The Horse Was an Economic Engine—Until It Wasn’t
A century ago, the horse wasn't a leisure animal; it was the undisputed engine of the American economy. In 1920, horses and mules powered 89% of draft power on U.S. farms, pulling plows and wagons. Cities, too, relied on their tireless backs for transport and deliveries.
This was a massive, living infrastructure. At its peak, America had roughly 26.7 million farm horses and mules. An entire industry revolved around them: over 440,000 Americans drove horse-drawn vehicles for a living in 1910, with another 100,000 working as stable hands, harness makers, and carriage builders.
Feeding this workforce was a monumental task, consuming about a quarter of America's cropland between 1880 and 1920—land that grew oats and hay instead of human food. Cities like New York, with 130,000 horses in 1900, contended with 2 to 4 million pounds of manure daily and the grim task of carting away 15,000 dead horses annually.
The internal combustion engine didn't just offer novelty; it offered relief. Tractors and motor vehicles solved the escalating logistical nightmares and immense resource demands of the horse economy, paving the way for a swift, brutal transition.
The Machines Didn’t Need a Better Horse
Machines didn’t need a better horse; they simply rendered the animal obsolete. The transition was swift and brutal. In New York City, cars outnumbered horses in traffic counts by 1912, and the city’s last horse-drawn streetcar ran in 1917. On farms, tractors exploded from about 246,000 in 1920 to more than 4.5 million by 1960.
This wasn't about improving the horse; it was about replacing it with a superior economic engine. A tractor didn't need feed, rest, or seasonal upkeep through winter, delivering more work hours at a lower marginal cost. Owners had every incentive to invest in machines rather than maintain animal power.
As machines took over, occupations shifted dramatically. Between 1910 and 1930, the number of horse drivers and stable hands plummeted by 75% and 90%, respectively. Simultaneously, chauffeurs and truck drivers surged from about 46,000 to over 800,000.
While some workers, like the Teamsters, adapted by learning to operate the new machines, the transition wasn't universally smooth. The horse couldn't learn to drive a tractor, and many human roles simply evaporated, leaving no equivalent for displaced workers. This stark reality underlines the brutal efficiency of technological displacement.
Why “Humans Aren’t Horses” May Not Be Enough
Of course, the immediate pushback to the "you are the next horse" thesis is that humans aren't horses. We adapt. When agricultural work automated, people moved to factories; when factories automated, we moved to offices. Unlike a horse, a human can learn new, more complex tasks as the economy evolves.
But what if the new tasks are precisely what AI excels at? McKinsey estimated that demonstrated technologies could theoretically automate 57% of American work hours. That’s not realized job loss, but it signals the unprecedented reach of AI into cognitive labor, which was once our exclusive domain.
The analogy has limits: humans are consumers with political agency, and past transitions did create new work. However, neither fact guarantees that new jobs will arrive quickly, or be accessible to displaced workers. While horses were simply economic inputs, humans also drive demand. For more on the current role of horses, see the National Economic Impact Study - American Horse Council.
The stakes are immense. If AI can perform cognitive work "better, faster, cheaper, and safer," as machines did for physical labor, then our comparative advantage erodes dramatically. This isn't just about retraining; it’s about the very nature of human economic relevance.
Enjoying this? Get one like it in your inbox each morning.
one email a day · unsubscribe in two clicks · no third-party tracking
The Real Question: Who Owns the Tractor?
The real question isn't whether humans can adapt, but who owns the new "tractor." We must distinguish between derived demand—the need for labor because it produces something—and essential demand, where human presence is valued intrinsically. Horses, once the ultimate source of "pull" for plows and wagons, were victims of derived demand.
Today, horses persist for sport, recreation, and therapy—because we like them, not because the economy needs them. This "leisure horse" model reveals a critical distinction: while essential demand for unique human services (like live performance or personal care) will endure, it’s unlikely to generate enough work to support a global labor market of billions. Niche demand doesn't equal widespread employment.
David Shapiro’s video proposes a stark alternative: if automation makes labor less critical to production, our focus must shift from retraining to distribution. Wider ownership of the machines and their output—the "tractors" of the AI age—could become more important than simply finding new tasks for dislocated workers. This isn’t a guaranteed solution, but a fundamental policy and distribution challenge that demands our attention.
Frequently Asked Questions
What does the horse analogy say about AI and jobs?
It warns that machines can eliminate demand for a type of labor, rather than simply changing how workers perform it.
Did tractors eliminate all jobs connected to horses?
They sharply reduced horse-related work, while some workers shifted into jobs such as driving trucks and operating machines.
Why do some economists reject the horse analogy?
They argue that people can learn new tasks, create demand as consumers, and influence technology through political and collective action.
What does the video propose as a response to AI automation?
It argues that broad ownership of automated systems could help people share in the value those systems produce.