Prior to the first Democratic debate last year, CNN.com ran a feature in which celebrities posed hypothetical questions to the candidates. Actor-turned-entrepreneur Ashton Kutcher asked a question that has been on my mind (and the minds of many economists, journalists and technologists): “During your presidency, you will be faced with a robotic revolution -- for example, driverless cars and semi-trucks -- as machines take skilled and unskilled jobs from Americans. This will further hollow out the middle class and divide society. What do you propose we do as a nation to bridge the gap without stifling innovation? What will you do as president to maintain a country where everyone has upward mobility?”
Kutcher is raising the specter of a future without enough work to go around -- a scenario often referred to as “technological unemployment.” Until recently, the idea that technology might, on net, replace human labor has been dismissed by economists as the “Luddite Fallacy.” Not to worry, economists have told us: History shows that in the aggregate technology always creates more jobs than it destroys.
But increasingly there is cautious acknowledgement that history may no longer be our guide. Last September, at the “World Summit on Technological Unemployment,” economists Joseph Stiglitz and Larry Summers as well as former Secretary of Labor Robert Reich all voiced qualified concerns about technology displacing workers and contributing to the growing economic divide.
Technologists weighing in on the subject have generally offered more dire warnings. Silicon Valley entrepreneur and author Martin Ford argues in his recent book, “Rise of the Robots,” that the coming wave of automation is an altogether different animal from what has come before because no sector of employment will be insulated from its impact. Filmmaker and educator CGP Grey makes a similar case in his compelling short film “Humans Need Not Apply.”
Indeed, it is difficult to imagine that the breathtaking advances occurring in machine learning, robotics and computer vision won’t have significant impacts on the labor market. Last October, Tesla Motors released a software update that enabled 50,000 of its already existing vehicles to drive semi-autonomously, and CEO Elon Musk has stated a goal of full automation by 2020. With Google and Uber pursuing driverless cars alongside traditional automakers, and President Obama proposing $4 billion of investment over 10 years for autonomous vehicle research, how much longer will driverless taxis remain the stuff of science fiction?
Similarly, Amazon, Walmart and other big-box retailers are investing heavily in drone-delivery technology that could threaten the jobs of many delivery drivers, and inside Amazon warehouses shelf-moving robots that locate items for shipment are replacing the legwork of humans. Meanwhile in China, the manufacturer Foxconn is ramping up “lights-out” factory production where robots in worker-less factories carry out their assembly tasks in the dark to cut costs.
Automation also looks poised to take a bite out of the professions. Predictive algorithms are now able to sift through mountains of legal documents and evaluate how relevant any particular document is likely to be to a case. Natural language-processing algorithms have made near real-time translation a reality (e.g. Google Translate) while other algorithms are now generating routine financial and sports articles.
Given time, it seems that nearly any work that is relatively predictable will be susceptible to automation, and “relatively predictable” turns out to describe a surprisingly large proportion of the work people do. A 2013 Oxford study estimated that 47 percent of total U.S. employment is at high risk of being automated over the next two decades.
To be sure, there are reasons to be skeptical of the claim that new automation technologies will permanently put a significant fraction of humanity out of work. Thus far, humans have always found ways to adapt to new technology and generate ever more goods and services to sell to each other.
Additionally, demographic trends may mitigate the impacts of automation. With the Baby Boomers beginning to retire, there may be increased demand for workers to fill the void.
Another line of skepticism acknowledges that automation technologies could cause significant job losses, but argues that as production of goods and services becomes automated, prices will fall and offset the loss in wages. Martin Ford counters this line of argument by pointing out that an economy with broadly falling wages and prices could be a recipe for a deflationary spiral -- the likes of which were last seen during the Great Depression.
A more nuanced take on the impact of automation on jobs is presented by labor economist David Autor of MIT. Autor has analyzed employment data from the U.S. and Europe and found that over the past several decades the share of total employment in middle-wage jobs has declined significantly while the share of both low- and high-paying jobs have increased. Autor refers to this bifurcation of the labor market as “job polarization” and attributes it to the advent of information technologies that have eliminated many of the routine clerical, office and sales jobs of previous decades.
