Companies around the world are dreaming up a new generation of technologies designed to monitor their workers—from Amazon’s new employee wristbands, to Uber’s recording whether its drivers are holding their phones rather than mounting them, to “Worksmart,” a new productivity tool that takes photos of workers every ten minutes via their webcams. Technologies like these can erode workplace privacy and encourage discrimination. Without disregarding the importance of those effects, I want to focus in this post on how employers can use new monitoring technologies to drive down wages or otherwise disempower workers as a class. I’ll use examples from Uber, not because Uber is exceptional in this regard — it most certainly is not — but rather because it is exemplary.
As a first example, consider how workplace monitoring generates data that companies can use to automate the very tasks workers are being paid to perform. When Uber drivers carry passengers from one location to another, or simply cruise around town waiting for fares, Uber gathers extensive data on routes, driving speed, and driver behavior. That data may prove useful in developing the many algorithms required for autonomous vehicles—for example by illuminating how a reasonable driver would respond to particular traffic or road conditions.
Of course, the automation of whole jobs isn’t always possible—today, for example, more and more observers are skeptical that fully autonomous vehicles will arrive anytime soon. But in those cases, companies may still use data generated from monitoring workers to automate particular tasks. And a rational firm, when deciding which tasks to automate, would have reasons to target the specialized skills that give workers bargaining power. To illustrate: with GPS data from millions of trips across town, Uber may be able to predict the best path from point A to point B fairly well, accounting not just for map distance, but also for current traffic, weather, the time of day, etc. In other words, its algorithms can replicate drivers’ subtle, local knowledge. If that knowledge was once relatively rare, then Uber’s algorithms may enable it to push down wages and erode working conditions.
Sometimes, of course, even task automation isn’t possible. But even then, firms may use monitoring technologies to push workers to perform harder, faster, and for less. Uber of course knows exactly where its drivers are and how much money they’ve made, which may allow it to price discriminate, sending extra fares to drivers who might soon log off. By managing drivers’ expectations, the company may be able to maintain a high supply of drivers on the road waiting for fares. The net effect may be to lower wages, since the company only pays drivers when they are ferrying passengers.
Meanwhile, many other companies use monitoring technologies to “sweat” workers in more mundane ways. For example, the “telematics” devices that trucking and other logistics firms use to monitor drivers’ speed and delivery efficiency may make it possible to ensure they are performing quickly and according to specific company protocols with minimal in-person supervision. Major retailers and fast food workers use point-of-sale devices that not only track inventory, but also determine how quickly cashiers and others are working. Such knowledge about how workers are using their time is quite valuable. With it, firms can squeeze out nearly all downtime—scheduling workers for short and unpredictable shifts, or reassigning them to other tasks whenever they’re unoccupied.
Finally, new monitoring technologies can help firms to shunt workers outside of their legal boundaries through independent contracting, subcontracting, and franchising. Various economic theories suggest that firms tend to bring workers in-house as employees rather than contracting for their services—and therefore tend to accept the legal obligations and financial costs that go along with using employees rather than contractors—when they lack reliable information about workers’ proclivities, or where their work performance is difficult to monitor. In a Coasean approach, the challenge of monitoring workers outside the firm may be a transaction cost that encourages the firm to bring them inside as employees. In some variants of “efficiency wage” theory, higher wages or implicit promises of long-term employment can incentivize workers to perform diligently. At a certain point the costs of employment—which include obligations under wage/hour, antidiscrimination, workers’ compensation, and collective bargaining laws—may be outweighed by the benefits of having a reliable, high-performing workforce.
Yet where firms can develop near-perfect knowledge about workers’ performance, the calculus changes. Again, Uber illustrates: through its software, and especially through customer ratings, the company can effectively oversee and discipline an enormous workforce of contracted drivers at very low expense. In other words, it need not classify drivers as employees to ensure that they perform well. To be clear, there are very powerful arguments that Uber drivers are in fact the company’s legal employees, especially under the “ABC test” for employment recently adopted, for certain purposes, by the California Supreme Court. My point is that Uber is highly unlikely to reclassify drivers as employees until forced to do so by courts, legislatures, or public opinion.
