Tractor vs. Horse: What Happened to 26.7 Million Working Animals

By Thorsten Meyer | November 4, 2026 | ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

Work After Work · Episode 3 · Arc 1: How the economy is built

In January 1918, almost 27 million horses and mules worked on American farms. They pulled the ploughs, the seed drills, the reapers and the wagons that hauled the harvest to town. They also ate. In those years, about a quarter of all the cropland American farmers harvested grew oats, corn and hay just to feed them.

Forty years later, most of them were gone. There was no strike and no protest. A machine had arrived that never got tired, and the horse economy quietly shut down. This is the first big automation story of the modern economy, and it comes with a warning label: some economists see the horse as a preview of what machines will do to human workers. In this episode we look at what actually happened to the horses, what happened to the people who worked beside them, how the same story played out in China, Germany and Japan, what today’s driverless tractors can and cannot do, and how far the horse analogy really carries.

Key takeaways

• The horse economy was huge. Horses and mules on US farms peaked at 26.7 million on 1 January 1918 (horses alone peaked in 1915 at 21.4 million). In 1915, about 93 million acres, more than a quarter of US harvested cropland, grew their feed.
• The tractor took over within one generation. About 1,000 tractors were on US farms in 1910, 246,000 in 1920 and 4.7 million in 1960. Around 1955, tractors outnumbered horses and mules for the first time.
• The number to remember: 26.7 million. By 1960 the count of farm horses and mules was down to 3.1 million, a fall of almost 90%. Nobody retrained them; farmers stopped breeding them.
• Farm work shrank, the economy grew. US farm employment peaked at 13.6 million in 1916 and fell to 4.5 million by 1970, while the US labor force doubled. The new jobs came in cities, so people had to move.
• The “peak horse” argument is an analogy, not a law. In 1983 Wassily Leontief compared workers to horses replaced by tractors. Leontief himself named a key difference: horses could not vote.
• Today’s farm robots are real but narrow. John Deere’s 16-camera autonomy kit (February 2025) does tillage on select large tractors. China says about 74% of ploughing, sowing and harvesting was mechanized in 2023.

Products Worth Considering

Samsung Galaxy Tab A7 Lite 8.7 32GB Wi‑Fi Android Tablet – Long‑Lasting Battery, Compact Slim Design, Sturdy Metal Frame (US Version)

AmazonView Latest Price

Samsung Galaxy Tab A11+ (6 GB RAM, 128 GB storage, expandable) – Gray

AmazonView Latest Price

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation

AmazonView Latest Price

As an affiliate, we earn on qualifying purchases.

The horse economy: 26.7 million engines

To see how big the change was, start with the animals. The US Census Bureau’s Historical Statistics of the United States, Colonial Times to 1970 (1975) reprints the Department of Agriculture’s yearly estimates of horses and mules on farms, taken every 1 January. Horses alone peaked in 1915 at 21.4 million. Mules kept growing a little longer, so the combined peak came in 1918: 26.7 million working animals (21.2 million horses plus 5.5 million mules).

These animals were the engines of the farm, and engines need fuel. The same statistics estimate how much harvested cropland was used to grow feed for horses and mules. In 1915 it was about 93 million acres, out of 340 million acres of harvested crops in total. That is more than a quarter of everything American farmers harvested, grown not for people but for the animals that did the pulling. In 1920 it was still 90 million acres.

That number explains why the tractor was so attractive. A horse eats whether it works or not, every day of the year. A tractor needs fuel only when it runs, and the land that fed the horse can grow something to sell.

Enter the tractor

Gasoline tractors existed before the First World War, but they were heavy, expensive and suited to large farms. In 1917, Henry Ford launched the Fordson. The Henry Ford museum describes it as “the first mass-produced and inexpensive lightweight tractor”, built “to meet the needs of small farmers”, and says it “rapidly became the most popular tractor in America”. On 21 February 1920, the 100,000th Fordson came off the assembly line.

The curve that followed is one of the steepest technology adoptions of the twentieth century (Historical Statistics, Series K 184, excluding steam and garden tractors):

Around 1955, the lines crossed: 4.35 million tractors against 4.31 million horses and mules. Two notes on the table: the tractor counts for 1920, 1930, 1940, 1950 and 1960 are census figures with slightly different dates, and from 1951 the Department of Agriculture counted horses and mules together.

