ROBOTS, ARTIFICIAL INTELLIGENCE AND THE 'FUTURE OF WORK'_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.

The Australian results were mixed, just like the US results, although Australian economists generally disagreed with the proposition that robots and AI were net substitutes for labour, whereas US counterparts skewed toward believe that they are on the whole substitutes. Both Australian and US economists on the whole tended to agree that the benefits of new technological adoption would create sufficient benefits that in principle redistribution could occur to ‘compensate the losers’. Whether this would actually occur, as a political act, the US economists were broadly skeptical.

There was a fair amount of ambiguity and wiggle room in the questions. First, labour market institutions and job training were assumed held fixed, which is a useful analytic simplification, but eschews a primary mechanism by which adaptation actually takes place. Also, ‘for long periods’ is open to interpretation along the lines of the economist’s priors. ‘Substantially’ negatively affected is a subjective judgment call. The upshot is that this poll was inevitably somewhat of a Rorschach test about political economy priors.

Economic theory is not that helpful here, because the effects could go either way depending on the sign and relative strength of income and substitution effects, elasticities of supply and demand, as well as parameters on hysteresis, path dependency, bounded rationality, and so on. Are new technologies more like substitues or more like complements for skilled (or unskilled) labour? On what margins? So it’s really an empirical question. But these also are difficult to generalize. Are all new technologies really alike in their economic impact?

But in another important sense, from the historical evolutionary perspective, we know the answer to both of these questions. The adoption of new ideas and technologies is the only thing that causes economic growth and development. And it does so because it both substitutes for existing resources (productivity growth) and by complementarity with existing resources (innovation).

Thus the only meaningful question here is about the distributional consequences, but the way the questions are implicitly set up these are indicated to be public policy questions – e.g. should the government forcibly redistribute gains from adoption of innovation; should governments tax some citizens to publically supply retraining, etc?

However, a better question is whether economists believe that market institutions (and price signals) actually work sufficiently well that the adaptation process of resources exiting from low values uses and moving into higher value uses (whether labour or capital) will proceed more smoothly than government (public policy) endeavours to intervene in these dynamic evolutionary processes. The public choice question about rent seeking opportunities, and risk of capture was not addressed nor implied in these questions.

Furthermore, the technological innovation (robots or AI) is surprisingly generic. What if we say, specifically, self-driving cars? In doing so you can see that the loss of employment from drivers (truck drivers or taxi drivers, say) is compensated on the consumer side by lower prices or better service. So it’s not obvious that the redistribution and welfare story is only about labour markets. The main beneficiaries of new technology adoption are largely consumers.

Share this with your friends