Wall Street is an eight-block stretch through Lower Manhattan, synonymous with American finance. After Wall Street crashed, the Great Depression spurred Hollywood's Golden Age. Hollywood, on the opposite coast, is a four-square-mile neighborhood that is synonymous with American cinema. Industries have long clustered, as firms benefit from a shared labor pool and location advantages. Where do industries cluster today? And is the industry or the shared labor pool more important? In this week's newsletter, we look for geographic clusters in Revelio Labs's location-mapped data from professional online profiles.
Which industry groups are most geographically concentrated?
Think of an industry group with no employment clusters. Take retail, for example. In every town, the share of America’s retail workers based there is about the same as the share of America’s total workforce based there. The geographic concentration of retail is almost zero. However, when an industry group is geographically clustered, its workers are overrepresented in a few locations. The G-index, from Ellison and Glaeser (1997), measures this. It ranges from 0 to 1 and captures how unevenly an industry group's workers are distributed across locations, relative to all workers.
To compute the G-index, we use location-mapped employment records classified into around 100 industry groups. Based on this measure, the figure below shows which groups are the most geographically concentrated today. There are no big surprises in the rankings. Public and multilateral institutions (like the World Bank, IMF, and the Fed), as well as many NGOs, sit close to the seat of federal power. Resource extraction sits close to resources. The geographic concentration of finance, law, and entertainment, however, reflects early locational advantages reinforced by decades of agglomeration.
A G-index of 1 would mean every employee in an industry group works in a single location. It is a theoretical extreme and, in practice, even the most concentrated industry groups are spread across many locations. These small G-index values are unsurprising given the coarseness of our classification, which sorts workers into roughly 100 industry groups. At this level of aggregation, no single location could plausibly contain an entire industry group. Even public institutions like the Fed have branches across the country.
What is the most overrepresented industry in each state?
The G-index told us whether industries cluster, capturing the extent to which an industry group is overrepresented in a few locations. To see where industries cluster, we can flip this measure around. Instead, we can look at the different states and see which industries are most overrepresented there.
The map below shows, for each state, the industry with the highest employment share relative to its national employment share. Healthcare is the largest employer in most states, echoing our recent newsletter on its role in propping up the labor market. So, this normalization matters. Adjusting for industries’ national employment shares brings regional patterns into sharper relief and highlights local specializations.
Take Oregon: 1.45% of Oregonians work in athletic apparel, compared to just 0.27% of Americans overall. That’s a 5.4x overrepresentation, which is the highest of any industry in Oregon. How did that happen? In the 1970s, Nike established itself in Beaverton, near founder Phil Knight’s childhood home and alma mater. Nike grew and then, twenty years later, Adidas moved into Portland. Today, dozens of competitors employ workers in Oregon, benefiting from a shared pool of design talent and the diverse terrain used for product testing. Still, the big three players – Nike, Adidas and Columbia – account for 80% of athletic apparel employment in the state.
Underlying the industry specializations on this map are various such combinations of historical accident, location advantage and agglomerations. There’s government anchors: public administration in the DC area and secret research in Los Alamos, New Mexico. There’s natural resources: oilfields in the South, metals in the West and fertile soil in the Wheat Belt. And then, like Wall Street and Hollywood, there’s the nickname-worthy clusters sustained by agglomeration forces. Though, among these, some names sound more glamorous than others. Hartford, Connecticut is known as the "Insurance Capital of the World" and Wilmington, Delaware as the "Credit Card Capital of the World". And what about Dalton, Georgia? That would be the “Carpet Capital of the World”.
Where is Silicon Valley?
Tech's absence so far is quite surprising. It didn’t feature among the most concentrated industry groups and it wasn’t the most overrepresented industry in California. Yet, among nickname-worthy clusters, Silicon Valley is probably the most famous. In fact, Ellison and Glaeser (1997), from whom we adapt the G-index, open their paper with the sentence: “[t]his paper discusses the prevalence of Silicon Valley-style localizations of individual manufacturing industries in the United States.” So, where is the tech hub?
Perhaps the first question to ask is: what are tech hubs? We know that tech workers — developers, engineers, scientists, product managers — flock to places like Silicon Valley, often chasing jobs at FAANG companies. But, at their founding, these companies had little in common. Facebook was a social network, Amazon a bookstore, Apple a computer maker, Netflix a DVD rental, and Google a search engine. Tech start-ups also run the gamut: ride-hailing platforms, dating apps, AI labs, self-driving cars, medical devices. Looking at the goods and services they provide, tech companies don’t cohere into one industry. Yet, we think of them as an industry because, differences aside, their workers perform similar roles.
With that in mind, Revelio Labs’ role taxonomy, rather than the industry taxonomy, is better suited for locating these tech hubs. Once more, we compute the G-index with location-mapped employment records. This time, however, we classify workers into around 200 roles and identify which among them are most geographically concentrated. The bar chart shows that many of the most geographically concentrated roles are software-related. Comparing this to the industry rankings, the geographic concentration of software-related roles stands out more than the geographic concentration of any industry group intersecting with tech.
Now we can revisit the question of where the tech hubs are. Across the US, roughly 3.2% of workers are in software-related roles, a rate that several states far exceed. Below, we show the states that employ software-related workers at the highest rate, and how those workers are distributed across the industry groups. In Washington, home to Microsoft and Amazon, half the software-related workers are in enterprise technology or digital commerce. In California, home of Apple, electronics has a larger share than it does elsewhere. And in Texas, home of CrowdStrike, cybersecurity has a larger share than it does elsewhere. Their industry mixes differ, but they share the thing that makes a tech hub a tech hub: they employ tech workers.
The most geographically concentrated industry groups don't concentrate for the same reasons. Resource extraction is tied to a location, while finance is held together by agglomeration forces. Across the US, local specializations emerge from combinations of historical accident and location advantages, and some endure because firms benefit from a shared labor pool. These shared pools of labor can become very important. Today, tech clusters might be better characterized by the talent that they house, rather than the specific industries that they host.