The combine and the farm worker who fed 47 people

The tractor was only the first machine. The combine harvester, which cuts, threshes and cleans grain in one pass, spread just as fast: about 4,000 on US farms in 1920, 190,000 in 1940 and more than a million (1,042,000) in 1960. Corn pickers, pickup balers and milking machines followed, together with fertilizer and better seed.

The result is a number that sums up the whole century. The Department of Agriculture estimated how many people one farm worker supplied with farm products. In 1910 it was 7.1 people. In 1940, 10.7. In 1970, 47.1 (Series K 407; the figure includes people abroad supplied by exports).

The number to remember: 26.7 million

That is how many horses and mules worked on American farms on 1 January 1918. By 1960, the count was 3.1 million, down almost 90%. The Department of Agriculture stopped its yearly horse and mule estimates soon after.

What happened to them is less dramatic than it sounds and more uncomfortable than it looks. Nobody retrained them, and there was no mass event. Farmers simply stopped breeding replacements, and the working horse population aged out. The land that had grown their feed shrank from 93 million acres in 1915 to about 5 million acres in 1960. That land could now grow food for people. It was the hidden dividend of the tractor, and one reason American farm output kept rising while the farm workforce fell.

The “peak horse” argument, and where it breaks

In 1983, at a symposium of the National Academy of Engineering on the long-term impact of technology on employment, the Nobel-winning economist Wassily Leontief put the horse at the center of the debate about automation. Computers and robots, he argued, replace human mental functions the way mechanical power replaced physical ones. And then:

“The role of humans as the most important factor of production is bound to diminish—in the same way that the role of horses in agricultural production was first diminished and then eliminated by the introduction of tractors.”

He went further. The idea that a worker who loses a job in one industry will always find work in another, after retraining if needed, was in his words “as invalid as would be the assertion that horses who lost their jobs in transportation and agriculture could necessarily have been put to another economically productive use.” Today this line of thought is often called the “peak horse” argument: there may be a point after which machines need fewer and fewer of us, just as farms needed fewer and fewer horses.

It is a powerful analogy. But it is an analogy, and it has limits:

• Horses could not vote. Leontief made this point himself. The market, he wrote, “efficiently eliminated the use of horses in agriculture. Had horses had an opportunity to vote and join the Republican or Democratic party, that solution would have been very different from what it actually was.” People can change the rules: working hours, social insurance, education, taxes. Leontief’s own answer was a mix of shorter working time and income policies.
• People are also the customers. Horses were an input. People are both the input and the reason the economy exists: every machine ultimately produces something for a human to use. That does not guarantee jobs, but it changes the economics. (This point is my reading, not Leontief’s.)
• People can learn new work. Leontief was right that retraining is no guarantee. But it exists as an option, and millions of former farm workers did find other work. Horses never had that option.
• The horse population “adjusted” by not being born. That is the darkest part of the analogy, and the clearest sign of where it breaks. A society cannot solve technological change for people by having fewer of them. It has to solve it through income, work and place.

What happened to the farm workers

The people did not vanish. They moved. According to the Department of Agriculture’s estimates, US farm employment, counting farm operators, unpaid family members and hired workers, peaked in 1916 at 13.6 million. In 1950 it was 9.9 million. By 1970 it was 4.5 million, about a third of the peak (Series K 174).

At the same time, millions of Americans left the countryside. Between 1910 and 1970, an estimated 6 million Black Americans left the South for cities in the North and West, according to the US Census Bureau, which calls it the Great Migration. The Census Bureau’s own teaching material names several reasons: economic opportunities in industrial centers, Jim Crow discrimination and harassment in the South, and the wish to follow friends and family who had already gone. During the First World War, factories in cities such as Chicago, Detroit, New York and Philadelphia were short of workers.

Where do machines fit into this? The easy story is “machines pushed people off the land”. The research suggests it ran both ways. In a study published in the American Economic Review in 2014, economists Richard Hornbeck and Suresh Naidu looked at the great Mississippi flood of 1927. Flooded counties saw an immediate and lasting out-migration of Black workers. In the years after, landowners in those counties mechanized and made farming more capital-intensive than landowners in comparable counties that had not flooded. Landowners had resisted the out-migration, because they benefited from the old labor-intensive system. In other words: sometimes the machine pushed people out, and sometimes people leaving pulled the machine in.

