Technological Change and Work: The Race Between Productivity and Demand
This is part one of a summary of my PhD thesis. You can find more information about the thesis here.
What is the future of work? Will robots take all (or most) of our jobs? In my PhD thesis, I analyzed historical examples of technological transformation to understand possible outcomes in the age of automation.
In this post, I will discuss what examples from history suggest for the future of employment: how many jobs will survive the rise of machines? I analyzed spinning (producing yarn for clothing and other textiles) and transportation in Britain and the US from 1750 to 1910, which were important sectors of the economy that experienced major technological changes, making them good examples to understand the possibilities of future innovation. They also have varying characteristics, which produced widely disparate outcomes.
Textiles were the second-largest part of the pre-industrial economy after agriculture. In Britain during the mid-1700s, spinning was a very large and growing source of work for women and children, with up to 15–20% of all women and children employed at spinning wheels for at least part of the year, usually working at home. Domestic and international demand for clothing (and therefore yarn) was rising, but the technology for producing it had changed relatively little for hundreds of years and each worker could not produce much.
In the late 1700s, three inventions allowed a rapid increase in productivity. One woman with a spinning wheel could produce about half a pound of yarn per day — and that yarn was only good for weaving into coarse, heavy clothing. By the 1790s, machines powered by water wheels or steam engines could produce ten times as much yarn that was finer and suitable for lighter fabrics. Prices fell, which allowed consumers to buy more, higher-quality clothing.
The losers are obvious: the women and children who had worked on spinning wheels saw demand for their output disappear in the last decades of the 1700s and the early 1800s. The new machines were so productive that they caused “technological unemployment”: job loss because of innovation.
In the United States, the outcome was different even though the technology was similar. In the 1700s North America was very sparsely populated, and there were fewer markets for cloth and intermediate products (like yarn, before it was woven into cloth). Most families of European colonists either spun yarn for their own use — homespun clothing — or purchased imported cloth from Britain. Women who spun for pay did so episodically, mixing earnings from spinning with many other paid and unpaid labor, such as housework.
When the same industrial spinning machines were introduced to the US, there was little technological unemployment because few women in the United States relied on spinning as their main occupation. Machines, based in new factories, provided a source of employment for young women, although there were tradeoffs to factory work that I will discuss in the next post. The relative underdevelopment of the US effectively protected it from technological unemployment, and tariffs on some textiles pushed American consumers to buy domestic clothing instead of British imports. Instead of replacing a large number of hand spinners, American factory spinning machines replaced a few irregular market-facing hand spinners and textiles imported from Britain.
The US transportation sector produced a third, distinct outcome. Since North America had a low population density and the few small cities were separated by long distances, there was little transportation infrastructure and few services at the end of the American Revolution. Population growth eventually increased demand for stagecoaches and boats to connect new towns, and consumers and shippers wanted better roads or canals for those services to use.
US investors and states began to build turnpikes (improved gravel roads with tolls) and canals in the 1790s and 1800s, but the most important technological break was the railroad. Railroads carried far more people and goods at higher speeds over long distances, although they were very dangerous in the 19th century.
Enthusiasm for this new invention stimulated a huge amount of public interest in and spending on railroads. Railroads were much more productive than earlier technologies (in terms of goods and people carried per employee), but because demand for transportation services rose the railroad workforce expanded rapidly through the 19th century. Railroads also had positive feedbacks: fast, reliable transportation encouraged economic growth, which produced more demand for rail services.
There is a familiar pattern across these examples: employment was in a race between productivity and demand, discussed by Richard Baldwin in his book The Globotics Upheaval.
In British spinning, productivity rapidly outstripped any increase in demand produced by falling textile prices, and employment collapsed. Even at the height of the British factory textile industry at the end of the 19th century, when Britain produced 70–80% of global cotton cloth exports, there were fewer workers employed than there had been before industrialization — and a much smaller share of the population.
In the United States, a small preindustrial textile industry with few dedicated workers meant that machines hardly replaced any jobs. Protected behind tariff barriers established after the War of 1812, American textile employment grew when factories were established, and its output took the place of British imports and some hand production.
The transportation sector in the US is the most positive example: new technologies raised productivity, but demand for transportation services also rose. Rapid, efficient transportation also led to a virtuous cycle for employment up to the early 20th century.
What are the implications for the future? The British textile example suggests that new technology in areas of “low-hanging fruit” — major innovations in sectors with low productivity and large-scale employment — can cause substantial unemployment and poverty.
However, the same innovations can have dramatically different employment effects in different countries. If it is successful, the long-running search for driverless vehicles may have a much larger impact on employment in the US, which has a very large share (about 65%) of freight transported by truck, than in other countries. Places where truck freight is negligible because of high labor costs, by contrast, could benefit from cheaper access to technology that moves goods but doesn’t replace many workers.
More optimistically, the US transportation example shows that some pathbreaking innovations can contribute to higher employment even though they allow a huge increase in goods produced or services provided per worker. The relationship between innovation and employment will depend on whether an invention has widespread benefits and on the economic context in which it is adopted.
In the next post, I’ll discuss my findings for a different dimension of the impact of technology on work: how does innovation shape access to good jobs?