And again, Uber is far from alone. Building managers that subcontract janitorial and security work may be able to use electronic badges, or biometric devices such as thumb-print access, to determine when janitors enter and exit a particular floor or office, and therefore to determine how quickly they are working. Franchisors in fast food often have detailed information about franchisees’ operations and therefore their labor costs, since franchisees use their proprietary order-tracking and management software. Advanced monitoring efforts can therefore give firms the best of both worlds: the powers traditionally associated with employment, like the ability to set wages and maintain a constant threat of termination, without the duties and costs associated with employment.
Now, even if automation, close management, and outsourcing of workers may lower wages and the like, don’t they contribute to overall economic efficiency? Yes, in many instances. Automation may limit the amount of time that must be spent on particular tasks, and may eliminate mind-numbing jobs that nobody really wants. Closer management may ensure that workers’ time is not wasted. Fissuring may create economies of scale, for example by enabling the growth of large building services contractors that specialize in cleaning and security.
But such considerations should not necessarily dominate social policy. Many basic labor regulations—including wage/hour laws, child labor laws, workplace safety laws, and collective bargaining laws—reflect a political commitment to values that stand apart from efficiency, such as individual autonomy, equality, and fairness. Moreover, as discussed above, firms can often use the same technologies in multiple ways—not just to improve efficiency (that is, to enable workers to produce more with the same effort), but also to sweat workers, by requiring them to produce more by working harder. Bar code scanners are a good example: they both simplify inventory and sales and help firms to set a faster pace of work.
Most importantly, firms deploy new technologies into a social context—the workplace—in which they enjoy substantial power over workers. Moreover, as Harry Braverman and Samuel Bowles have explored, firms often introduce workplace technologies for the express purpose of maintaining their power over workers. By limiting workers’ discretion, employers can often reduce labor costs, and therefore maintain profits. Monitoring technologies help them to do so.
Now, to be clear, not all employers will do so. Some will treat workers quite well regardless of what existing law and technology enables; others will attempt to skirt moral and legal obligations whenever possible; and still others will be in the middle, and thus quite responsive to legal and moral regulations. Moreover, many workers benefit from this process, as technological innovation creates new production opportunities or demand for more highly-skilled labor. But all else being equal, firms will have incentives to use technology to gain a power advantage over workers, especially in highly competitive markets.
Why can employers do this? As I explored in an earlier post, the background rule of employment-at-will is quite important here, as is the deeply-rooted assumption that management “owns” the enterprise and therefore can organize production as it wishes. Employees’ limited privacy rights are also a part of the story. Under U.S. law, employers have near-plenary powers to monitor workers’ behavior in the workplace through technological means, so long as doing so is not discriminatory. The major limitations are that firms cannot search or monitor workers in ways that would be “highly offensive to a reasonable person,” and that health and genetic-related data are generally off limits. But work-related acts, in the workplace, are generally fair game; and all the acts I’ve described above fall into that category.
What to do? Algorithmic accountability in hiring and management processes, while a laudable goal, wouldn’t really address this issue, since it only aims to ensure equal treatment of similarly-situated workers. Another option would be to sharply limit firms’ abilities to gather and utilize work-related data via legislation. This strikes me as a political non-starter, and not necessarily desirable in any event. Automation often replaces monotonous tasks, so hampering its development would often be counterproductive. Moreover, the optimal rules around data gathering and usage may vary greatly from worksite to worksite.
This suggests, in my mind, a strategy of worker empowerment and deliberative governance rather than command-and-control regulation. At the firm or workplace level, new forms of unionization and collective bargaining could address the everyday invasions of privacy or erosions of autonomy that arise through technological monitoring. Workers might block new monitoring tools that they feel are unduly intrusive. Or they might accept more extensive monitoring in exchange for greater pay or more reasonable hours.
Workers could also be woven into state and federal policy-making in a more sustained fashion. They could be guaranteed seats on new administrative boards established to consider responses to technological change, for example, or given a formal role in a more robust industrial policy that aims to create high-skill jobs and to train workers to take them on. Such proposals update a classic theme in information law: the potential for new technologies to encourage greater democratization. The twist is that to exert democratic control over contemporary technologies, we may need to repurpose a paradigmatic “old-economy” tool: labor unions.
Brishen Rogers would like to thank Jesse Williams for excellent feedback and editorial assistance.