Author: Jeff Rosenthal
Different analysts have pointed toward the split of the labor market over the past 30 years into low- and high-wage workers. The sociology literature points to “good” and “bad” jobs (Kalleberg 2011) with social class implications, while the economics literature focuses on a polarization thesis where ‘middle jobs’ (requiring a moderate level of skills) are vanishing in favor of jobs that require more or fewer skills (Autor & Dorn 2013).
Recently, Third Way released work by economists Henry Siu and Nir Jaimovich from the University of British Columbia and Duke University, respectively. Their research showed evidence of job polarization and the loss of routine jobs — those with limited sets of tasks that are often instruction or rule based. Siu and Jaimovich argued that the meager job gains following the past two recessions are a result of the loss of routine jobs. Their point is that automation eliminates routine jobs, while nonroutine jobs can’t be automated, and thus can continue to grow.
They separated positions along the lines of earlier work by Autor, Levy & Murnane that explained a new division of labor where nonroutine cognitive tasks would dominate future jobs. They analyzed the data using a ‘2*2 Matrix’ of occupations based on degrees of routine and cognitive requirements for jobs. Examples of each of the four categories can be seen below:
As mentioned above, routine jobs are often rule based, and do not provide workers with discretion over their tasks. Routine jobs that rely on mental tasks are seen in the upper left quadrant of the matrix above, and include occupations such as Office Clerks or Bank Tellers. Routine jobs that rely on manual tasks (upper right quadrant) are production occupations like Team Assemblers or Machine Operators.
Nonroutine jobs are often not rule based, and thus are harder to automate. Examples of nonroutine jobs that rely on manual tasks (bottom right quadrant) are service occupations such as Housekeeper or Barista. Nonroutine jobs that rely on cognitive tasks include managerial, professional, and technical jobs like Physicians or Computer Programmers.
LEAD staff decided to see if this played out similarly in North Carolina over the same period of time. We analyzed Current Population Survey data from North Carolina from the past 20 years to see if jobs grew or shrank based on their routine or cognitive nature. To do this, we replicated Siu’s and Jaimovich’s routine/cognitive scheme for jobs (above) for detailed occupations across multiple coding systems (1990, 2000, 2010 Census Occupational Codes).
We found similar results in North Carolina as Siu and Jaimovich found nationally — routine jobs had shrunk over the past 20 years, while nonroutine jobs grew both in percentage and frequency.
The first figure shows the percentages of primary jobs in each category. Nonroutine cognitive positions are growing their percentage among the employed, while routine jobs are shrinking their percentage.
The main difference between North Carolina and the U.S. trends can be seen by separating manual from cognitive routine positions. While in the U.S., the decline in routine jobs comes from both manual and cognitive jobs; in North Carolina, the decline in routine jobs primarily comes from manual jobs. This finding makes sense given the decline of manufacturing in North Carolina over this time period.
When we look at the actual number employed in these types of positions, we see growth for each category except routine manual. This reinforces the composition found above and highlights the growth of nonroutine cognitive positions across North Carolina, as well as routine cognitive jobs.
These figures — by and large — show a continuing transformation of North Carolina’s occupational structure toward nonroutine jobs. This may be an indication that the jobs that are growing are ones that are not or cannot be automated away (at least yet).