ESA National Economic Panel Polls

Robots, artificial intelligence and the 'future of work' - October 2017

The public policy debate about the economic and social impact of disruptive technologies, transitioning industries and the 'future work' is finally gaining momentum in Australia.

This month, we wanted to ask the ESA's National Economic Panel (NEP) for their views on two related propositions - the same propositions put to the eminent US and European economists of the IGM Economic Experts Panel in September 2017. The only change we made was to focus our panellists on the impacts in Australia rather than in (all) 'advanced countries'.

Question A: "Holding labor market institutions and job training fixed, rising use of robots and artificial intelligence is likely to increase substantially the number of workers in Australia who are unemployed for long periods."

Question B: "Rising use of robots and artificial intelligence in Australia is likely to create benefits large enough that they could be used to compensate those workers who are substantially negatively affected for their lost wages."

** Question credits: this month's poll questions were adapted - from those put to the eminent US and European economists of the IGM Economic Experts Panel in September 2017 - to focus on the impacts in Australia rather than in all 'advanced countries'. For the results of the US IGM poll, see: http://www.igmchicago.org/surveys/robots-and-artificial-intelligence-2 For the results of the European IGM poll see: http://www.igmchicago.org/surveys/robots-and-artificial-intelligence.

Overviews of poll results by Professor Jeff Borland and Professor Jason Potts

Professor Jeff Borland

By Jeff Borland, Professor of Economics, The University of Melbourne

The impact of technological change on the Australian labour market and income distribution should be relatively limited, according to respondents to this poll.

Professor Jason Potts

By Jason Potts, Professor of Economics, RMIT University

The October 2017 poll conducted by the ESA replicated a US survey of economists on topical but broad policy issues, this one specifically in relation to labour markets and new technology. The two questions were about labour-capita substitution effects, and the distributional welfare consequences of continued innovation in robots and artificial intelligence.

Question A

Question B

Responses (33)


Rachel Armstrong	A - Agree B - Strongly disagree


Garry Barrett	A - Disagree B - Agree		Productivity gains are the foundation of improved standards of living across the community. Technology change can cause dislocation, and different groups of workers - for example, by skill level and type, experience and location - will be affected differently. An important distributional issue is to ensure groups more strongly impacted are appropriately supported.


Harry Bloch	A - Strongly disagree B - Agree


Alison Booth	A - Disagree B - Agree


David Butler	A - Agree B Agree		I recommend the following short ‘duel’ where the main points are debated nicely: https://www.mruniversity.com/courses/econ-duel/will-robots-take-our-jobs


Matthew Butlin	A - Uncertain B - Agree		On the first question, robots have been a significant element of manufacturing for many years. It is very likely robots and AI will displace many jobs. Whether or not this translates into long term unemployment depends significantly on the existence of other concurrently growing areas of new jobs and also the capacity of the education and training systems (broadly defined) efficiently retrain and reskill displaced workers. In short, having flexibility in skills formation is very important and is an essential part of structural adjustment. On the second question, historically deep structural changes driven by technological change (e.g. the industrial revolution in Britain) have had a net increase to income but the distributional consequences have been particularly adverse for some groups in the adjustment.


Lisa Cameron	A - Agree B - Uncertain		Many routine aspects of jobs will be automated which will reduce labour demand, although I suspect not as dramatically as many people are predicting as human judgement will remain important in many roles. The benefits to those who own capital will probably be substantial and maybe enough to compensate the unemployed but the distribution of these benefits will be highly concentrated and it is unlikely that governments will be able to extract enough revenue to compensate those who are negatively affected.


Bruce Chapman	A - Strongly disagree B - Strongly agree		Technical change is always with us, and it is part of the process of innovation. There is no evidence that rapid technical change creates higher unemployment, although for some workers there will inevitably be displacement. Seen any blacksmiths around lately? We must remember that technical change, including the use of robots, increases efficiency and this tends to decrease prices and increase incomes. High purchasing power leads to higher levels of expenditure and this helps create employment. although the types of jobs inevitably are different.


Ken Clements	A - Disagree B - Agree


Deborah Cobb-Clark2	A - Disagree B - Agree		Robots may create non-employment in the traditional sense; it is less clear that they will produce unemployment.


Janine Dixon	A - Agree B - Disagree		In a monetary sense, technological gains will provide enough to "compensate" workers, but poorly-managed welfare dependence leads to many problems. The premise of these questions, "holding labour market institutions and job training fixed," is key. Training strategies need to anticipate, as best they can, areas in which jobs are not at threat of automation, and encourage trainees in these directions.


Uwe Dulleck	A - Disagree B - Strongly Agree		On question A: I find it likely that further automation (robots and AI) will shift more employment towards personal service jobs, I find it likely that labor demand in these sectors will on the aggregate be more stable, hence people will transition quicker. On question B : The crux is in the could ... as usual there will be more than enough gains from automation but that they will be shared in a way that allows the workers losing from this transition to be compensated is very unlikely.


