Political machinery: did robots swing the 2016 US presidential election?

Investigating whether the perception of having lost out to technology correlates with support for "radical political change", Carl Frey, Thor Berger and Chinchih Chen discount offshoring and exposure to globalised free trade as explanations for Donald Trump's success among working-class voters, and find instead that his vote was higher in "local labour markets more exposed to the adoption of robots", to the extent that, had exposure to robots not increased in a series of key electoral battlegrounds in the years preceding the 2016 election, Trump would not have won.

In their analysis, Frey et al remain focused specifically on the question of machines, rather than other pervasive technologies currently reshaping work, like algorithms or software. Automation, in this respect, is isolated as a factor insofar as there is '"elatively limited overlap between robot exposure and exposure to Chinese imports, offshoring, and specialisation in routine work", enabling its relevance to be independently assessed as an influence on political behaviour.

Frey et al make the important point that "the economics of automation cannot be separated from its politics", specifically insofar as workers will use political action to challenge the role of technology in reshaping "labour market outcomes" if deprived of other routes, creating unrest that threatens the position of "incumbent political leaders". This is why economic growth was held back through much of human history by leaders keen to stymy new technology in order to avoid unrest, and the industrial revolution later unleashed in Britain by governments keen to remove rather than erect barriers to innovation. Today, the question is whether democracy can establish such barriers through the ballot box.

Frey et al follow David Autor in seeing automation, and not "globalization, immigration, deunionization and manufacturing decline" as being principally responsible for changes in the distribution of wages and incomes across occupations. Rather than an "end of work" scenario, automation has seen workers move from middle-income to low-income jobs. The authors follow Autor, as well as Goos and Manning, in arguing that the displacement of workers in routine and skilled occupations has polarized the labour market. In particular, as routine jobs have been impacted by automation, workers have moved into lower-income service jobs less amendable to automation.

Specifically, these shifts have impacted on men, and specifically men without higher education. The authors cite Acemoglu and Restrepo (2017), who find that the introduction of robotics in the US has reduced wages and employment specifically among blue-collar workers without higher education - in turn the group that decisively swung behind Trump in 2016.  Trump stood to benefit from the  anxieties of this group, Frey et al propose, because his promise to repatriate and restimulate manufacturing jobs in the US also implied resistance to automation.  

Frey et al claim that "many voters are unlikely to have recognised the true causes of their concerns" in political terms, but nonetheless are revealed in polling conducted in 2014 to have blamed technology, or associated factors like lack of good jobs and unsuitable education and skills, rather than offshoring, for instance. Frey et al cite a Pew Research Centre study evidencing widespread fear of the computerization and robotization of work among voters, and extensive support for restrictions upon the industrial application of machines.