國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.2.12. McKinsey’s Views on Automation, Employment & Productivity - 2017 The McKinsey Global Institute (MGI) since its founding in 1990 has established an objective to create a deep understanding of the evolving global economy. As a business and economic research arm for the McKinsey & Company, MGI’s goal is to provide leaders in the commercial, public and social sectors with diverse facts and insights, which aid in strategic decision-making and the development of policies.
In the year 2017, MGI published the report “A Future That Works: Automation, Employment and Productivity” in an attempt to explain how advances in robotics, artificial intelligence and machine learning are introducing a new age of automation. The report analyzes the automation potential of the global economy as well as the factors that will determine its pace and the extent of the adoption and acceptance in workplaces. Their research is similar to that of Arntz, Gregory and Zierahn on OECD economies potential for automation (introduced in the previous subchapter), both of them base their approach on specific task displacement since they consider them more relevant due to occupations being made up of a range of activities with different potential for automation. MGI’s research stresses the fact that automation can enable businesses to improve their performance, by reducing the human error, standardizing quality and increasing speed. In many cases achieving outcomes that go far beyond human capabilities due to our own limitations. The scenario modeling estimates automation could raise productivity growth globally by 0.8% to 1.4% annually (McKinsey Global Institute, 2017).
Their study uses the state of technology with respect to 18 performance capabilities grouped in five major categories:
1. Sensory Perception: Sensory abilities
2. Cognitive Capabilities: Retrieving information, recognizing patterns, generating patterns, logical reasoning, optimizing & planning, creativity, articulating and coordinating
3. Natural Language Processing: Language generation and understanding
4. Social & Emotional Capabilities: Social and emotional sensing & reasoning, Social and emotional output
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
These performance capabilities span across 2,000 work activities/tasks and 800 occupations in order to calculate their potential to be automated using currently demonstrated technologies.
Their results show that less than 5% of all occupations can be fully automated (much less than the 47% of automation risk proposed by Frey & Osborne in 2013) and that about 60% of all occupations have at least 30% of tasks that could be automated. This constitutes more than half of the activities people are paid to do in the world’s workforce that will transform and evolve due to technology rather than be automated away.
Figure 7: Automation Potential Based on Demonstrated Technology in the United States (McKinsey Global Institute, 2017) The activities with the highest susceptibilities to automation involve physical activities in highly structured and predictable environments and those involving the collection and processing of data. These jobs are highly prone to become fully automated due to their predictable routine tasks and lower skill entry. Together they make up 51% of the activities in the economy accounting for almost US$ 2.7 Trillion in yearly wages. These jobs are more prevalent in industries like manufacturing, accommodation and food service, retail trade and some middle-skill jobs.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The study points out that the pace an extent of automation will vary from country to country and it will depend on technical, economic and social factors. Technical progress affecting the performance capabilities will give ground for more tasks to be automated. The cost of technology, supply and demand dynamics, performance analyses and regulatory policies will also affect the pace and scope of automation. Humans will need to adapt and work alongside machines in order to contribute to growth in per capita GDP. This implies an inevitable change in the nature of work, in which processes are transformed by the automation of certain tasks causing humans to perform complementary activities to the work done by machines (and vice versa). For policymakers, it is recommended to embrace the opportunity for economic development by creating policies to encourage investment and market incentives that will ensure continued progress and innovation. The magnitude of shifts in work activities is not unprecedented. In the United States of America, the share of farm employment fell from 40%
in 1900 to 2% in 2000, while the share of manufacturing employment fell from 25% in 1950 to less than 10% in 2010. In both cases, new previously inexistent jobs were created across various industries once again demonstrating the effects of the “creative destruction”.
Figure 8: Employee Overall % of Activities that can be Automated
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.2.13. PricewaterhouseCoopers & the Impact of Automation - 2018
PricewaterhouseCoopers (PwC) is a multinational professional services network; it is currently ranked as the largest professional services firm in the world. PwC’s network extends across 158 countries in the world and they provide services to 420 out of 500 Fortune 500 companies (PricewaterhouseCoopers, 2018). Due to its vast network, the firm is able to contribute data analysis to a wide range of areas and industries, including to one in particular which is important to this research: Automation and its effect on employment.