Saul Eslake	A - Disagree B - Agree		I'm not persuaded that robots and AI will have any greater impact on long-term unemployment than previous waves of technological change (which were often accompanied by similar predictions). I could more readily accept that these kinds of technological change could result in more widespread short-term unemployment. In answering these questions I have not interpreted the "job training fixed" stipulation too narrowly.


ALLAN FELS	A - Disagree B - Agree


Gigi Foster	A - Disagree B - Uncertain		AI is the latest form of capital that throughout its history has combined well with skilled humans, leading in the latter part of the last century to what economists christened "skill-biased technological change" in which people with higher levels of human capital saw their wages go up faster than those with lower levels of human capital due to being able to combine their skills with technology. As with previous technological changes, the AI revolution will not only increase the marginal productivity of labour for highly-skilled workers, but will significantly change the content of many jobs. Human mediation will nevertheless still be required to apply AI effectively to production and distribution processes, and some of those mediation and support jobs will not require exceptional levels of intelligence or skill. Human service provision in certain industries will also continue to constitute a differentiated product from AI provision, with accompanying demand for it. The binding constraints we see now on productivity in many industries will change, and the nature of many jobs will change, but we will adapt without huge increases in the count of long-term unemployed even if the government does nothing to help.


John Freebairn	A - Uncertain B - Disagree		Robots and AI is likely to be a game of "job destruction and creation" along the lines of the steam engine, electricity, mechanisation,telephone, TV, etc of history. Automation and AI will be introduced if they offer lower cost ways of production and new or better products than the replaced technology in a positive sum game. The surplus generated by the positive sum game of lower production costs and more utility per dollar spent finds its way into additional outlays and some job creation. Yes, the process of job destruction and job creation also involves devaluation of some skills and additional demands for other skills creating structural unemployment. Government taxation of a share of the additional real income generated by adoption of robots and IA provides additional resources for short term income support and retraining of displaced employees to better adjust to the additional new jobs created. But, this increase of funds is limited.


Paul Frijters	A - Disagree B - Strongly Agree		A very tricky question that touches upon many issues. So I am going to take the questions literally. I think that in stead of being officially unemployed, any large group of displaced workers is going to be in hidden unemployment or non-participants, but not in official employment. I also think the number of displaced workers will be relatively small. On the second question ('could be used to compensate') I can only answer "of course benefits from technological progress could be used to compensate the displaced. They could be used to save the Murray-Darling wetlands. They could be used to paint parliament pink. 'Will they be used to compensate'? No idea. Maybe.


Geoffrey Kingston	A - Disagree B - Agree		Economic history since the invention of the Spinning Jenny suggests a negative answers to Question A and a positive answer to Question B.


Michael Knox	A - Disagree B - Agree		Like previous expansions of the capital base, we might reasonably expect that the jobs that robots fit will be replaced by others jobs that may not yet exist. We also expect that just like previous cycles of job replacements, the process will be far from seamless.


Tony Makin	A - Uncertain B - Uncertain		It seems unrealistic to assume labour market institutions and job training would remain fixed as workers became displaced. The IT revolution saw jobs lost (eg typists, clerical staff) but new ones were created in the services sector that required specialist skills. Unlikely that as many service sector jobs be created with robotics however.

Doug McTaggart

A - Uncertain

B - Agree

Other things constant, technological progress changes the mix of jobs on offer with different skills required. In the absence of retraining opportunities, some workers may enter long-term unemployment. But other workers might find employment.

We have a long history of technological progress increasing productivity, including labour productivity, leading to rising living standards. As the skills mix of the workforce changes with technological change, we typically use the tax and transfer system to compensate losers in the process. This will not change with the current and prospective technological shifts arising from the use of robotics.


Flavio Menezes	A - Disagree B - Agree		The unparalleled technological innovation experienced since the industrial revolution has resulted in significant improvements in well-being and no great shifts in trends in the unemployment rate. Of course, technological innovation destroys particular jobs, and there may be high social adjustment costs, but there is no basis to suggest that rapid technological change will raise the unemployment rate in the long run.


James Morley	A - Strongly disagree B - No opinion		It seems highly unlikely to me that AI will lead to a big increase in long-term structural unemployment. The realms in which AI can substitute for labour are quite limited in comparison to the size of the current economy (meanwhile, if AI non-trivially increases the size of the economy, that will just lead to more and/or better paying jobs). Also, most of any workers displaced by AI are likely to have transferable skills and don't live in the equivalent of the U.S. mid western rust-belt. Instead, they live in large dynamic cities with other employment opportunities... This whole "robots-taking-white-collar-jobs" scenario is overblown science fiction. Its only benefit is to focus minds on the value of general skills education rather than specific professional training.