The same story on different clocks: China, Germany, Japan

Every rich country went through this shift, at different times and speeds. China shows it most clearly, because it happened within living memory. Here is China’s sector clock, from the National Bureau of Statistics (year-end employment; the series has a break in 1990, when the figures were revised after the census):

• 1952: 83.5% of workers in the primary sector (mostly farming), 7.4% in industry, 9.1% in services.
• 1978, when the reforms began: 70.5% primary.
• 2000: exactly 50.0% primary.
• 2024: 22.2% primary, 29.0% industry, 48.8% services. Services overtook farming in 2011.

Even at 22.2%, that is about 163 million people in China’s primary sector, so the story there is far from over. China’s farm ministry reports that in 2023, 74.29% of the work of ploughing, sowing and harvesting was done by machine, and 97.81% for wheat. The country had about 21.1 million tractors.

West Germany went through the change earlier. In 1955, 18.5% of workers in West Germany were in farming, forestry and fishing, according to Destatis. In 2025, in united Germany, it was 1.2%.

Japan shows what may be the next chapter. Its number of core farmers (people whose main job is farming their own household’s land) fell from 1.36 million in 2020 to 1.04 million in 2025, a drop of 24% in five years, according to the Ministry of Agriculture, Forestry and Fisheries’ agricultural census. Their average age is 67.7, and about 70% are 65 or older. In Japan, machines are not pushing people off the land. They are filling the gap that people leave behind.

Today’s farm robots, without the hype

The driverless tractor is no longer a concept. John Deere first showed a fully autonomous tractor at the CES technology show in 2022. On 6 January 2025, it presented a second-generation autonomy kit with 16 cameras for a 360-degree view, for tillage on its large 9RX tractors, plus an orchard tractor, a quarry dump truck and a commercial mower. On 27 February 2025, Deere released the kit as a precision upgrade for model-year 2022 and newer 9R/9RX and 2020.5 and newer 8R/8RX tractors, with availability through dealers from fall 2025. The farmer starts and watches the machine through a phone app.

Deere frames this as an answer to a labor shortage, not as a way to cut farm jobs: “there is not enough available and skilled labor to do the work,” its chief technology officer said in the January 2025 release. That is the company’s framing, and it fits Japan’s numbers. It is also worth being precise about what these machines do today. The autonomy Deere describes covers specific tasks, starting with tillage (preparing the soil before planting), on specific machines. A person still decides what to grow, sets up the field, fixes what breaks and carries the risk. The horse was replaced by a machine that did every pulling job on the farm. Today’s robots replace one task at a time.

Myth check: “Machines destroyed rural jobs, and nothing replaced them”

Verdict: partly true.

• The true part: American farm jobs fell by two thirds, from 13.6 million in 1916 to 4.5 million in 1970. Many farm towns lost the work that held them together, and for many families the only way forward was to leave.
• The false part: the jobs did not disappear from the economy. Over roughly the same period, the US labor force doubled, from about 37.5 million in 1910 to 74.1 million in 1960 (Historical Statistics, Series D 167, Lebergott’s estimates). The new jobs came in factories, shops and offices, mostly in cities.
• The catch: “replaced” is doing a lot of work in that sentence. The new jobs were in different places and needed different skills. For the economy, farm mechanization was an enormous gain. For a given village or family, it often meant a move, and for millions of Black Southerners, that move was also an escape.

What it means for businesses

• Plan for the work around the machine. The tractor did not just replace the horse. It needed dealers, mechanics, fuel, spare parts and credit, and it changed which farms could compete. AI is the same: before you count the jobs it replaces, map the new work it creates around it, from checking outputs to running the systems.
• Decide early who does that new work. Ideally, the people you already have. They know the business, and moving them is usually cheaper than hiring from outside and dealing with the fallout of layoffs.
• Automation follows scarcity. Deere’s pitch and Japan’s farm census point the same way: machines arrive fastest where people are missing. If your industry struggles to hire, automation pressure will be strongest there, and the case for it is easier to make to your own staff.

What it means for you

• Mobility beats panic. As I read this history, the people with the most options were the ones who could move, to a new place or into a new skill, before they had to. You may not need to move cities. You may need to move tasks.
• Watch the task, not the job title. Today’s farm robot takes one task (tillage), not the whole job of farming. AI tends to work the same way. Ask which of your tasks could be handed to a machine first, and which remaining tasks become more valuable.
• Remember the horses’ real disadvantage. It was not strength or speed. It was that they had no voice. People do: in their companies, their unions and associations, and their votes.

