Yair Elbaz sent to Russ and me this very interesting interactive graph from McKinsey Global Institute. It arrays more than 750 paid occupations in the United States according to their average hourly wage rates* and the estimated percentage of how much of these workers’ tasks can be automated with existing technologies. Not surprisingly, the occupations that are most able to be automated – that involve the performance of the greatest number of tasks that can be automated with existing technologies – are generally at the lower end of the wage scale. The percentage of a surgeon’s tasks that can be automated is lower than is the percentage of a baker’s tasks that can be automated.
The reader can decide for himself or herself just what these data mean and imply. For example, what do these data tell us about the likely consequences of a minimum-wage hike to $10.10 per hour? To $15.00 per hour? Here are some points to keep in mind if and when such questions are asked:
- Whether or not substitution occurs as a result of a hike in the minimum wage depends upon the relative cost of implementing and supporting the automation compared to the cost of alternatives to the automation – alternatives that include (but are not limited to)
– reducing output by using fewer workers and without taking any steps to have the tasks that were once performed by the now-gone workers performed by some other means
– replacing a given number of lower-skilled (meaning, less productive) workers, not with automation, but with a smaller number of higher-skilled (meaning, more productive) workers
– working lower-skilled workers harder in order to get more hourly output from them.
- “Existing technology” is not exogenous to the economy – that is, existing technology is not just something that happens to be whatever it happens to be. Instead, existing technology is largely a response to existing economic rewards and costs. Therefore, if the minimum wage rises, so too does the payoff to investing in efforts to improve technologies that enable machines to substitute for low-skilled human labor. This fact means that for most, perhaps even all, of the occupations that you’ll click on in the interactive graph, the reported “% of time automatable” will rise if the minimum wage rises. These existing technologies will improve in response to the higher minimum wage. (Of course, this “% of time automatable” will not rise immediately: innovation and its adoption take time. But such innovation and innovation-adoption will indeed occur as time passes. The reality of this time lapse is one reason to discount the validity of many of the empirical studies of the minimum wage that find little or no negative employment consequences of raising the minimum wage: such studies, including the famous one by David Card and Alan Krueger, examine time horizons that are too short.)
* I assume that the wage rate listed for each occupation is that occupation’s average hourly wage, but it could be that occupation’s median hourly wage. (I was unable to discover which of the two aggregate summaries of hourly wages for each group is used.) Of course, for the vast majority of occupations – especially at the low-end of the pay scale – the difference between the mean and the median hourly wage is insignificant.
The above originally appeared at Cafe Hayek.