Gonçalves, Kevin Z. Hu, Alessandro Vespignani, Steven Pinker and César A. Hidalgo. “Links that speak: The global language network and its association with global fame.” Proceedings of the National Academy of Sciences (111) 52: E5616–E5622, 2014.
Weaver, Warren. “Science and complexity.” American Scientist (36) 4: 536–544, 1948.
 In the 1965 article “The Structure of the U.S. Economy,” Wassiliy Leontief wrote: “‘Gross national product,’ ‘Total output,’ ‘Value added by manufacture,’ ‘Personal consumption expenditures,’ ‘Federal Government expenditures,’ ‘Exports’ — these headings in the book of national accounts describe the familiar external features of the economic system. In recent years the students and the managers of the system have been confronted with many questions that cannot even be clearly posed in such aggregative terms. To answer them one must now look ‘under the hood’ at the inside workings of the system.”
 Robert Lucas (1988) argued that “a successful theory of development (or of anything else) has to involve more than aggregative modeling.”