Quantifying knowledge and why it matters for new companies

Cristian Jara-Figueroa
Published in
5 min readDec 14, 2018


Photo by Samson Creative. on Unsplash

Back in the 1960s, instead of being the high-technology producer we know today, Tokyo imported almost everything — even bicycles. The small shops that specialized in repairing those imported bicycles branched out by making bikes of their own. By combining the right set of skills and processes, they created a new industry that went on to create other industries of its own. There is nothing special about bikes, but they illustrate a simple principle of how cities create new industries when the right knowledge is available. Just like the semiconductors in South Korea in the 1970s, bicycle manufacturing in Tokyo was not new-to-the-world, but it was new-to-the-place and thus, had the ability to transform the productive structure of the city by breeding new knowledge out of old knowledge. In order for this process to happen, however, and for these pioneer companies to succeed, cities need workers with the right knowledge. But what is the right knowledge?

Since Adam Smith’s description of the pin factory in his 1776 book “The Wealth of Nations,” the role of knowledge in economic growth had been somewhat of a fringe idea until the work of Paul Romer in the early 90s. Knowledge is often conceptualized as an amount of something. We often say phrases like “that person knows a lot.” Yet, there is something odd about the idea of “a lot” of knowledge. Does, for example, Marie Curie, have more knowledge than Michael Jackson? Maybe about chemistry, but I’m pretty sure that if you asked Curie to organize the mid-time show for the Super Bowl, it would not be very good. What this example illustrates is that knowledge is better conceptualized not as a measure of quantity but as a measure of relatedness. Curie knows more about chemistry than Jackson while Jackson knows more about performing arts than Curie. The same is true for less accomplished individuals, for whom we can infer what they know based on their experience. In our recent paper, we build on this conceptualization to study how knowledge impacts the chances of survival of newly formed companies.

We use the work experience of all the employees that participate in the formal sector economy of Brazil for a ten-year period to study how the knowledge of the first hires of a new company impacts the company’s chance of survival. We focus on pioneer firms and their first hires. Pioneer firms are the first firms in an industry to operate in a city: think about that first company to manufacture bicycles from the example at the beginning.

Figure 1: Using the work histories of more than 30 million employees of the formal sector economy of Brazil for a period of 10 years, we can quantify how much industry-specific knowledge and how much occupation-specific knowledge a pioneer firm has. If the firm hires people from related industries, it has a lot of industry-specific knowledge, and if it hires people from the same or from similar occupations, it has a lot of occupation-specific knowledge.

By borrowing methods from network science, we infer how much knowledge about, for example, car manufacturing a worker has if her work experience is mainly in the motorcycle industry. In the same manner, we can infer how much knowledge about cooking a waiter has. Figure 2 illustrates this technique, which gives us two different knowledge quantities: knowledge about the industry (industry-specific) and knowledge about the occupation (occupation-specific). For example, a salesperson coming from a car dealership has a lot of occupation-specific knowledge about sales and a lot of industry-specific knowledge about the car industry.

Figure 2: Based on the mobility of workers between pairs of industries (left) and pairs of occupations (right) we can infer how related two industries and two occupations are. This quantity reveals a network of knowledge-relatedness.
Figure 3: The chance of survival depends much more on the level of industry-specific knowledge that the pioneer firm has than on its level of occupation-specific knowledge. The curve for the survival rate of pioneers with high levels of industry-specific knowledge remains higher than the curve for the survival rate of pioneers with low levels of industry-specific knowledge (left), while the curves for the survival rates of pioneers with different levels of occupation-specific knowledge pretty much track each other (right).

Our main finding is that pioneer firms that hire workers with industry-specific knowledge are much more likely to survive than pioneers that hire workers with occupation-specific knowledge. We also find that the difference between occupation-specific and industry-specific knowledge becomes much smaller when looking at the survival of companies that are not pioneers, suggesting that hiring workers with industry experience is more important when pioneering an economic activity.

One way of conceptualizing the difference between industry-specific and occupation-specific knowledge can be found in the evolutionary economics concept of organizational routines. Routines are collections of tasks carried out by independent actors and are specific to a firm. Firms from similar industries are more likely to share similar routines. Two workers with the same occupation in different industries are likely to perform similar tasks that fit differently into different routines. In simple terms, a database engineer for a bank has different industry-specific knowledge than a database engineer in a hospital not because the task of maintaining the database is vastly different, but because the role that this task plays in the routines of each firm is different. Moreover, the institutional framework in which they operate is vastly different. Under this light, our results suggest that when pioneering economic activities, knowing how the tasks fit together and how they relate to the institutional framework is more important than knowing the task itself perfectly.

Although our data is specific to Brazil, there are reasons to believe that our results might generalize. Brazil is a vastly heterogenous country, which makes it an interesting scenario for studying industrial development since it combines the challenges of middle-income countries with the data reporting quality of high-income countries. The richest Brazilian region had an average income per capita in 2013 of about 30 thousand in 2018 U.S. dollars, which is comparable to that of Spain, Italy, or South Korea; while the poorest microregions had an average income of about 5 thousand dollars, which is comparable to that of Paraguay, Jamaica, or Algeria.

Finally, the importance of knowledge for economic growth was emphasized with this year’s Nobel Prize in Economics, awarded to Paul Romer for “integrating technological innovations into long-run macroeconomic analysis.” Before Romer, as David Warsh pointed in his book “Knowledge and the Wealth of Nations,” goods were thought to be made by combining labor and capital: people and things. Romer added a third component: ideas.

In the same way that Tokyo’s bicycle-repair shops used their knowledge to develop a new industry, other cities will use their own local knowledge to develop new economic activities. As cities grow and economic activities become more knowledge-intensive, new methods to quantify knowledge can help by expanding the empirical toolbox available to researchers, policy makers, and entrepreneurs to understand the role of knowledge in growing economies.

The research cited in this post was conducted by researchers from The MIT Media Lab and Harvard University. For more details, please see the PNAS publication.



Cristian Jara-Figueroa

PhD student at the MIT Media Lab working on the economic value of knowledge and its implications for regional industrial diversification. @Cristian_jf