Lionel Page

A - Uncertain

B - Strongly agree

Robots and AI will offer opportunities but also challenges. Part of the work force will see lower demand for the skills which could foster structural unemployment. At the same time the simplification of tasks associated with these technical changes could facilitate transitions to new activities.
In any case, new opportunities for economic activities are bound to appear. The associated gains for the economy should make it possible to compensate the workers who are negatively affected (e.g. early retirement, welfare benefits). A problem though that this win-win scenario has often failed to eventuate in the past.


A Abigail Payne	A - Uncertain B - Disagree


John Piggott	A - Agree B - Uncertain


John Quiggin	A - Agree B - Agree		The crucial factor is " Holding labour market institutions fixed". Currently labour market institutions are designed to drive down wages and conditions, and weaken the bargaining power of workers. Any technological change, even if it is neutral or labour-complementary is most likely to be implemented in a way that reflects this design. The use of the Internet to drive the "gig economy" is an example. The gains of technology could be distributed more fairly but, with current institutions, they will probably not be.

Rana Roy

A - Uncertain

B - Strongly agree

Question A: No, I do not think the outcome in terms of the unemployment count can be derived from these specifications alone. The extent to which robots and artificial intelligence are put to use and their impact on the number of hours worked and the number of workers employed is a composite result of factor prices, aggregate demand and government policies – and these in turn will reflect the relative social power of the relevant “factors” (aka social classes). Where labour is strong, the reduction in hours required for a given level of output can be translated into a reduction in hours worked per worker rather than a reduction in the number of workers employed – or into an increase in aggregate demand and a sufficient increase in output to maintain the level of hours worked at the higher level of productivity. Where labour is weak, the result may be greater and more prolonged unemployment – or it may be a displacement from high-wage employment in the mainstream economy into low-wage employment at its periphery. But then, so long as labour is weak, the downward pressure on wages will act as a brake on continuing automation. Today in Britain, as Paul Mason observed recently, automated car-washes are being replaced by six men with a rag!

Question B: Yes, this statement is true under almost all conditions. Up to and including the point of total automation, the increasing use of robots and artificial intelligence delivers a net gain to society as a whole. It is then a question of how we choose to distribute that gain. At one extreme, we could continue to allow the primary gain to fall to the owners of capital and then attempt to recover part of it in taxes with which to “compensate the losers” to a partial extent – be it in the form of the dole, or “work-for-the dole”, or “retraining” for new “jobs” outside the trade-exposed sectors (for example: in car-washes, shoe-shine stands, etc.). At the other, we could recognise the fruits of centuries of scientific and technological progress as the collective patrimony of all and distribute to each, via a universal income, the consumer goods produced at a decreasing and ultimately zero marginal labour cost. Right now, however, we are stuck in a trap of stagnant wages, stagnant productivity, and stagnant growth. The immediate question therefore is how to spring this trap and return to the high-wage, high-productivity growth path that could take us to the threshold of abundance – not how to distribute the fruits thereof once we arrive thereat.


Jeffrey Sheen	A - Disagree B - Agree		How employment will be affected depends on whether robots and AI are complements or substitutes for human effort. Substitution will compete away jobs, and if this were always true, eventually none of our labour may be able or willing to compete. However I think this is unlikely, and my best guess is that more of these technologies will complement human effort in the short and longer run, and thus require improved updating of our skills to maximise the potential.


Hugh Sibly	A - Disagree B - Agree		History suggests that new technology does not increase unemployment after a period of adjustment. New technology would be expected to raise productivity and thereby raise national income. In principle it would therefore be possible for winners to compensate losers. Whether this happens will depend in part on policy. It might be expected that appropriate education and training policies could assist with the adjustment process and thereby reduce some of the negative impacts of the technological change.


Julie Toth	A - Disagree B - Agree		To paraphrase a paraphrase, change is the only constant. Every new technology fosters creative destruction to a greater or lesser degree (or in modern business-babble, disruptive innovation). Smashing the Spinning Jenny (or the smart-phone/smart-car/smart-telly) can only delay its adoption, not avert it entirely. It is up to all of us how we manage these transitions, and for whom. ( apologies to every famous economist who has ever commented on the aggregate benefits of tech change, which is probably all of them)


Joaquin Vespignani	A - Disagree B - Agree


Elizabeth Webster	A - Uncertain B - Strongly agree		Whereas potentially, artificial intelligence can deliver strong benefits to our residents, whether it will or not depends on how well we use this technology to increase our exports and position in global value chains; measures governments take to support well-paid jobs; and the collective efforts of civil society to dampen extreme wage inequality. A growth in low-paid, low-hours jobs is a more likely scenario than a growth in unemployment. Training programs with no end job do not have a track record for helping the unemployed. Its best to get people into jobs or enable them to start new businesses before you train them.