Technology is likely to substantially reshape labour markets in the future, dramatically altering the kinds of skills that middle-class workers will need. As such, policymakers must act now.
The debate over the impacts of new technologies on labour markets is centuries old. However, in recent years, a chorus of trailblazers, led by Erik Brynjolfsson and Andrew McAfee, have begun to argue that new advances in technology are likely to fundamentally reshape the kinds of skills that tomorrow’s middle class will need to cope with increasingly automated labour. This time, they say, things are different. If we believe that such advances might lead to what Keynes called ‘technological unemployment’, then European policy-makers must plan well before the effects are felt.
In the current debate, sceptics like Robert Gordon argue that we are entering a new era of low economic growth in which new technological developments will have less impact than past ones. Against him are the futurists, like Brynjolfsson & McAfee, who predict dramatic economic shifts from the coming of the ‘Second Machine Age’. They expect a spiralling race between technology and education in the battle for employment, which will dramatically alter the kinds of skills required by workers. According to this view, the automation of jobs threatens not only routine tasks with rule-based activities, but also, increasingly, jobs defined by pattern recognition and non-routine cognitive tasks.
Technology has always shifted labour markets by making some jobs obsolete, creating others, and changing the day-to-day nature of many. But what is fundamentally different about recent advances – in artificial intelligence or machine learning, for example – is that automation is moving beyond narrowly-defined mechanical processes and into areas that we previously thought needed the human touch. Consequently, a new range of professions are becoming susceptible to change. Whether these changes will lead to an overall decrease in employment is uncertain – a 2014 Pew survey of academic experts found an even split between those predicting net employment decreases and those expecting little overall change.
But, even if the overall impacts on employment are uncertain, the probability that worker skills will have to be reallocated is high. The question, then, becomes what the policy implications of increasing levels of automation will be and how policy-makers need to plan.
Jobs to be impacted
To answer this, we need to know, firstly, what kinds of jobs are likely to be impacted on and, secondly, the time horizon in play. Authors like Frey and Osborne argue that automation of a given job is restricted by the persistence of particular engineering bottlenecks – namely, social and creative intelligence, and perception or manipulation tasks. Using the incidence of these bottlenecks across different jobs, they posit that about 47% of all American jobs (currently representing over 50 million people) are threatened by computerization over the next two decades, with an inverse correlation between wage levels and the risk of automation.
Given the magnitude and new range of jobs potentially affected by these advances, the first policy implication relates to education. Changing how citizens are educated before entering the workforce is surely socially preferable to retraining and lengthy periods of unemployment as individuals search for new employment in different areas. Initiatives to better train young people in the use of technology are a natural area on which to place emphasis and to ensure that advances benefit, rather than hinder, tomorrow’s workers. For example, in mid-2012, Estonia launched an ambitious programme to teach children from the ages of 7 to 19 how to code. By teaching these skills so young, the government aims to change thinking around computers and programmes and encourage the smart use of technology. This programme is part of a larger government initiative to encourage computerization, an initiative which has already made Estonia the first European country with all schools connected to the Internet.
Cost and regulation
Beyond education, other policy responses depend on the time horizon of the predicted changes. The key point here is that automation is not a deterministic process, bound to affect labour markets according to some pre-set clock: the existence of labour-displacing technologies and their uptake are very different things. Two factors most affect uptake, with similar overarching policy implications. The first of these is cost: firms will only adopt automation technologies when there is an economic case to be made and the comparative cost of human labour is higher. This is why, for example, automotive manufacturing is so highly automated in countries such as the US and Japan, but much more reliant on human labour in low-wage countries such as India. New technologies often carry very high initial costs and, consequently, the role of government funding to these industries becomes pertinent.
The second factor is regulation. Automation has always run up against entrenched sectoral interests. Consequently, government sensitivity towards labour unions and the interests of those most affected by advances will determine the time horizons of labour market changes. Consider, for example, the firm Uber. Using advances in data analysis to efficiently broker a market between those with cars and those seeking lifts, the company has simultaneously undermined the collective action of its supply-side (by inducing competition between drivers) and benefited consumers (by reducing prices in the market). With strong regulatory opposition in a number of European markets, such situations demonstrate clearly how entrenched interests can clash and undermine the diffusion of new approaches to old areas of employment. Policy-makers must fundamentally decide how to balance the long-term spillover benefits that technological advances so often represent, versus the short-term constraints of those likely to be affected the most and earliest – whether in terms of inducing research and investment in areas likely to temporarily displace workers or in regulating sectors that are most vulnerable.
Long term decisions
Lastly, at the international level, not all countries will be similarly affected. In previous work applying the findings of Frey and Osborne, I found striking disparities in the expected degree of use of new technologies across the EU, and such disparities will surely be more pronounced globally. For example, the incidence of computerization varied from over 60% of jobs potentially being affected in Romania (where there is a comparatively high proportion of low-wage, low-skilled employment) to a low of 46% in Sweden. Implications for migration policy and even intra-EU redistribution should, therefore, not be overlooked, although it is doubtful that specific new policy mechanisms would need to be created to tackle these shifts.
Predictions are fickle and the contours of technological development are hardly mechanical. Nonetheless, technology is likely to substantially reshape labour markets in the long run, causing reallocations of the types of skills that middle class workers need. As such, policy-makers must consider a number of areas. Education and skills training is the most obvious area in the short term, but long-term decisions over research and investment policy, policy towards sectoral interests, and even migration and international redistribution all deserve attention. We should not succumb to the Luddite fallacy of naively fearing technological advances; but, still, policy must play a critical role in mitigating the risks and maximizing the benefits of these potentially dramatic shifts.