Trends in automatability over the past 20 years

State of Australia’s Skills 2021: now and into the future

Trends in automatability over the past 20 years

Automation has also affected the labour market over the past several decades and will continue to do so in the years ahead. The development and use of technology has changed many jobs and encouraged growth in higher skilled jobs. Over the past 20 years, there has been a net increase in jobs that are less easily automated and this trend is likely to continue.

To observe the impact of automation on the labour market, the NSC estimated the average automatability of occupations based on research by Duckworth et al 23. Using the Work Task Automatability model of the original study, a data set of predicted automatability of direct work activities was concorded to the specialist tasks of around 600 occupations in the ASC.

Based on this concordance, a weighted automatability score was derived for each of these occupations at the ANZSCO 4-digit or ANZSCO 6-digit level. The weighted automatability score is the weight of task automation by proportion of time spent on each task. The automatability score is rated from 1 to 4, with 1 being not at all automatable, and 4 being completely automatable. Further detail on the methodology can be found in Chapter 8 where the Work Task Automatability model is applied to outcomes from the scenario modelling exercises undertaken by the NSC.

Over the past 20 years, there has been a net increase in less automatable jobs. Figure 21 shows the gradual decline in the average weighted automatability score in the labour force with a weighted automatability score of 2.90 in February 2000 and 2.79 in February 2020, accounting for a difference in the score of 0.11 over the 20-year period. The average weighted automatability score was derived by combining the weighted score of all occupations by the employment size of each occupation group at any given point in time, at each quarter over the past 20 years 24.

The model to predict automatability assumes the specialist tasks for each occupation have been consistent over the past two decades. Therefore, the downward trend in the automatability of occupations is likely due to the net increase of occupations with lower weighted automatability scores.

Figure 21: Change in the average weighted automatability score in the labour force, 2000 to 2020

This time series chart shows the decline in the average automatability score of the entire labour market over the last 20 years. The average automatability score of the Australian Labour Market was 2.89 in February 2000 and this had declined to 2.79 in November 2020.

Occupations with strong growth in employment are relatively less likely to be automated. Figure 22 shows the average weighted automatability score for each occupation group (y-axis) and the change in employment growth over the last 20 years (x-axis). The size of each bubble represents the current employment size of each occupation group as of February 2020.

The 23 occupation groups are based on the NSC occupation matrix 25. The average weighted automatability score for each occupation group was based on the weighted average of each occupation (at the ANZSCO 4/6 digit level) underlying each occupation group.

The median weighted automatability score across all occupation groups is 2.71 and the median employment growth over the past 20 years is 47%.

The purple bubbles each represent 10 occupation groups that have a weighted automatability score above the median and experienced employment growth below the median over the past 20 years. Manufacturing and agriculture and animal and horticulture have both declined in employment size (-13% and -7%, respectively) and have weighted average automatability scores above the median (3.22 and 2.83, respectively). These occupation groups also employ more than 450,000 and 400,000 workers respectively as at February 2020.

The pink bubbles each represent nine occupations with a weighted average automatability score below the median which experienced above the median employment growth. Executive and general management and health and community services have grown by 153% and 132% respectively over the past 20 years and have weighted average automatability scores of 2.18 and 2.55 respectively. These occupation groups employ more than 200,000 and 1.5 million workers respectively as at February 2020.

The green bubbles each represent four occupation groups with either a weighted average automatability score above the median and employment growth above the median or vice versa. As at February 2020, hospitality, food services and tourism employed almost 800,000 workers and had experienced 39% growth over the previous 20 years. The average weighted automatability score (2.65) for this group is below the median score across all occupation groups.

Figure 22: Weighted automatability score by employment growth for each occupation group, 2000 to 2020

This bubble chart compares the weighted automation score of 23 occupation groups to their  employment growth and size between February 2000 to November 2020. The median automatability score is 2.71 and the median employment growth was 47 per cent. 10 occupation groups are relatively more automatable and experienced relatively lower growth in the last 20 years. This includes Sales, Retail, Wholesale and Real Estate which has a weighted automation score of 3.17 and grew by over 20 per cent. Nine occupation gr

Table 11 includes the weighted automatability scores and employment growth for all 23 occupation groups. The five occupation groups that are the least likely to be automated have experienced an average growth of about 100% in the last 20 years. By comparison, the five occupation groups that are the most likely to be automated have only experienced an average growth of approximately 22% over the same period.

Table 11: Weighted automatability score by employment growth of occupations, 2000 to 2020

Occupation Group

Automatability

Growth

Education and Training

1.81

62%

Executive and General Management

2.18

154%

Advertising, Media and Public Relations

2.32

135%

Legal and Insurance

2.33

100%

Sports and Recreation

2.52

51%

Health and Community Services

2.55

133%

Science

2.55

54%

Information and Communication Technology (ICT)

2.63

127%

Electrical and Electronics

2.65

47%

Hospitality, Food Services and Tourism

2.65

39%

Government, Defence and Protective Services

2.67

78%

Arts and Entertainment

2.71

75%

Mining and Energy

2.77

131%

Automotive

2.78

3%

Agriculture, Animal and Horticulture

2.83

-7%

Personal Services

2.88

34%

Construction, Architecture and Design

2.89

46%

Administration and Human Resources

3.02

44%

Transport and Logistics

3.08

44%

Engineers and Engineering Trades

3.09

29%

Accounting, Banking and Financial Services

3.11

29%

Sales, Retail, Wholesale and Real Estate

3.17

20%

Manufacturing

3.23

-13%

Source: NSC analysis using Duckworth et al.

Footnotes

23

P Duckworth, L Graham and MA Osborne, ‘Inferring work task automatability from AI expert evidence’, [conference paper], AIES ’19 (Artificial intelligence, ethics, and society), Honolulu, 2019.

24

This differs from the 2.85 score cited in the ‘Occupation analysis’ section earlier in this chapter, as analysis in that section is limited to the ANZSCO 4-digit level of detail.

25

The NSC Occupation Matrix differs from the more commonly used Australian and New Zealand Standard Classification of Occupations. Titles in the matrix have been grouped into broad categories based on field of work to assist users to better explore the labour market. More information on the occupation matrix can be found in Department of Jobs and Small Business, Australian Jobs 2019.