In our recent report, “Are Robots Taking our Jobs or Making Them?,” we argued that one reason the alleged robotic invasion of the workforce (e.g., very high productivity through machines) was unlikely to happen was that the workforce was made up of such a wide variety of occupations. While there has been work done examining the susceptibility of work to automation through the decomposition of tasks or similar methods (see here, here and here), we thought that doing our own back-of-the-envelope calculation might be interesting. So we coded each occupation in the BLS Occupational Employment Statistics and ranked it according to our perceived ease of automation, with 1 being only moderately difficult to automate and 3 being very difficult. Then we added up each category to see how much of the economy was liable to be automated.
We admit that these classifications are not based on science, but rather educated guesses. Additionally, we should note that occupations in the “moderately difficult to automate” category can still be uneconomical or challenging automate for other reasons: we based our criteria on whether we could envision current or moderately better technology being able to automate them. Our main point is not to predict that job X will be gone in 2 years while job Y will be around for 20 years, rather just to show that there are a large variety of jobs that will in all likelihood not be automated any time soon. Technology is simply NOT on the verge of replacing all of our jobs. Preschool teachers are certainly safe for a while, for example.
The bottom line: we came up with roughly a 20-50-30 split between jobs that are moderately difficult to automate, difficult to automate, and very difficult to automate. Perhaps more interestingly, the weighted average wage of each group increased significantly according to the difficulty of automation.
Twenty percent of the economy for jobs that are moderately difficult to automate may sound large, but the U.S economy has a yearly job turnover of approximately fifteen percent. If there are already a large number of jobs being created and destroyed all the time, it’s possible that even large amounts of automation could be absorbed into old or new industries. However, assuming that even all the easy-to-automate jobs disappear within a year is quite a stretch of the imagination. Replacing workers in an entire industry or occupational category —and the business processes that surround them—with automated devices is seldom a quick and easy task, and it will not happen all at once. For example, even though automatic checkout technology has been available for years at grocery stores it has led to an immediate disappearance of cashiers.
Breaking down the groups into occupations, we can see the top 10 occupations by annual salary for each group. This data shows us several things. While there does appear to be wage bias to automation, there is nevertheless a wide variety of wages within each occupation. Although most of the occupations in the difficult-to-automate category are higher-skilled, the easy and moderate categories have a varied mix of low, moderate, and some high-skilled jobs.
Download our microdata and calculations here.