POLITICS
Here’s Why Robots Make People Vote Republican
New survey evidence connects dots between automation and voting
In the U.S., following the 2016 election, both Washington Post articles and recent research suggest individuals living in areas with more exposure to advanced technology in the workplace — what is often called automation — were more likely to support the right-wing populist Donald Trump. Similar studies in Europe show that individuals exposed to technology in the workplace form a reservoir of votes for the far right, and demonstrate greater support for the populist right and mainstream left parties.
This matters because in the coming decades millions of workers in affluent democracies are expected to lose their jobs to advanced technologies like robots and artificial intelligence. Yet, it remains unclear whether exposure to technology leads individuals to support certain candidates and parties in the U.S.; that is, whether the relationship is causal.
In our forthcoming article in Political Research Quarterly, “Technology Threats to Employment, Issues and Candidate and Party Preferences in the United States,” we examine how exposure to technology in the workplace can drive support for different candidates and parties, and we consider if candidate policy priorities play a role in any shifts in party support.
Our hunch was that individuals exposed to technology in the workplace would prefer candidates and parties that prioritize the economy, employment and the welfare state, either in contemporary elections, or with historical reputations for doing so. In the U.S. the Democratic Party has long been known as the “pro-worker” party, but in order to examine the possible effect of current party priorities we used two experiments in which the policy priorities of hypothetical candidates were randomized. What we found surprised us.
While we cannot randomize actual exposure to technology in the workplace because most people are not very knowledgeable about how likely it is that their job can be replaced by technology, we were able to approximate technology threat in another way. We randomizing whether we presented individuals with the odds that their job can be replaced by technology in the near future, using estimates from the influential Osborne and Frey study.
We conducted experiments in which we presented two hypothetical U.S. Senate candidates with varying issue priorities and asked “voters” who they would support. We used two variations both with and without candidate party labels, expecting that people might be hesitant to abandon their usual party even if they are highly exposed to technology.
While all people in the treatment condition would have been primed to think more about technology exposure, some people had very high odds of being replaced by technology, while others have very low odds. If exposure to technology causes shifts in party support on the basis of party priorities we should see that individuals highly exposed to technology who were made aware of that exposure through our randomization become more supportive of candidates prioritizing welfare, jobs and the economy.
We also consider whether there are any changes in support for candidates that prioritize immigration and free trade, since these are more salient causes of job loss that people may blame for technology-induced labor precarity.
The results for the no party labels for candidates experiment are presented below (Figure 2 in the article). While we do see that prioritizing certain issues by candidates results in higher levels of support there is almost no difference between those who are highly exposed versus not highly exposed to technology, or those that were informed of their odds of being replaced by technology and those who were not.
The results for the party labels experiment below show some differences depending on the party of the voter and candidate. But these differences are not consistent and quite modest. There is no consistent evidence that voters exposed to technology and made aware of that exposure systematically react differently from those unaware of that exposure.
Finally, we investigated whether averaging over all issue priorities, whether individuals who are exposed to technology prefer one party or the other. This would capture preferences for parties on some other dimension, though we cannot say what exactly this is, as well as account for interactions of dimensions. The figure above presents the change in probability of voting for a Democratic candidate when shown automation odds.
On average we see that job exposure to automation does lead to a pro-Republican shift of the type observed in observational research discussed above. However, we see that this effect is largest for Republicans and low-exposure independents. The fact that partisans might become more partisan in the face of threat makes some sense, but we do not see this among Democrats. This cannot explain the result for independents, of course, but perhaps low-exposure independents are more likely to be Republican leaners.
Overall, we show that at least some voters appear to shift toward the Republican Party when primed to think about automation threat, particularly if they are personally at-risk and also a Republican. This is consistent with existing research showing that automation threat is associated with support for populist right parties, and that Trump performed better in areas where technology penetration was greater. However, we can conclude that this association is not a result of candidate or party issue priorities based on our research. Clearly, many questions remain to be examined regarding how exposure to technology at work shapes voting and party support in affluent democracies, and as technological unemployment increases in the coming decades these questions will only become more important.
Author Information:
Tobias Heinrich is Associate Professor in the Department of Political Science at the University of South Carolina.
Christopher Witko is Professor of Public Policy and Political Science, and the Associate Director of the School of Public Policy at the Pennsylvania State University.