However, Autor also finds that information technologies have complemented the skills of many knowledge workers and, rather than substituting for them, has made them into more productive and in-demand employees. Autor believes that for the foreseeable future jobs involving creativity, adaptability, problem solving, interpersonal interaction and dexterity will remain relatively insulated from new automation technologies.
Of course, no one knows how things will unfold. To quote the late Yogi Berra, it’s tough making predictions -- especially about the future. I’ll hazard a fairly tame prediction here: It is unlikely that our next president will face a crisis of mass technological unemployment during his or her term of office. The clearest threat to millions of jobs will be posed when fully autonomous cars and trucks are adopted en masse. While this day will almost certainly come within our lifetimes, the transition to autonomous vehicles will likely be incremental and unfold over a decade or two as technologies improve, prices fall and older fleets of vehicles turn over.
In the near-term I think the bigger concern is not outright technological unemployment but rather technological underemployment. New technologies could well accelerate the erosion of stable middle class and service industry jobs, leaving more workers juggling multiple part-time jobs or finding only short-term employment without benefits.This is particularly worrying because for far too many Americans, our current economy is not delivering. Despite the recovery, wages are stagnant and many are burdened by debt. In a recent household survey conducted by the Federal Reserve, only 53 percent of respondents said they could cover a hypothetical emergency expense of $400 without selling something or borrowing money. Thirty-one percent of respondents reported forgoing some form of medical care in the previous year because they could not afford it.
Without a sound policy response, accelerating automation is likely to exacerbate the economic insecurity so many are experiencing. But with appropriate policy we can enjoy the benefits of technological innovation while ensuring economic security for more Americans.
What might some of those policy responses be? A number of policies have been suggested by forward-thinking economists and policy wonks, ranging from modest proposals to more contentious remedies that are unthinkable in today’s polarized political climate:
1.) Tuition-Free Community College: President Obama has already put this on the table. Making community colleges tuition-free will help increase availability of vocational training for young people and, for mid-career adults, community colleges offer one of the most accessible avenues to job retraining.
2.) Invest in Infrastructure and Green Energy: America’s infrastructure is in disrepair and we need to shift to renewable energy sources. Repairing our infrastructure, installing solar panels and weatherizing inefficient buildings will be a labor intensive undertaking. With historically low interest rates these investments should be no-brainers.
3.) Expand EITC and Increase the Minimum Wage: Automation technologies are likely to put downward pressure on wages for many workers. A more generous and comprehensive Earned Income Tax Credit can help offset the loss in wages and reward work. A concurrent boost of the minimum wage will create a wage floor that prevents employers from holding down wages.
5.) Work Sharing: If there isn’t enough work to go around, reduce the work week and share what remains. Germany successfully implemented a work-sharing policy with wage subsidies during the Great Recession.
6.) Employee Ownership: Owners of automation technologies stand to reap considerable rewards as they shed labor costs. Giving employees an ownership stake will ensure more broadly shared profits as automation technologies are deployed. Tax incentives and subsidies can give a boost to companies offering employee stock ownership plans and support the establishment of worker cooperatives.
7.) Job Guarantee: Willing and able to work but can’t find a job? Then the federal government must become the “employer of last resort,” as it was during the Great Depression with programs like the WPA and CCC. Duke’s own William Darity Jr. in the Sanford School has advocated for a job guarantee. Other economists have recommended that the federal government fund local non-profits to hire those who can’t find work.
8.) Basic Income: Give all adult citizens a modest source of unconditional non-labor income to cover the essentials ($500-$1000/month). This would ensure a floor beneath which no one could fall and eliminate the welfare trap that can occur with traditional means-tested benefits. Advocates propose that a basic income could be funded through a variety of sources including a carbon tax and the establishment of a sovereign wealth fund modeled after the Alaska Permanent Fund.
David Winski, from Chicago, is a Ph.D. candidate in Computational Biology and Bioinformatics at Duke.