PwC released their report “Will robots really steal our jobs? An international analysis of the potential long term impact of automation” in February 2018, in this comprehensive report the firm attempts to demonstrate the actual impact in economies that technologies such as Artificial Intelligence, robotics and other forms of “smart automation” have on production levels, creation of better products and services and contributions to economic measures such as GDP. The research was primarily built around two previous studies on the subject: 1) Frey and Osborne’s research regarding the future of employment published in 2013, which utilized the Occupational Information Network (O*Net) Content Model. 2) Arntz, Gregory and Zierahn’s 2016 research focusing on the impact of automation on OECD member countries utilizing the Programme for the International Assessment of Adult Competencies (PIACC). Both of these studies were previously covered in this literature review.
The final report estimated the proportion of existing jobs that could be at high risk of automation by the year 2030 for 29 member countries of the OECD, different industry sectors, occupations within industries and workers of different gender, ages and educational levels. PwC also elaborated a timeline on how this process might unfold in the time between 2018 and 2030 through three overlapping waves:
Algorithm Wave: Focus on automation of simple computational tasks. This wave is already present at the time of this study
Augmentation Wave: Focus on automation of repeatable tasks and the communication and exchange of information through dynamic technological support. Also includes the analysis of unstructured data in semi-controlled environments. This wave is also underway at the time of this study but is expected to reach full maturity in the 2020’s.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Autonomy Wave: Focus on automation of physical labor, manual tasks and problem-solving in real-world situations that require responsive actions in industries such as manufacturing and transportation. The technologies enabling this wave are under current development but will become fully mature on a worldwide scale in the 2030’s.
Their research shows the proportion of jobs at high risk of automation by 2030 varies significantly by country. Some of the highest automation risk levels (> 40%) are observed in countries like Slovakia and Slovenia whose economies are more reliant on industrial production which tends to be easier to automate due to the routine tasks and programmable features. Lower automation risk levels (< 25%) are observed in countries like Finland and Korea which relatively higher education levels across their population. Service focused economies like the UK and USA lie in the middle-risk tier (Between 30% and 40%) due to their vast amount of lower-skilled workers that could see intermediate levels of automation in the long run (PricewaterhouseCoopers, 2018).
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Results show that countries with similar labor market performances and economic structures have broadly similar levels of potential automation. The figure above also details the levels of automation each of the countries may incur in throughout the firms proposed waves, it provides a short term and long-term view depending on the actual economic conditions in each country and the different technologies already in place. According to the study, the relative automatability of jobs across countries depends on a range of factors such as level of training, education, and skill enhancement. Therefore, countries with higher levels of education are estimated to have lower potential automation rates.
Different industries also see variations in automation risks across different waves. In the short term, financial services and other programmable sectors are more exposed due to algorithms outperforming human capability in a wide range of tasks involving pure data analysis. On the long run, sectors such as transportation are highly exposed as the worldwide scale introduction of driverless vehicles takes place. An industry’s task composition and educational requirements are the primary drivers behind its automatability. Industries where large number of workers are involved in routine tasks are likely to see more automation. Less automatable sectors have a larger portion of time spent on social and literacy-based tasks, these sectors generally possess higher educational requirements.
Figure 10: Potential Job Automation Rate over Time across Industries
(PricewaterhouseCoopers, 2018)
‧
potential boost of productivity and the creation of better products and services. The rate at which these technologies will affect businesses will vary across industries and countries. Nonetheless, companies should start investing now to reap the benefits of new trends. The direct impact of the implementation of new technologies on different levels of business will also vary depending on the diversity of tasks performed:Table 2: Key Impacts in the Three Waves of Automation
Phase Description Tasks Impacted Industries
Impacted
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Figure 11: Task Automation across the Three Waves
(PricewaterhouseCoopers, 2018) The introduction of new technologies will undoubtedly disrupt labor markets across the world, the extent to which automation will affect employment will vary due to a series of factors. Some of these factors relate to industries and categories of workers; however, countries will also need to consider the public policies implications and requirements. This means that the extent of automation in different economies may be more or less due to economic, legal, regulatory and organizational constraints.
PwC’s research suggests that job losses due to automation are more likely to be offset in the long run by new jobs made possible by these new technologies. The process of creative destruction observed through periods of major technological changes since the industrial revolution will undoubtedly repeat itself in the era of industry 4.0.
‧
國立 政 治 大 學