The age of computing power — computerization, automation and robotics — is transforming today’s global labour markets in unprecedented ways, presenting great uncertainty about the future of the employment landscape. In this age, we’re told driverless cars, pilotless planes and chefless kitchens are all within the realm of possibility.
Of course, this is not the first time the economy has faced uncertainty at the hands of technological change. Perhaps most notably, the 19th and 20th Centuries were shaped by three big technological revolutions — the invention of steam power, followed by the arrival of electrical power and then finally, the development of the internal combustion engine — that replaced existing jobs and greatly economized on labour. These major developments also led to a widespread increase in manufacturing and service-sector jobs and improved standards of living.
Is the current age different? A telltale sign is the well-documented job polarization phenomenon, the thinning out of middle skill jobs in the U.S. labour market and an increase in the share of low-skill, low-wage and high-skill, high-wage jobs. With the current global share of manufacturing at close to 15 per cent and the share of services nearly 70 per cent, the latter sector is where the impact of technology and jobs is likely to be felt most in the future. Understanding how this technology-job nexus will evolve, as well as the role of public policy in the transition is now a central issue for industrial and emerging market economies.
A conceptual framework: tasks vs skills
Recent economic research provides insights to help understand and organize thinking on the present and future impact of technology on jobs. It draws an important distinction between tasks and skills. A task is work that produces output. Skills are capabilities a worker has for performing a variety of tasks, and generally accumulated or enhanced through college-level, or higher, education. With this distinction, we can meaningfully speak of different types of tasks that can be performed by a range of skills (low, mid, high) or by computer software and automation.
Tasks can be “routine” if they involve following well-understood or repetitive procedures, or “non-routine” if they require active contextual awareness, adaptability and precision. Additionally, many tasks require manual labour, while others depend on cognitive abilities such as problem-solving, intuition, persuasion and creativity. This conceptual framework delineates four categories of jobs, namely, Routine-Manual (RM), Routine-Cognitive (RC), Non-Routine Manual (NRM) and Non-Routine Cognitive (NRC).
While there may be an overlap between these four categories, each one represents the most important aspects of any particular task in that category. Examples of RM jobs are assembly-line work, housekeeping, warehousing, packaging services, supervision and inspection, picking and sorting. RC jobs include clerical work, bookkeeping, customer service, banking services and administrative services. The NRM jobs category refers to such occupations as child care, construction, health care, personal services, food services, cleaning services and truck driving. And finally, NRC jobs are primarily science, technology, engineering, math (STEM) jobs, software development, legal services, managerial work, nursing and teaching.
The most at-risk jobs
Routine jobs face the highest risk of being substituted by advancing technology, in particular, through computerization and automation. These jobs are also prone to off-shoring to countries with lower costs of production. Indeed, the income share of routine jobs in the U.S. has fallen from about 40 per cent in 1970 to under 25 per cent now, as shown in Chart A. At the same time, the income share of non-routine jobs has risen about 25 per cent to nearly 35 per cent over the past four decades.
Non-routine jobs, especially in the NRC category, are seen as more immune to technological substitution, as they are viewed as providing a complement to information and communication technologies. However, the stupendous doubling of the transistor density of a microchip every two years, as famously predicted by Intel’s co-founder, Gordon Moore, in 1965, reflects the ever-increasing computing power. When combined with Big Data, machine learning and machine-environment control, it is conceivable that automation costs will continue to fall rapidly and machines could replace even highly skilled humans performing NRC jobs in the future.
An intriguing recent finding that corroborates this view is that there has been a remarkable slowdown in the demand for cognitive tasks since the year 2000. As a result, more high-skilled workers have moved down the occupational ladder, pushing out low-skilled workers, sometimes out of the labour force entirely. Chart B shows, despite the cyclical fluctuation, the general decline in the median wage growth of high, low and mid-skill jobs.
Policies for the (near) future
While the trends themselves are not in dispute, there are optimistic and pessimistic views and predictions about how economies will adjust to these trajectories. On the one hand, some believe that the service sector will transform in a way to generate a sustained demand for jobs that complement evolving technology and that the widespread use of robots in NRM and NRC jobs is still several decades away.
From this perspective, the design of public policy must contend with at least two challenges. First, a central concern is how to improve the effectiveness of relevant advanced education and retraining programs, and relatedly, to study and assess how different cohorts of students and workers will adapt to these trends. A second challenge is to sustain and support innovative activities that will lead to complementary jobs. Evidence favours geographic clustering of innovative activities and ideas-generation, as seen, for example, in Silicon Valley, as a way to foster such necessary innovation. Startup hubs, such as the one in Kitchener-Waterloo, are examples of such clustering in Canada.
Others believe that job losses in all four categories listed above will mean sharp increases in income inequality, which is already on the rise, and that these large-scale job losses will occur in the near future (within 10 years). If this occurs, there will, at the same time, be a reduction in tax revenues. Under this scenario, the main challenge for public policy will be to design policies that simultaneously address these two developments. For example, a serious consideration of redistributive basic income proposals will continue to gain momentum. These types of proposals and policies are not new territory for Canada; the Canadian guaranteed annual income field experiment (MINCOME), conducted in Manitoba in the 1970s, provides a historical example from which we can learn. Another pertinent proposal in this context is taxing robots to sustain revenues as recently suggested by Microsoft co-founder Bill Gates.
The realistic scenario is likely a mix of these two views and none of these policies is mutually exclusive. Another dimension of the technology-job future is that no country’s labour market will remain insulated from the change. Wherever and whenever cheaper automation takes hold in manufacturing and services, for example, currently off-shored low-skill RM, RC jobs in Asian, African and Latin American countries may start to shrink. These developments are likely to put significant pressure on domestic economies. For industrial economies, adopting inward-oriented policies focusing on increasing jobs in the manufacturing sector, as proposed by the Trump administration, may only be a temporary solution to job creation by forcing workers to accept lower-market wages. The incentives for substituting cheaper automation will restrict wage growth. Navigating the age of computing power will present enormous opportunities for global policy co-ordination, especially in the services sector.
Hashmat Khan is a full professor and co-director of Carleton University’s Centre for Monetary and Financial Economics.
by