Published July 3, 2023 in Audio articles • 8 min read • 
Audio available
It is hard to know where you are going if you don’t have a decent map – either a physical map, or a mental map based on experience. First-time visitors to London can be forgiven for wondering why there are so many “High Streets” in the same city, or how they wound up in Westminster or St John’s Wood when they thought they were in London. The London Tube map – stylized, simplified, and necessarily inaccurate – turns out to be an indispensable guide for getting around a charmingly idiosyncratic city.
What happens when the tools and categories you use for your map no longer work? For much of the 20th century, you could reasonably describe the US economy using a simple set of nested categories; industries consisted of firms that operated establishments (factories, offices, stores) staffed by employees who engaged in occupations. Today, not so much: when crypto exchange Coinbase went public in 2021, its prospectus noted that it did not have a physical headquarters for its (minuscule) workforce because it was a remote-first company, and declared its industry to be “business services, not elsewhere classified”. We still use tools created during the Great Depression to map an economy in which work-from-home contractors directed by placeless “algos” contribute to virtual goods sold on global app stores.
Our outdated tools make it hard to answer even basic questions, much less create thoughtful policy. It’s time to create a better data architecture for mapping the 21st century economy – particularly when it comes to employment.
Twenty five years ago, Yale’s James C Scott published a remarkable book, Seeing Like a State: How Certain Schemes to Improve the Human Condition have Failed. The book described high modernism, where governments aim to make societies more legible and easier to manage by imposing simplified categories on a complex world. States like to have a synoptic view – like a London Underground map – and often seek to impose a map that does not fit. Sometimes these mapping efforts are benign – say, naming roads and numbering houses that make it easier to deliver mail. Other efforts to impose a map have been disastrous, from “urban renewal” that bulldozed thriving neighborhoods for highways to Stalin’s forced collectivization.
Scott describes an early version of high modernism with the birth of German forestry science in the 18th century. One of the important sources of royal wealth was ownership of forests, and forestry science was developed to help estimate how much board wood and cord wood could be produced each year. The forest contained much beyond this, but monarchs were most concerned with the revenue-producing parts of their holdings, and not so much the flowers, fruits, shrubs, mushrooms, and rodents. Of course, a forest would be a lot more efficient if it included only cash crops like Norway spruce or Scotch pine, planted in even rows and columns – easier to plant, manage, harvest, and count. On the other hand, a monocrop production forest is far more susceptible to disease and blight, which provides a telling analogy for other high modernist efforts.
Those who design our world often aim to make it legible and easy to navigate. A typical Walmart store contains 142,000 different products on its shelves, and yet most shoppers can find what they are looking for thanks to the intuitive layout of the store, with different sections for dairy products, men’s clothing, sporting goods, electronics, and so on. This is high modernism in action.
Contrast this with an Amazon warehouse, which is a riot of items with no obvious categorization in place. As Amazon describes it: “Instead of storing items as a retail store would – electronics on one aisle, books on another – all of the inventory at Amazon fulfillment centers is stowed randomly. Yellow-tiered ‘pods’ stack bins full of unrelated items, all of them tracked by computers. This counterintuitive method makes it easier for associates to quickly pick and pack a wide variety of products.”
Information and communication technologies (ICTs) mean that the warehouse doesn’t need a map readable by humans. Technology has made a map superfluous – as if the German production forest had lapsed back into a jungle.
The same process has happened to the economy as ICTs reshuffle the basic categories we use to create our economic maps. Government efforts to map American industries date back to the Napoleonic Wars and the initiation of a census of manufacturing establishments. But our contemporary mapping system was created during the Great Depression, when the need for actionable data to guide economic policy was evident.
For information about corporations, the big breakthrough was the passage of the Securities Exchange Act of 1934, which created the Securities and Exchange Commission (SEC) and empowered it to require listed corporations to disclose balance sheets and income statements annually to investors and the broader public – the origin of the Form 10K we all know and love today. Three years later, the Central Statistical Board launched a committee to create a system of industry categories, which led to the creation of the Standard Industrial Classification (SIC) system, launched in 1939 and updated regularly until 1987. The same year, the US Employment Service published the first Dictionary of Occupational Titles (DOT), classifying occupations based on extensive fieldwork.
The SIC system was replaced by the North American Industry Classification System (NAICS) in 1997, although SIC codes are still in wide use today. The DOT was replaced by the Occupational Information Network (O*NET) in 1998. But despite valiant efforts to stay up to date, there are some basic ontological issues with our categories. Moreover, the fact that several diverse and often uncoordinated government agencies are responsible for collecting, cataloging, and sharing economic data means that standardization is often low.
