Automation is a Distraction
Alternative Work Arrangements are the Real Threat to American Jobs
by Natalie Meltzer
Rapid advances in artificial intelligence (AI) and robotics have caused panic that American jobs will be automated out of existence. News outlets are littered with headlines like “Robots: Now coming to a workplace near you,” “Find out if a robot will take your job,” and “The robots-are-taking-our-jobs threat is real.” Billionaires Mark Zuckerberg and Elon Musk are sounding the alarm that driverless cars, robotics, and other forms of automation will lead to large job losses. Democratic presidential candidate Andrew Yang proposed a universal basic income to compensate for mass unemployment resulting from robots stealing Americans’ jobs.
But the fear of automation ignores a more insidious way American jobs have been replaced: alternative work arrangements. Companies outsource jobs through staffing agencies and contract work, undermining basic labor protections that many of us take for granted. Policy has already failed to keep up with the changing nature of work, and the window for change may be closing as companies aggressively work to codify substandard forms of employment.
Putting Today’s Fears in Context
Fears that automation will make many jobs obsolete and cause widespread unemployment are not new. The nineteenth century Luddite movement emerged in response to gig mills and knitting machines replacing wool and cotton artisans. In the 1960s, fears of automation and unemployment led President Lyndon B. Johnson to create a National Commission on Technology, Automation, and Economic Progress that would “identify and describe the impact of technological and economic change on production and employment, including new job requirements and the major types of worker displacement, both technological and economic.”
History shows that although technological advances make some jobs obsolete—handloom weavers, blacksmiths, and candlemakers are no longer common occupations, for example—new jobs emerge in their place. The Johnson-era commission found that technological change was not a driving force in unemployment and that there was no evidence this would change in the near future.
Many argue that today’s automation is different than that of the nineteenth and twentieth centuries due to the pace of change, the advent of machine learning, and the relationship between technology and workers. Exponential increases in computing power, as observed by Moore’s Law, have driven rapid advancement in computing and robotics capabilities. At the same time, costs have decreased, allowing for greater access to sophisticated technology. The availability of large data sets and highly sophisticated learning algorithms threatens to automate occupations that were once considered too difficult to program such as doctors, lawyers, and teachers. And while previous periods of technological innovation augmented workers’ abilities by allowing them to perform their job functions more efficiently, today’s technologies threaten to replace workers entirely.
The Automation-Induced Jobs Apocalypse Is Overstated
A closer look at the most influential academic research on automation illustrates how their findings have been misinterpreted to fuel the automation apocalypse frenzy. A 2013 study by Carl Benedikt Frey and Michael A. Osborne is arguably the most prominent paper examining how susceptible jobs are to computerization. The paper estimates that 47 percent of America’s jobs are at risk of being automated in the next fifteen to twenty years. The study, which has been cited more than 7,000 times according to Google Scholar, has significant limitations. First, it only evaluated the technical ability to automate tasks and did not take the costs or feasibility of doing so into account. The researchers also assumed that the data necessary to develop automation algorithms was currently available—a big assumption, given the costs and time required to collect a sufficiently large data set for machine learning.
Even if many of the tasks that make up a job can be automated, it is unlikely that jobs themselves will be mechanized out of existence. Jobs are complex, made up of many different types of tasks and skills. A McKinsey study found that while fifty percent of the activities people are paid to do have the potential to be automated by current technology, less than five percent of occupations can be fully automated. Therefore, just as in previous periods of rapid technological transformation, machines will augment existing jobs. In other words, robots will be our new coworkers rather than our replacements.
There are also significant economic, political, and social factors that may impede the adoption of robotics and automation technologies. Incorporating AI into existing processes can be an expensive, complex task, and the change would require employees with specific, in-demand skills to develop. Customers may resist machines replacing humans, especially in contexts like healthcare and teaching that involve care and nurturing. Serious mistakes during implementation—for example, self-driving cars killing pedestrians—may lead policymakers to impose regulation on the rollout of these technologies.
This is not to say that we should not be concerned about automation-related job displacement. There are certain occupations that are already being automated or at high risk of becoming automated—and these jobs are concentrated among certain populations. It is much easier to automate jobs that require mechanical precision, such as food preparation or manufacturing, and those with repetitive cognitive tasks, such as clerical and administrative office jobs. These occupations are more likely to be held by men, young people, and minorities. They are also concentrated in rural areas, heartland states, and small cities. Policies to address the future of work must take these discrepancies into account.
