Automation in Kentucky: A Look at Automation Risk in Occupations and KCTCS Programs

Background:

Since the industrial revolution, technological advancement has increasingly shaped the workplace by changing the duties, performance, and productivity of workers. As workplaces change, the tasks associated with occupations also change and the skillsets of workers must adapt to keep pace. A recent study by Schwab and Zahidi (2020)1 estimated 40% of workers will need reskilling over the next five years which suggests new challenges and opportunities are emerging that will test the ability of educational institutions to meet the needs of industry. Technological change creates challenges for educational institutions by producing uncertainty about the future needs of industry, but it also creates opportunities for educational institutions to significantly impact local, national, and global economies. One of the most pressing concerns related to technological change and higher education is occupational automation. Key questions related to occupational automation include: What impact will automation have on job demand and wages? What jobs are likely to automate in the future? What skillsets will workers need as workplaces automate? How can postsecondary institutions anticipate and quickly act to meet the changing needs of industry?

 These questions are difficult to answer because they require knowledge about what the future might hold as it relates to technological change within a wide range of industries. It is likely the uptake of automation will be uneven across industries, so it is important to consider which occupations and programs are at risk of automating first. This report explores these questions by analyzing data on occupational automation risk and how it relates to job demand, wages, education level, and programmatic offerings. A key challenge that KCTCS must face as technology advances is aligning the skills produced in its programs with the needs of a modernizing workforce. Understanding how occupational automation will impact the need for programs will impact strategies for aligning educational production with workforce demand.

A popular example of automation impacting the workplace is the emergence of driverless vehicles and the potential of those vehicles to displace workers in the transportation and warehousing industry. A key question many education systems are considering is whether investment in transportation programs is prudent given the potential risk of automation on the horizon. A recent study by Gittleman and Monaco

¹Schwab, K. & Zahidi, S. (2020). The Future of Jobs Report 2020: World Economic Forum. Available at: WEF_Future_of_Jobs_2020.pdf (weforum.org)

(2020) ²found truck driving occupations are not currently in imminent danger of losing many jobs due to automation, and only specific segments of trucking like long hauling are likely to automate in the near term. However, other research suggests the pace of technological adaptation within the workforce is expected to grow and is likely to accelerate in the future (Schwab & Zahidi, 2020)³. Currently, trucking and CDL occupations are in high demand due to increases in delivery demand during the COVID-19 pandemic and the general trend toward online retail in place of brick-and[1]mortar stores. However, it is unknown when (if ever) we might see demand for transportation jobs decline and when we might see driverless vehicles begin to significantly influence demand for truck drivers.

While researchers are attempting to understand how and when automation will impact occupations like truck driving, there remains a great deal of uncertainty because there are many factors that impact the uptake of automation. For example, machines will likely need to be nimbler on construction sites than they will in factories, which means there are likely to be additional design costs to increase uptake in the construction industry. Theoretically, occupations that have specific needs for automation or occupations that have large incentives to automate (e.g., decreasing worker injury) are most likely to automate first while automation will take longer in occupations that are better performed by humans.

It is also unclear if occupational automation will necessarily lead to the loss of human jobs. Schwab and Zahidi (2020) report that automation during the COVID-19 era forced companies to adapt to worker shortages. Forty-three percent of companies surveyed indicated that they reduced their workforce during the pandemic and 34% indicated that they planned to increase technological integration. The report also suggests that while many jobs are likely to be destroyed due to automation, the number of jobs created is likely to be higher. The report estimates that by 2025, 85 million jobs may be displaced by automation while 97 million new jobs may be created. Thus, while companies are intending to make up for worker shortages through automation, they are also creating new demand for employees who have skills needed to work in modern automated workplaces. Postsecondary institutions must be able to anticipate these shifts to ensure companies have the workforce needed to fill these jobs.

²Gittleman, M., & Monaco, K. (2020). Truck-driving jobs: Are they headed for rapid elimination? ILR Review, 73(1), 3-24. 3 Schwab, K. & Zahidi, S. (2020). The Future of Jobs Report 2020: World Economic Forum. Available at: WEF_Future_of_Jobs_2020.pdf (weforum.org)

Data:

The Statewide 2019-2029 Occupational Outlook file and the 2021 Statewide Employment and Wage file from KYSTATS were used to measure occupation-level job demand and median annual wages. The Occupational Outlook file includes 709 unique occupational codes4 and several projections of labor market conditions. To explore the relationship between job demand and automation we utilized the projected annual growth rate, which is the average predicted annual change in employment between 2019 and 2029. The Employment and Wages file was used to measure median annual wages in 2021 of the occupations listed in the Occupational Outlook file. The Employment and Wages file includes 740 occupations with several variables measuring wage characteristics. The data were linked to the Occupational Outlook file by matching SOC codes in each file resulting in matches for 573 occupations.

