"If, thanks to greater automation, the rail sector can thrive in the future, this is likely to translate into more job opportunities than job losses. If, on the contrary, rail is overtaken by the pace of changes in other transport modes, there is no question about what that will mean for rail jobs." - Luisa Moisio
Technological advances and their impact on jobs have inflamed the debate and provoked anxiety since handloom weavers fought against the mechanisation of the British textile industry in the early 19th century.
Over the past two centuries there has been pain and turmoil as certain jobs vanished, others changed beyond recognition and totally new ones emerged, but technological progress and associated automation have not made human labour unnecessary and obsolete, or lead to the 15-hour work week predicted in 1930s by John Maynard Keynes for his grandchildren.
Quite the contrary, the employment-to-population ratio rose, particularly in the 20th century as women moved from domestic jobs to market occupations. In the UK we currently work an average of 37.5 hours per week and whilst unemployment rates fluctuate cyclically, they do not show signs of increasing over the long term.
Welcome to today’s debate
The current wave of new technologies is wide raging, including digitisation, data capture and storage, robotics and artificial intelligence (AI). Known as ‘the second machine age’
(Brynjolfsson and McAfee 2014) or ‘the fourth industrial revolution’ (Schwab 2015), this has raised new, significant concerns over the future of jobs.
There have been many attempts to put numbers against these concerns. The 2013 study by Frey and Osborne found that 47% of all jobs in the United States were highly susceptible to automation over the next decade or two. Using a similar approach, 30% to 35% of UK jobs were found to be at risk of automation as reported in the 2016 report by Citi and the Oxford Martin School at the University of Oxford and the 2017 report by consultancy firm PwC.
However, the 2016 OECD study by Arntz, Gregory and Zierahn distinguished between jobs and tasks automation, and concluded that a much smaller share of jobs – 10% in the UK– was likely to be under threat from automation. This lower estimate assumes that if less than 70% of the tasks performed within a job are ‘automatable’, job transformation is more likely than job obsolesce. A similar approach was taken by the 2017 McKinsey Global Institute ‘A future that works: automation, employment and productivity' and concluded that about 26% of jobs have 70% or more of their tasks that could be automated.
Studies looking at past examples of automation have often pointed out how it changes the nature of jobs and leads to increased demand, thanks to an increase in productivity. For example, 2016 ‘How Computer Automation Affects Occupations: Technology, Jobs, and Skills’ by Bessen shows an average annual increase of legal clerks by 1.1% between 2000 and 2013 following the introduction of software capable of analysing large volumes of legal documents which reduced the cost. Similarly, whilst the introduction of automated teller machines in the US, led to bank clerks falling from 20 per branch to 13 between 1988 and 2004, the reduced cost of running a bank branch meant that banks could afford to meet customers’ demand for more branches. This led to the number of branches increasing by 43% over the same period, with a net increase of the total number of employees.
Not just the quantity of future jobs, but also their quality is in question: the fear is that this wave of new technologies eliminates many middle-skilled middle-income jobs and complement only a few high-skilled high-income ones. If new jobs are about ‘smart people’ working with technology, resulting in middle-skilled workers either moving out of the labour force or moving down the occupational ladder, taking on jobs traditionally performed by low-skilled workers, this could increase inequality which has already soared over the last several decades.
So, can we draw any conclusions?
There are some tasks which are more likely to be automated: tasks that involve physical activities in highly structured and predictable environments and those that involve the collection and processing of data. More complex tasks which require perception and manipulation, creativity, empathy and social intelligence are unlikely to be fully automatable over the next couple of decades.
Estimates of the potential job losses are generally based on anticipated technological capabilities by the early 2030s (which is already very tricky to predict!) but actual automation uptake may vary significantly for a combination of economic, organisational and regulatory constraints. Economic attractiveness (why take the risk of investment in automation if there is a lower-risk and lower-cost human alternative? But higher cost of labour would make investment in automation much more attractive), organisational inertia (for example the life cycle of existing systems and adaptability of legacy systems), and legal hurdles (for example liability for accidents with driverless vehicles) will all impact on the actual speed at which automation will be introduced.
While the combination of digitisation, big data, and AI will not cause mass unemployment any time soon, over the next two decades some jobs will cease to exist and many more will change substantially. Job losses are likely to be offset by job gains elsewhere with a highly uncertain net effect on total human employment. But what is certain is that both new and changed jobs will require workers to acquire new skills quicker and more routinely than in the past.
What does this mean for transport in general?
Automation is increasingly being used in transportation, and the technological improvements underway, in particular in visual and environmental processing, will change the transport industries radically. Some of the world's biggest companies – both automotive and technology – are investing billions in a race to be the first to create self-driving vehicles with on road demonstrations underway.
This leads many to conclude that many transport related jobs will be lost. For example, the 2016 report by Citi and the Oxford Martin School at the University of Oxford estimates the probability of automation for ‘Transport and mobile machine drivers and operatives’ to be 52.7%. And Vivek Wadhwa, author of ‘The Driver in the Driverless Car', estimates that millions of driving jobs will be lost in the early 2020s, as vehicles achieve full autonomy and drones that drop packages on door steps and in backyards are further developed.
Others point out how driving jobs often involve other tasks, such as unloading merchandise, speaking and interacting with customers, and coping with emergencies. Automated solutions currently either cannot perform or do not perform well at such tasks, hence while some tasks will be lost, jobs will change and adapt. Also, the autonomous vehicle industry will likely create new tasks, such as IT and comms specialisms to update and maintain novel customised information systems like passenger chatbots. A 2015 PwC study estimated that around 6% of all UK jobs in 2013 were of a kind that did not exist at all in 1990.
More rarely considered is the fact that the magnitude of potential job/task losses by sector is driven not only by the proportion that have a high likelihood of automation (which tends to be the focus of the debate), but also by the present and future level of share of that sector. The boost of sector productivity thanks to automation can generate extra demand for its core traditional services, and also demand new related services, including those that are less automatable. Looking back, the introduction of automobiles led to a disappearance of horse-related jobs. However, not only did the automobile industry itself grow quickly, it also led to more travelling, and this in turn created jobs to provide (new and improved) services to significantly greater number of motorists and truck drivers.
What are the implications for rail?
Ultimately rail, as any other sector and transport mode, needs to be focussed on making it a better experience for its customers. Some elements of this might come from automation (eg. Automatic Train Operation can enable more capacity and increase punctuality) but many other important elements will not necessarily be automatable or will be better performed by a combination of machine and humans (eg. management of disruptions).
Automation has already transformed the rail industry, and it will continue to do so with driverless train trials already underway in different places and the use of robots and drones to perform inspection and maintenance tasks on the rise. This will eliminate certain jobs, but it will also redefine many others and create new ones.
If, also thanks to the adoption of automation, the rail sector can build on its past success and thrive in the future, this is likely to translate in more job opportunities than job losses. If, on the contrary, rail stagnates and is overtaken by the pace and nature of changes in other transport modes, there is no question about what that will mean for rail jobs.
The former scenario comes with its own challenges. But if rail fails to take advantage of this wave of automation and use it to deliver benefits to both our customers and workforce, we risk losing customers and fuelling staff resentment. For automation to bring a successful future for the rail, we need to pay as much attention to the ‘social’ challenges as we do to the technical ones, and avoid using the former as an excuse for inaction.
In the next article David Hardman – Senior Research Analyst – will consider what employers, employees, and government can do to respond to the workplace transformations brought about by growing automation.