Why the Republican Party wins when robots take your job
Research suggests that job automation is only a matter of time. Here's why the American right will benefit.
i. Me and you and everyone we know
One evening in late July of last year, as a friend’s birthday celebration wore on and the night steadily progressed from wine to beer, the conversation turned to Uber. We’d trotted out the usual talking points: the ossified taxi industry was about to face the same fate that befell horse and buggy drivers; driverless cars were the future. The rising tide of progress, we observed with some regret, doesn't always lift all boats; while many would win as a result of automation, drivers would lose.
With the coming presidential election a constant presence in the backs of our minds, I remember arriving at an alarming thought amidst the banter: what would happen politically if the other jobs that allowed millions of people to earn a middle-class living — without the need to attend college — went the way of the taxi drivers? If broad support for outsiders like Bernie Sanders and Donald Trump was brewing when the economy was strong, I wondered, what would happen if countless workers were made redundant? I decided to turn to the data.
While machines tended to replace manual labour throughout the past century, computers today devise new ways of predicting earthquakes based on imperceptibly slight audio patterns, and identify tumors in x-rays better than their human counterparts. It’s not far-fetched to worry that machines may render vast swaths of the job market obsolete.
Which jobs are most likely to vanish in the near future? In 2013, Oxford University’s Carl Frey and Michael Osborne broke down the individual tasks that make up hundreds of occupations, and assessed the probability that each of these will be automated (for a rundown of their methodology, check the bottom of the page).
Curious to see if your accounting major is going to end up being a viable long-term career choice (spoiler: unlikely)? Or, should you take on nursing, or perhaps become a teacher? Check the chart below (circle sizes represent the number of Americans in these occupations as of 2015):
If you’re in a career that involves a good deal of social perceptiveness, such as a healthcare social worker or a speech language pathologist (both at < 1% likelihood of automation), there's little need to worry about automation. If your days are spent filled with repetitive, manual tasks, however, such as those of telemarketers or data entry clerks, the prospects of long-term employment in your field are discouraging.
Most of us fall somewhere in between those two extremes, and may be happy enough knowing that robots won’t be stealing our jobs quite yet (here’s looking at you, 601,000 medical assistants). Nonetheless, there’s a an unnerving takeaway here: if only the top 15% of the most vulnerable occupations disappear, over 50 million people risk being out of work. These jobs, of course, won’t vanish all at once; neither will automation eliminate professions altogether, but dull and hone the importance of human involvement in particular tasks as technology progresses. Bearing these caveats in mind, however, yields little comfort: if even a small number of these occupations disappear completely, and substitutes fail to arise in time, the impact on civil society is likely to be colossal.
ii. The 85%-ers
Considering the number of people who are liable to lose their jobs with few alternatives, I began to wonder where the greatest impact would be felt. To find out, I looked at where highly automatable jobs — those with an 85% likelihood of automation — are concentrated.
Why the 85% cutoff? For months, I’ve been thinking about a map which showed truck drivers — who have a 79% chance of becoming automated — to be the most popular job in over 20 states. Truckers make up over 1% of the U.S. working population, and driving rigs is one of the few options for people with no college education to make a decent, middle-class income. Considering their ubiquity, one begins to wonder what would happen if their numbers were, say, reduced by even a tenth in a short period. That’s altogether likely: in October of 2016, a self-driving truck made the 100+ mile highway leg of its delivery trip to bring 50,000 beers from Fort Collins, CO, to Colorado Springs. If truck drivers’ jobs were so close to being automated, limiting my analyses to jobs with an even higher, 85% automation likelihood, struck me as a more conservative way to assess realistic job loss.
Below is a map of the proportion of jobs that would be lost in each county if these highly automatable occupations were to disappear.
Hover over the legend to see all counties with likely similar job losses, or find your county to see what proportion of its jobs are likely to become robo-jobs in the future:
Were jobs with a high risk of automation to vanish, a staggering number of people in nearly every county in America would be out of a job. Even in the northeast, where states can be expected to be somewhat less affected, many counties would lose nearly 30% of jobs.
The larger question, however, remained: what does this mean politically?
iii. The White Drift Right
In light of America’s governmental volatility in 2017, I naturally wondered how the loss of an inordinate number of jobs to automation would impact the country’s biggest political contest — the presidential election. All other things being equal, then, would the Republicans or the Democrats benefit?
Although there’s little research on the political effects of automation in the current era, an unusual paper by David Autor, David Dorn, Gordon Hanson, and Kaveh Majlesi emerged in December 2016, entitled “Importing Political Polarization? The Electoral Consequences of Rising Trade Exposure.” The authors carefully explored the political impact of America opening its borders to trade with China, and unearthed a critical finding:
When trade adversely affected employment prospects in white areas, voters reacted by veering to the right, electing conservative Republicans. Conversely, in predominantly non-white districts, voters were much more likely to elect progressive Democrats. The political shift from the center to the extremes was clear.
If voters react to their jobs being usurped by robots in similar ways as they did to Chinese economic competition, what impact will this have on presidential races? To answer this, I looked at the number of white and non-white workers with highly automatable jobs (i.e., with an 85% likelihood of becoming automated) in each state.
In 90% of states, jobs that are highly susceptible to automation are filled by white Americans. Politically, however, most states are considered safe — regardless of most social or economic shifts, they consistently vote for one party. These states are all but locked in, and automation is less likely to impact elections therein.
The states that truly decide elections are few in number, and are called "swing states." While these vary from one election to the next, they've recently consisted of this small group.
In these states, people at risk of losing their jobs due to automation tend to be white — this tips the scales in every single swing state in favor of the Republican party.
These trends are neither predictive, nor isolated from a myriad of other, perhaps more impactful, factors, such as candidate appeal and the state of the economy. Nevertheless, the more I considered the impact of job automation in the context of the current political climate, the more I thought of Kurt Vonnegut’s first novel, The Player Piano (1952), wherein automation has rendered human workers obsolete. At one point, the protagonist, a well-off factor manager, happens upon the leader of an anti-automation resistance group, who declares,
“Things, gentlemen, are ripe for a phony Messiah, and when he comes, it's sure to be a bloody business… Sooner or later someone's going to catch the imagination of these people with some new magic. At the bottom of it will be a promise of regaining the feeling of participation, the feeling of being needed on earth — hell, dignity.”
In light of the current political fervor, such forecasts seem all too prescient.
Methodology and Data Sources: The first visualization relied on Frey and Osborne's study of the likelihoods that various careers will be automated, and the BLS' figures for employment in 2015. Frey and Osborne employed a novel methodology in order to create their list of automation likelihoods: after breaking down 702 occupations into tasks, they compiled a list of 70 careers whose automation likelihood they judged as either certain (probability of 1) or completely unlikely (probability of 0). They then used a machine learning algorithm to classify the automation likelihood of the tasks in the remaining jobs.
The second visualization, depicting county-based job loss likelihood, was compiled using a PUMA to county crosswalk compiled by The University of Michigan's Population Studies Center. It employed Frey and Osborne's job classifications, paired with U.S. Census Data on county-based employment, sourced from IPUMS, from 2011. This was the final year during which the PUMA code crosswalk was available prior to the redrawing of PUMA boundaries.
The third visualization entailed the aggregation of state-level jobs by race (I counted non-Hispanic White individuals as White and all others as non-White), and relied on the same U.S. Census Data as used in the second portion of the piece.