To measure the predicted risk of occupation-level automation, data were utilized from Economic Modeling Specialists International (EMSI) that measures automation risk using a combination of metrics on the percentage of time spent on work that is considered high-risk for automation, the percentage of time spent on low-risk work, the number of high-risk jobs in compatible occupations, and the overall risk of automation within each industry. The automation risk measure is a scale with a mean of 100 and standard deviation of 15. EMSI calculates the automation index with data from O*NET and Frey and Osborne’s (2013)5 research on computerization within occupations. The automation risk data are collected at the SOC code level and are linked with the Occupational Outlook file to create the final dataset used for the analysis in this report. The final dataset includes 546 occupations that have complete data on demand, wages, and automation risk.

To explore the risk of automation in KCTCS programs a Classification of Instructional Programs (CIP) to Standard Occupational Classification (SOC) crosswalk file was used that links KCTCS programs to specific occupations. The CIP-to-SOC crosswalk is developed through a collaborative effort between the KCTCS Office of Research and Policy Analysis, Academic Programs and Policy, and with input from KCTCS Curriculum Committees. The file is based on a National CIP-to-SOC crosswalk file that is produced by the National Center for Education Statistics and is also informed by prior iterations of the KCTCS CIP-to-SOC crosswalk. The CIP-to-SOC crosswalk was used to link the occupations included in the main dataset (i.e., the dataset including occupations that have both demand and automation risk scores) to the current list of active programs at KCTCS. The linkage between KCTCS programs and occupations allows an automation risk score to be applied to each of the KCTCS programs with available data. An important caveat is that KCTCS graduates find jobs in a wide range of occupations that may be quite different than occupations the program is linked to using the crosswalk. This means the automation risk applied to KCTCS programs should only be viewed as an estimate and not as a definitive determination or prediction of automation risk in KCTCS programs.

423 of the 709 occupational codes in the Occupational Outlook file are industry aggregates and are excluded from the analysis.
5 Frey, C. B., & Osborne, M. A. (2017). The future of employment: How susceptible are jobs to computerisation? Technological Forecasting and Social Change, 114, 254-280.

Findings:

Figure 1 highlights the result of a linear regression model that examines the relationship between the automation index and the annual growth rate for occupations. The graph shows that a one unit increase in the automation index reduces the predicted annual growth rate by 0.20%, which is significant at the 99% confidence level. While this may not seem like a huge effect on the surface, the result suggests that occupations with higher automation risk are growing at a slower rate when compared to occupations that have a lower risk of automation.

Figure 2 highlights the result of a regression analysis that examines the relationship between the automation index and the median annual income across all occupations. The graph shows that the relationship between the automation index and the median annual income is non-linear, which required the use of a non-linear regression approach to produce the estimated effect. The graph shows a one unit increase in the automation index decreases the predicted median annual income by an average of $856.15, which suggests jobs with higher automation risk are associated with lower median incomes. It is noticeable in the graph that occupations on the low end of the automation scale are most likely to have incomes greater than $100,000. Conversely, nearly all the occupations with automation index scores greater than 120 have median annual incomes less than $50,000. Again, the principal explanation for this pattern in the data is the type of work required in the highest paying occupations. Occupations that are more human-centered like those in the education and healthcare fields are among the least likely occupational fields to experience automation. For example, in the education field, decision-making linked to determining the best educational 5 interventions for students with disabilities is best determined by trained educators and psychologists who observe, interact with, and communicate with children with disabilities. Machines will have great difficulty determining how best to educate children who are non-verbal or are not capable of providing direct inputs into any electronic devices and will have even more trouble helping parents through the stress and difficulty of determining the best solutions.

While the results of the regression analyses show the automation index is associated with lower job growth and wages, it is likely that the automation index is linked to other variables that also impact the outcome. For example, it was previously mentioned that the type of work commonly conducted within an occupation has a strong impact on automation risk. Another factor that impacts automation risk is the level of education required for the occupation; it is likely that jobs that require higher levels of education involve more complicated tasks that are not easily automated with machines. Figure 3 shows the relationship between the automation index and the typical education level required for entry into occupations. The graph shows that occupations with lower education requirements have higher risk of automation, but only occupations with typical entry education levels at “high school diploma or equivalent” or “no formal education” have above average automation risk (i.e., automation risk greater than 100). From a postsecondary perspective, Figure 3 clearly shows occupations that typically require a postsecondary degree are less likely to face automation risk in the future.

