Driving a taxi used to require a specialized skill — deep knowledge of local streets. But the arrival of GPS tools made navigation easier to do and much less valuable. There are more people providing transportation services now, because of GPS and ride-share companies, and as a result, wages have fallen for the profession.
In short, technology reduced the expertise required to do drivers’ jobs. According to MIT economist David Autor and principal research scientist Neil Thompson, this is a good example of how and when automation can affect workers: It depends on how much expertise is required to do the remaining, nonautomated tasks.
“Which workers gain and which lose as automation advances depends on whether automation makes their expertise scarcer … or redundant,” the researchers write. Another way they put it: “Does automation replace experts or complement expertise? The answer is yes.”
In a new paper, Autor and Thompson explore how automation affects the value of expertise, or a worker’s capability to complete specialized tasks. To illustrate this — and to show how similar jobs can be affected differently by automation — Autor and Thompson compared two roles: accounting clerks, who complete tasks such as recording transactions, reconciling statements, and identifying discrepancies; and inventory clerks, who evaluate stock, conduct inventories, and compute prices.
Both jobs include tasks that are targets for automation, but they have not been affected by it the same way. Wages for accounting clerks have gone up, though overall employment in the field has gone down. The reverse is true for inventory clerks: Wages have fallen, but the number of jobs has gone up.
Here's why: The expertise required for the nonautomated tasks varies between the two roles. Despite automation, accounting clerks still need to complete complex problem-solving tasks and make difficult decisions that require expertise. But the nonautomatable tasks of inventory clerks, such as counting and stocking inventory, require less training and certification — most people can do them.
This research shows that managers’ responses to automation should be based on what kind of work is being automated, Thompson said.
If the nonexpert parts of a job are automated, wages will go up but the number of people needed in the role will shrink. In this case, companies can often adapt gradually, such as by not replacing workers when they leave.
But if the job becomes less specialized and wages go down, the adjustment is harder — lower pay can hurt worker morale and make recruitment and retention more difficult.
“Depending on which tasks are automated, you’ll see different effects in different parts of your workforce,” Thompson said. “You want to think about that ahead of time because it calls for different responses from the company.” These responses could include designing more specialized, high-value roles that retain expert tasks or making certain jobs easier to access to expand the talent pool.
