Elasticity
In the early 1800s, agriculture was the engine of the American economy, employing over 80% of workers and accounting for nearly 20% of GDP. In 1900, when automation began its march, it took 56 man-hours to grow an acre of wheat. By the late 20th century, that dropped to 3 hours. Today, agriculture employs less than 2% of the workforce and contributes only 1% to GDP.
Automation—tractors, threshers, combines, and the cotton gin—increased productivity massively. People, however, can only eat so much. Beyond a certain point, lower prices do not drive incremental demand. Economists call this an inelastic response. It means demand is not sensitive to changes in price. Since higher agricultural productivity was not met with a higher demand, employment fell.
Air travel is the opposite. Deregulation in 1978 and technologic advances drastically lowered the cost per mile flown. In 1970, a round-trip ticket from New York to London cost approximately $550, or 140 hours of labor for the average worker. Today, the same flight costs about 13 hours of labor. As the real price of airfare fell, demand exploded. Global Revenue Passenger Miles (RPMs) have grown from roughly 350 billion in 1970 to over 8 trillion in 2025.
Demand for air travel was highly sensitive to a change in price. People did not fly the same amount for less money. They flew more: low-cost fares created demand for weekend trips, international vacations, and in-person business meetings.
Moore’s Law describes one of the highest price elasticities in human history. Since the mid-1960s, the semiconductor industry has delivered an average 20% to 30% annual decline in the cost of manufacturing a transistor. In 1987 the global sales of semiconductors was $33 billion. In 2024 it was $775 billion.
As the price of a transistor fell, we didn’t just use the same amount of compute for less money. We put computers into everything from greeting cards to cars.
Historically, demand for software has looked more like airfare and semiconductors than wheat. It is highly elastic.
As programming languages became more efficient (moving from machine code to high-level languages like Python and JavaScript), the price of building an app dropped. Instead of hiring fewer programmers, companies built exponentially more complex software for every aspect of life, leading to a massive hiring boom in the sector. Between 2019 and 2025, the number of professional software developers worldwide grew from 14 million to nearly 21 million.
We didn't use the efficiency gains of high-level languages to hire fewer programmers; we used them to migrate the entire human experience—banking, social life, and infrastructure—into the cloud.
Software development jobs continue to trend higher and are meaningfully higher year over year. This runs counter to the narrative that AI is destroying software development jobs. As W. Edwards Deming, the father of the lean manufacturing movement, said: “In God we trust. All others bring data.”