In addition to education level, the industry sector of occupations may also impact the risk of automation due to within-industry standards, common industry practices, and cultural acceptance of automation. Figure 4 shows construction; accommodation and food service; administrative and support and waste management and remediation services; mining; manufacturing; utilities; retail trade; agriculture, forestry, fishing, and hunting; transportation and warehousing; wholesale trade; and other services (except public administration) have higher than average automation risk. Industry sectors with lower levels of automation risk that are most relevant to KCTCS programs include healthcare and social assistance; professional, scientific, and technical services; finance and insurance; and educational services. Again, the industry sectors that include technical tasks that typically require a higher level of education and that have a human-centered focus are least likely to experience automation in the future6 . However, there is substantial variation of automation risk across several industry sectors, and that variation translates directly to KCTCS programs.

Figure 5 shows the top ten programs in each KCTCS targeted industry sector by risk of automation (See Appendix A for a full list of programs by automation risk). To link programs to occupations KCTCS regularly updates a CIP-to-SOC crosswalk that connects specific KCTCS programs to specific occupations. Overall, healthcare has the fewest number of programs with above average risk of automation (i.e., mean automation index above 100). Alternatively, manufacturing and construction have the highest number of programs with above average risk of automation.

In healthcare, the highest risk programs are Advanced Phlebotomy Technician, Phlebotomist, Phlebotomy Healthcare Worker, and Physician Office Laboratory worker, but these programs are only slightly above average on the automation index. This finding is consistent with other studies on automation that show occupations that are human-centered have a lower risk of automation when compared to occupations that focus on repetitive tasks. Many of the job-related tasks within the healthcare field require employees to be delicate, personable, and comforting to patients. These are tasks that are difficult for machines to achieve. However, there are likely tasks within the healthcare field that require precision, repetition, and the ability to recognize patterns in data. As healthcare costs continue to increase for most patients, the healthcare industry will continue to search for ways to reduce costs, improve tasks like appointment scheduling and patient privacy, and will incorporate automation in many areas. 

In transportation and logistics, the top ten programs align to the same occupation, which is Bus and Truck Mechanics and Diesel Engine Specialists. These are careers that will likely continue to automate as vehicle technology continues to computerize. As a result, workers within the transportation field will be required to learn more about technical processes that will be required to maintain, operate, and control new technologies. For example, as more electric vehicles enter the market new programs will be needed to train mechanics on electric motors, batteries, and other related technologies. While the need for heavy equipment mechanics is not going away any time soon, educational institutions should pay close attention to the needs of the transportation industry as it relates to mechanics and other technical services.

in business and IT, the top programs at risk of automation are aligned to occupations related to telecommunications. The programs Alternative Energies, Broadband Technology, Industrial Automation & Robotics Tech, Industrial Networking, Voice and Data Wiring, and Broadband Installation have the highest risk of automation. Again, these programs are highly technical and are at the intersection of human-machine interaction and may be in fields that experience a higher number of jobs resulting from automation. For example, Industrial Automation and Robotics technicians are likely in occupations that are becoming more computerized and technical, but as more companies begin to use robotics there will be demand for technicians who can repair and adjust robots.

In manufacturing, CNC Operators and various welding programs have the highest risk of automation. CNC operators align with the occupation Structural Metal Fabricators and Fitters and welding programs align with the occupation Welders, Cutters, Solderers, and Brazers. The manufacturing industry has been on the leading edge of the automation increase for decades. The reason for this are fairly clear. Manufacturing involves the need to mass produce items for sale into the marketplace. The repetitive 

nature of mass production initially led to the creation of the assembly line, which then was improved multiple times until companies perfected the processes. A recent study by Frenette and Frank (2020)7 found that not only was manufacturing among the industries with the highest risk of automation, it ranked first for the percent of employees at high risk of jobs transformation due to automation. Thus, as we think about the type of occupations that are likely to automate it is important to also consider how many employees work in those occupations to determine how many people could potentially be displaced. Within manufacturing, the evolution of the assembly line will continue as companies determine how to incorporate robotics, artificial intelligence, and other forms of automation that greatly improve productivity and make work environments safer.