We Don’t Need to Look to the Future to See the Changing Nature of Jobs
Fears of future automation-induced mass unemployment ignore the fact that jobs in the U.S. have already been supplanted—not by robots but by “alternative work arrangements.” When people refer to jobs, they are generally talking about employment characterized by a wage or a salary, an implicit or explicit contract for a continuing employment relationship, a predictable work schedule and predictable earnings, and supervision by the firm paying the salary. Over the last 40 years, however, these kinds of jobs have been have been replaced by temporary, contract, and on-call work. A lack of systematic, longitudinal studies makes it difficult to evaluate just how many people rely on alternative work arrangements for their entire income. However, a comparative analysis of Bureau of Labor Statistics, Internal Revenue Service, and RAND data suggests that the percentage of workers in alternative work arrangements rose from approximately eleven percent in February 2005 to almost sixteen percent in late 2015.
Companies use alternative work arrangements to decrease costs and liability by shedding employees that are not part of their core business functions. Workers in alternative work agreements are not covered by the minimum wage and overtime requirements of the Fair Labor Standards Act, Occupational Safety and Health Administration regulations, the National Labor Relations Act, and Social Security. Employers are not required to pay payroll, state, and federal unemployment insurance taxes for these workers, nor are they required to provide workers’ compensation insurance and disability insurance. Alternative work arrangements also result in lower wages; the National Employment Law Project estimates that full-time temporary workers earn 41 percent less than their counterparts in standard work arrangements.
The growth in contingent labor converges with fears of automation through the “gig economy,” in which digital platforms use algorithms to connect workers with customers and manage their relationships. Platforms like Uber, TaskRabbit, Instacart, and Handy do not directly employ the workers engaged through their websites but instead designate them as independent contractors. While this approach suggests a degree of autonomy, the platforms aggregate data and use their technology to influence their independent contractors’ behavior. For example, Uber drivers are expected to take the route presented on the app; if an alternate route is taken, customers are entitled to complain and Uber reserves the right to retroactively adjust the fare. As a result, Uber drivers report feeling disempowered by the surveillance and lack of transparency in algorithmic management.
While only one percent of Americans regularly use online platforms to connect with work opportunities, the use of algorithms is playing an increasing role in management outside the platforms by enabling firms to automate decision-making. In the retail and service industries, scheduling algorithms crunch historical data to predict labor demand and ensure employees do not work more than the full-time threshold where they become eligible for benefits. This practice often results in continual or last-minute changes to employees’ work schedules, inhibiting their ability to find additional work to supplement their earnings and exacerbating childcare challenges.
The rise of contingent labor comes at significant social cost. Alternative work arrangements create a significant financial burden for local, state, and federal governments due to lost tax revenue and the added cost of providing services to uninsured workers. These costs are amplified by an estimated ten to thirty percent of employers who misclassify their employees as independent contractors to cut costs. Misclassification further distorts the market by giving firms that engage in the practice an unfair advantage over law-abiding competitors.
The Urgency for Action
Focusing on the needs of workers in alternative employment arrangements as opposed to those who may be displaced by technological advancement prompts different policy priorities. While reskilling programs and a universal basic income might address concerns around automation, addressing the rise of alternative employment relations requires making it harder for employers to classify regular workers as independent contractors. A provision to do just that is included in the Protecting the Right to Organize (PRO) Act, which would amend the National Labor Relations Act’s definition of “employer” to include those that exercise “direct or indirect” control over employees. The PRO Act passed the House of Representatives in February 2020 and the labor movement is mobilizing around its passage in the Senate.
Shifting the policy discussion to the concerns around alternative work arrangements also increases the urgency for action. In November, California passed Proposition 22, a ballot referendum which exempts app-based, on-demand companies from California labor law and keeps gig workers classified as independent contractors instead of employees. The proposition requires seven eighths of the California State Legislature to amend, making it nearly impossible to roll back. And executives for gig companies plan to bring similar laws to other states; Anthony Foxx, Lyft’s chief policy officer stated, “We think that prop 22 has now created a model that can be replicated and can be scaled.”
The passage of Proposition 22 has prompted a flurry of organizing among both app-based companies and workers. Companies formed a coalition to lobby for a Proposition 22-inspired policy in New York. Organized labor has mobilized in opposition: app-based drivers, rideshare consumers and labor unions filed a lawsuit challenging Proposition 22 in California, and more than 70 organizations sent a letter urging Congress to extend federal labor protections to app-based workers. But the cards are stacked in the companies’ favor: they spent a historic $205 million to pass Proposition 22, dwarfing labor’s $20 million for the campaign against it.
We are already seeing the devastating effects of Proposition 22. California’s major grocery chains are replacing full-time, benefits-receiving delivery workers with DoorDash subcontractors. Delivery workers are complaining that a Proposition 22-related change to the food delivery app Grubhub discourages tipping, decreasing their earnings by as much as 50%. Uber only applies the Proposition 22-guaranteed wage to time spent with passengers in the car, which accounts for only half of drivers’ working hours, and requires drivers to almost 50 hours per week to receive the basic health benefits included in Proposition 22.
So forget the theoretical robots. We need policies that address the erosion of fundamental labor protections caused by the rise of contingent work.