Thorsten’s thesis: what farming says about it

My thesis for this season is that automation could become its own sector, up to 90% of the economy within a decade or a few decades. That is my thesis, not a fact, and this season tests it with evidence for and against.

Farming is the strongest evidence for it: from about three in four American workers around 1800 to fewer than two in a hundred today, with output far higher. If any sector shows how far automation can go, it is this one. But farming is also evidence against the “no human hand” part: even the most automated farm in 2026 still has a farmer, and the newest machines automate one task at a time. Episode 48 will test the thesis in full.

Limits of this episode

• Long series, changing definitions. The farm statistics cover a century in which the definition of a farm, of farm employment and of the labor force changed. From 1951, horses and mules were counted together. China’s employment series has a break in 1990. Germany’s figures are West Germany until 1990.
• Estimates, not counts. The horse and mule numbers are Department of Agriculture estimates as of 1 January, checked against census years. The feed-acreage figures are calculated from assumed rations per animal.
• Two kinds of farm-worker counts. Episode 2 uses Stanley Lebergott’s labor-force series (Historical Statistics, Series D 170: about 11.8 million people employed in agriculture in 1910). This episode uses the Department of Agriculture’s farm-employment estimate (Series K 174: 13.6 million in 1916), which also counts unpaid family members working on farms. The two series measure different things, so their levels differ.
• Causes of migration. Farm machines were one force among many in the Great Migration. This episode does not claim that tractors caused it.
• Company claims. Deere’s descriptions of its autonomy products and the labor shortage are the company’s own statements.

Series navigation

• Previous episode: Episode 2, When Almost Everyone Farmed.
• Next episode (tomorrow): Episode 4, The Luddites: What They Really Feared. In 1811, in the English Midlands, men began smashing machines at night. Were they afraid of technology, or of something else?

Sources

• US Census Bureau, Historical Statistics of the United States, Colonial Times to 1970, Bicentennial Edition, Part 1 (1975), Chapter K (Agriculture): Series K 174 (farm employment), K 184–191 (farm machinery), K 407 (persons supplied per farmworker), K 496–498 (harvested cropland by use, including feed for horses and mules), K 570–573 (horses and mules on farms). census.gov (PDF)
• US Census Bureau, Historical Statistics of the United States, Chapter D, Series D 167 (labor force, Lebergott estimates). census.gov (PDF)
• The Henry Ford, “100,000th Fordson Tractor on Final Assembly Line, February 21, 1920”, artifact THF126008. thehenryford.org
• Wassily Leontief, “National Perspective: The Definition of Problems and Opportunities”, in The Long-Term Impact of Technology on Employment and Unemployment: A National Academy of Engineering Symposium, June 30, 1983, pp. 3–7. National Academies Press, doi 10.17226/19470. nationalacademies.org
• US Census Bureau, “The Great Migration, 1910 to 1970”. census.gov · Statistics in Schools, “Exploring the Great Migration” (teacher version). census.gov (PDF)
• Richard Hornbeck and Suresh Naidu, “When the Levee Breaks: Black Migration and Economic Development in the American South”, American Economic Review 104(3), 2014, pp. 963–990. aeaweb.org
• National Bureau of Statistics of China, China Statistical Yearbook 2025, Table 4-2 “Number of Employed Persons at Year-end by Three Strata of Industry”. stats.gov.cn
• Ministry of Agriculture and Rural Affairs of China, Department of Agricultural Mechanization, “2023 National Agricultural Mechanization Development Statistical Bulletin” (published 26 September 2024). moa.gov.cn
• Statistisches Bundesamt (Destatis), “Erwerbstätige im Inland nach Wirtschaftssektoren” (long series, data as of 30 September 2026). destatis.de
• Japan Ministry of Agriculture, Forestry and Fisheries, “Statistics on agricultural labor” (core persons mainly engaged in farming, Census of Agriculture and Forestry 2020 and 2025). maff.go.jp
• John Deere, “John Deere Reveals New Autonomous Machines & Technology at CES 2025”, 6 January 2025. deere.com · “John Deere’s Next Generation Perception System Brings Autonomy to Tillage”, 27 February 2025. deere.com

This article is for information only. It is not career, legal, financial or investment advice. Mentions of companies refer to their published statements, not to their shares.

Researched and drafted with AI assistance, including Anthropic’s Claude.