Finally, in construction, the programs at highest risk of automation are Painters, Stone Masons, Bricklayers, Carpenters, Drywallers, and Residential Roofers. While the manufacturing industry has been leading in the increase in automation for decades, the construction field has the highest risk to automate moving forward. This likely means the construction industry will attempt to catch up to the level of automation within the manufacturing industry. Again, construction has many repetitive and dangerous tasks that companies will look to automate to improve efficiency and safety, but automation within the construction industry has additional challenges that have made uptake of various technologies slower and more difficult (e.g., difficult work conditions).

Conclusions:

The results of this research brief highlight several important findings related to the impact of automation on occupations and programs. First, the results showed automation risk has a significant negative relationship with the annual growth rate and median annual income of occupations. While the analysis performed for this research brief do not establish a causal relationship between automation and the annual growth rate and median annual income, they do show that there is a significant association. More research is needed to evaluate the relationship between automation risk and occupations to better identify causal mechanisms. Further, more research is needed to determine whether valid predictions of occupational automation can be made with the automation risk scores produced by EMSI.

The results of the report also suggest automation impacts occupations requiring lower levels of education. This result likely suggests occupations with higher levels of education attainment involve fewer tasks that machines can perform. Additionally, this finding suggests KCTCS should continue to promote the value of higher education because occupations that require higher level credentials are more likely to be sustained as the economy continues to automate. Additionally, KCTCS should continue to work with business partners to determine how best to respond to the emerging needs of companies as they introduce more automated processes into their businesses.

Another finding from the report shows automation risk varies substantially across industry sectors. Again, the reason for the variation in automation risk across industry sectors is related to the type of tasks that are commonly performed within various industries. The manufacturing industry has successfully integrated automated processes at a faster rate than other industry sectors primarily due to the nature of repetitive tasks that are commonly found in manufacturing processes. The report showed construction currently has the highest overall risk of automation, but several other industries are also facing significant automation risk.

A key caveat to industry automation is that the implementation of automation does not necessarily mean jobs will be lost. For example, commercial airplanes have relied on automated piloting systems for many years, but planes still require pilots to safely fly. However, the educational and training demands for pilots changed with the introduction of automated piloting systems, which required the aviation industry to change its training practices. For KCTCS to continue to maintain its place as the key provider of workforce education in Kentucky, it must track changes by conducting research to understand the how the skillsets of automated industries will change in the future. This will likely require research using available data sources that will help track when, where, and to which occupations automation is likely to impact, but also will require KCTCS to continue its outreach efforts to companies to promote a better understanding of specific skill needs within the context of automation.

Finally, the results of the report showed there is significant variation in automation risk across KCTCS programs. KCTCS programs are linked to occupations based on the type of skills that the program trains for and the typical skills needed for occupations. Not surprisingly, programs that align with the construction and manufacturing industries have the highest risk of automation. As KCTCS continues to prepare students for the workforce its leadership, faculty, and staff must continue to stay informed on the patterns of automation within various industry sectors so new programs can be created and existing programs modified to meet industry needs. This will require working with faculty, advisors, and subject matter experts including industry experts and accreditation experts who can help guide colleges along the path toward producing high-quality training programs. 

Recommendations:

Based on the results of the analysis we make the following recommendations related to automation, occupations, and KCTCS programs.

1. More research is needed to determine the validity of the automation index for tracking automation risk in occupations. Simultaneously, more research is needed to identify alternative methods for quantifying occupational and programmatic risk of automation.
2. As KCTCS and its Workforce Solutions departments continue to engage with business leaders and organizations to determine future education and training needs, the impact of automation and technological advancements should be considered. This is particularly important in industry sectors with more 12 pronounced automation risk.
   1. How can KCTCS incorporate high-quality industry feedback to help advise KCTCS in identifying needed programs.
   2. Is there work on the horizon to automate and streamline current processes?
   3. How might technology change in future years? When is this expected to occur?
   4. What skillsets will new technology require as occupations become more automated?
3. KCTCS should continue to conduct research into education and workforce alignment to promote a better understanding of patterns of graduate production and industry demand. Additionally, research must address the skill alignment between education/training programs and industry needs to accommodate automation and other technological advances.
4. KCTCS should continue to recognize the regional differences that may impact the uptake of automation technologies and how those regional differences may translate to both challenges and opportunities for KCTCS colleges.

Appendix: KCTCS Targeted Industry Sector (TIS) Programs by SOC and Automation Risk