We still use tools created during the Great Depression to map an economy in which work-from-home contractors directed by placeless ‘algos’ contribute to virtual goods sold on global app stores.
What, exactly, is an “industry” today? DoorDash is a food delivery platform that dispatches drivers, and Olo is an online ordering and payment platform that conveys orders for dine-in and takeout customers. You might imagine that these companies are in the restaurant industry, or at least restaurant-adjacent. But Compustat classifies them as “Web Search Portals and All Other Information Services”, where they join Bumble (dating), Qualtrics (online surveys), and Buzzfeed (viral content – don’t ask). Meanwhile, Apple is “Radio and Television Broadcasting and Wireless Communications Equipment Manufacturing”, Netflix is “Video Tape and Disc Rental”, and Amazon is “Electronic Shopping and Mail-Order Houses”. Not exactly wrong, but not very helpful either.
Tech firms, particularly platforms, just don’t fit the category schemes of the SIC and NAICS system, and “platform” is not really an industry. Basic ideas like industry concentration don’t scan for these companies – they are clearly upending how competition works, but not in ways contemplated by our old system.
The difficulty is especially acute when it comes to employment. We know surprisingly little about the American workforce. Companies that list shares on the stock market have to disclose a vast catalog of financial information every year, detailed in income statements, balance sheets, letters from management, and endless footnotes. What do they tell you about the people who work there (and who are not top executives or board members)? Until recently, two things: how many employees they have in total, and how many are union members. In recent years, we also got to know what the median annual compensation was (on the proxy statement), and a bit about the company’s employment philosophy (in a mostly uninformative section of the annual report called “Human Capital”).
What if you wanted to know how many contractors and temporary workers are retained by the company? No dice. In 2019, it was revealed that Alphabet had substantially more temps, vendors, and contractors (TVCs) than employees, but this information only came out due to a leak. Given the minuscule (reported) workforces of younger tech companies, it’s a safe bet that large sections of the tech economy’s employment are effectively invisible. How about turnover rates, investments in training, or safety records? Nope. Without a mandate, we have no real way to find out.
Of course, the COVID-19 pandemic and the large-scale move to work-from-home has encouraged more companies to consider switching even more work from employees to contractors, and the revolution underway sparked by the release of the latest version of ChatGPT and other AI tools seems likely to lead to large-scale job displacement – at least if Wall Street’s valuations are good predictors. It seems highly likely that many workers today are juggling multiple “jobs” online and offline, and there will be more of that in the future. But given the way we collect data, we really don’t know.
Are industries becoming more concentrated, as many have claimed? It is almost impossible to know in a world where tech platforms reshape competition and the very definition of an industry.
Imagine finding your way around Berlin today with a map from 1988. That is the situation we are in now. Are industries becoming more concentrated, as many have claimed? It is almost impossible to know in a world where tech platforms reshape competition and the very definition of an industry (e.g., Airbnb and hotels, Uber and taxis, DoorDash and restaurants). This makes competition policy complicated. How much of the labor force consists of TVCs (who, in the US, typically do not get employer-provided healthcare and other benefits)? Is gig work a small slice of the labor force or mostly a part-time side hustle? Is it a major part of employment today? Again, it’s complicated – and until we have better data, we will not be able to create smart policy.
What is to be done? Given how outdated our tools and categories are, it’s an opportune moment for a significant re-thinking of our methods of mapping the economy. One promising portent is a petition to the SEC seeking more comprehensive disclosures on the workforce (filed by a group of scholars that included me), which is supported by a large group of investors seeking to figure out how to evaluate companies that are light on tangible assets but heavy on knowledge and human capital. It would only apply to listed corporations (a dwindling part of the American economy) – but at least it’s a start.
Research by the Boston Consulting Group exposes an inconvenient truth about longevity: awareness has never been higher, but action lags far behind. Here's how to turn aspirations into reality....
Audio available
The US President is gouging out chunks of America’s long-established economic data apparatus because he doesn’t like what he sees. It’s unlikely to end well....
Audio available
Understanding what makes your brain and body perform at their optimum is the foundation for long-term performance....
Audio available
Increasing longevity is forcing a rethink of public policy, individual lifestyles, and private finances. Yet it is crucial not to overreact or grow unduly alarmed about shifting population dynamics....
Audio availableExplore first person business intelligence from top minds curated for a global executive audience