A recent article in the New York Times described new computer software that in an instant sifts through thousands of legal documents looking a few litigable items; they replace hundreds of hours of lawyers reading the documents. This is not the start of a joke about how many lawyers you need to . . . But it does raise the question of how many lawyers you need. Economist-columnist Paul Krugman used the story to explain that computerization threatens to replace many white-collar jobs that are now held by college graduates. (And if you don’t need college graduates, do you need college professors? Uh-oh.)
It did not help settle anxieties that the story appeared shortly after IBM’s Watson computer beat two super-humans at Jeopardy. And now there are reports of software programs winning big pots on internet poker. The specter of automation unemploying us all may have finally arrived.
For decades, ages before personal computers, learned observers wrote about how machines were going to replace humans – for better or for ill. Some worried that the masses of dispossessed workers would form a revolutionary mob; others suggested introducing people to uplifting hobbies since we would have so much more leisure time on our hands. But the mass job shrinkage that these observers all expected did not come. Has it finally come?
How Ya Gonna Keep ‘Em
The biggest occupational displacement in American history was the virtual end of farming. Around the time of the Revolution, about 90% of Americans were involved in farming; they were farm owners, farm wives, farm kids, farm hands, farm slaves. Mostly they farmed to keep themselves alive and then farmed some more so they could barter or sell some surplus. Selling on the market became increasingly important in the 19th century as roads, canals, and rails linked the farmers to towns and harbors. Big ships then took farmers’ crops to Europe. Farming increasingly became a cash business.
But as the agricultural industry grew, it needed fewer and fewer workers to produce a bushel of corn, a gallon of milk, or a head of cattle. The absolute number of people who worked at farming and ranching hit its peak about 1910 – at around 11 or 12 million – and then the number dropped off rapidly. Today, agriculture provides fewer than 2 million jobs.
What happened? In great part, automation happened – better plows, planting and sowing machines, harvesters – as did scientific farming, better seeds, and the like. Millions of farmers and farm hands now made superfluous had to move on. The percentage of American workers who were farmers dropped from that early 90% or so in 1800 to about 40% of the labor force in 1900 and then to under 2% in 2000. Yet Americans as whole were not automated out of work; the farmers – or more typically, the farmers’ sons and daughters – found new kinds of jobs in a growing economy. A lot of those jobs were in manufacturing. Those jobs both paid better and usually provided better working conditions than did farming — one reason that every rural generation moaned about how hard it was to keep the kids down on the farm.
Machines Run Machines
The early water- and steam-powered factories that employed many formerly-rural Americans themselves displaced millions of craftsmen, a process some scholars have labeled “deskilling.” For example, early 19th-century shoemakers hand-crafted shoes, starting with the raw leather and ending with the laces, but by mid-century assembly-line shoes were undercutting their business. In one North Carolina town during the 1830s, church elder and shoemaker Henry Leinbach complained that “Rough times, these. It appears there is little love among us any more . . . .” One of Leinbach’s neighbors who may have shown too little love wrote that she preferred to order her shoes from Philadelphia, because they “wear and fit better than any I have ever owned” – and they were probably cheaper, too (see here). In place of craftsmen making shoes, machine-handling factory workers made them.
The number of factory jobs increased about 6-fold between 1860 and 1920 (while the population grew only about 3-fold). The percentage of American workers in manufacturing rose from about 15 to about 25 percent. Then, the factory jobs got harder to find, in part because of automation. It was one reason – along with a shift to foreign suppliers – that the number of manufacturing jobs in the U.S. peaked in 1979 at 19 million [see here]. That number dropped to 13 million in 2008, even before the Great Recession really hit.
Millions of manufacturing jobs have gone away, as millions of farm and craft jobs went away before. Yet, through all that, more new jobs appeared. Between 1960 and 2010, the population of the U.S. grew 1.7 times; but the number of employed Americans grew about 2.4 times. How can that be? Answer: The magic of growing productivity (combined with mothers [see here] and immigrants joining the labor force to fill those jobs).
The automation of farming, craft work, and manufacturing made products – most importantly, food – incredibly cheaper. For example, around 1900 a pound of bread cost an American about a half-hour of hard factory work; around 2000 a pound of fresher, more nutritious bread cost about five minutes of much easier work. The savings from cheaper food, shoes, and the like went into buying all sorts of new goods like cars and refrigerators and especially into paying service-providers: entertainers, doctors, waiters, teachers, software creators, bankers, police officers, yoga instructors, and the like. Many of the displaced farmers, craftsmen, and factory workers — or much more often their children — ended up in pink-collar, white-collar, and professional jobs.
This story – tragic at the level of the displaced worker, happy at the level of the national labor force – summarizes the work experience in America for centuries. Will it continue? Will the computerization of, say, document-searching eliminate jobs today but yield savings that will create newer, perhaps better jobs tomorrow?
Or has history turned a corner? Is AI (artificial intelligence) a new sort of automation, one that undercuts the brain work that became the mark of late-20th century employment, one that will only eliminate the better jobs? Will AI machines take over the best occupations such as systems analysts and biomedical engineers (paying humans about $75,000 a year) and leave people to be home health and personal care aides (at about $20,000 a year) – the four jobs that the Bureau of Labor Statistics currently expects will grow the fastest in the next decade?
The historical trend in American work suggests some optimism that better jobs for humans are coming, but history also suggests that few trends move in the same direction for very long.