Featured Indicator: Risk of Workforce Automation by State

The average job has a 61% chance of being automated at some point in the future.

That’s what you find when you stack employment data for the United States against automation probabilities. Just how much automation could endanger the typical American job? And, how does that probability vary from state to state? 

The data was provided by a 2013 Oxford University research paper by Carl Benedikt Frey and Michael A. Osborne, entitled “The future of Employment: How Susceptible are Jobs to Computerisation?” Note that for the paper, “computerisation” refers to “job automation by means of computer-controlled equipment” (p. 2), and will be used interchangeably with the word “automation” in this article. For more info on the methodologies and findings of the paper, check out our other Navigator article, “How Automation Will Help Define Workforce Development.”

The Approach

To help guide data analysis, all 703 studied occupations were divided into four quartiles, then further divided these quartiles into three risk pools:

  • Low-Risk Jobs: Jobs with a 25% chance or less of being automated in the future. These jobs require high levels of the skills like perception and manipulation, creativity, and social intelligence, and as such can be considered safe from automation for now. The three occupations with the lowest likelihood of automation are Recreational Therapists (SOC 29-1125; 0.28% chance of automation); Emergency Management Directors (SOC 11-9161; 0.3% chance); and First-Line Supervisors of Mechanics, Installers, and Repairers (SOC 49-1011; 0.3% chance).
  • Medium-Risk Jobs: Jobs with a probability of automation between 26% and 75%. For these occupations, the jury’s still out on if or when automation will occur. Despite the size of this increment, the share of medium-risk occupations amounts to roughly a quarter of all jobs. Three jobs can be considered a literal coin flip (exactly 50% probability) based on the data: Court Reporters (SOC 23-2091); Installation, Maintenance, and Repair Workers (Unclassified) (SOC 49-9799); and Loading Machine Operators (Underground Mining) (SOC 7033).
  • High-Risk Jobs: with all jobs with a probability of automation above 75%, this risk pool makes up roughly 50% of all occupations. For many of these jobs, the question of automation is less “if” and more “when”. There are 93 jobs with a probability of 95% or higher, and 12 jobs with a 99% probability. These jobs include Telemarketers (SOC 41-9041), Data Entry Keyers (SOC 43-9012), and Tax Preparers (SOC 13-2082).

We then combined these values with publicly-available 2016 employment data from the Bureau of Labor Statistics for every SOC occupation included in the paper. The Average Probability of Automation for each state was found by multiplying the number of workers for each occupation by that occupation’s probability of automation, taking the sum of those values, and dividing that sum by the total number of workers.

The graphic below is a choropleth (sort of like a heatmap) of the United States, shaded in by the average likelihood of job automation for that state. Hover over a state to see more detailed information and a breakdown of how prevalent jobs of each risk pool are in each state’s economy. 

US State Average Probabilities of Workforce Automation, 2016

 
 
Hover for Info
Low-Risk Jobs Medium-Risk Jobs High-Risk Jobs

 

 

The Findings

For the most part, the level of variation in probability from state to state is not going to be particularly high, because larger geographies usually have industry concentrations that are similar to the national average. For comparison, here are the five states with the lowest average probability of automation:

  • Pennsylvania: 55%
  • Texas: 56%
  • Florida: 57%
  • Oregon: 58%
  • Virginia: 58%

As well as the five states with the highest average probability of automation:

  • Ohio: 67%
  • California: 66%
  • Nevada: 64%
  • Missouri: 64%
  • Wyoming: 64%

After reviewing the data, we could not help but wonder: “what is happening with Ohio and Pennsylvania?” The two states rank first and last in terms of average likelihood of automation, respectively. The average job has a 22% higher probability of being automated in Ohio than in Pennsylvania, despite the two states sharing a border. The use of location quotient analysis, which compares the density of an industry of a given region to the whole United States, can provide some idea of what may be going on to cause such a disparate result in two bordering states. In Ohio, for example, Manufacturing (NAICS 31) has a location quotient of 1.51, meaning it is 51% more concentrated in Ohio than in the entire US. Pennsylvania, by comparison, has a manufacturing location quotient of only 1.14, indicating a higher industry concentration, but not as high as that of Ohio. It is possible that the larger concentration of a highly-automatable industry could contribute to this stark contrast.

As discussed in How Automation Will Help Define Workforce Development, the accelerating pace of automation means that communities will need to narrow the scope of their workforce development plans in the future, and place a greater focus on ensuring higher quality education that is either specialized enough to outpace automation or abstract enough to allow for better adaptation to automation trends. The exact form that these plans will take will vary drastically from community to community, depending what mix of automatability is evident in that community’s industry mix. 

Notes: 

  • In the spirit of automation, we are also debuting some new interactive graphics, which may not work on some older browsers or smaller displays. For browsers that do not support these graphics, a static image will be displayed.
  • The map was built using a Javascript library called D3.js, and expands on code and mapping provided by the library’s developer. 

Contact Information

120 West Avenue, Suite 303
Saratoga Springs, NY 12866
518.899.2608
info@camoinassociates.com

Contact Us