Kiosks, robots, and the minimum wage: does a higher floor speed up automation?

Walk into a fast-food restaurant in most American cities today and there is a decent chance you will order from a touchscreen rather than a person. For people who argue about the minimum wage, that screen has become a kind of Rorschach test. One side sees proof that higher wage floors push employers to replace workers with machines. The other sees a technology that was coming anyway, wages or no wages, and a debating point dressed up as evidence.

Both readings contain something true. The honest version of the argument is not "every wage increase summons a robot." It is narrower and, we think, more durable: a wage floor changes the relative price of labor and capital, and over time firms respond to relative prices. How fast, how much, and for which workers are empirical questions, and the answers so far are suggestive rather than settled.

A question of relative prices

Start with the textbook logic, because it is the part nobody seriously disputes. A business producing meals, packages, or clean hotel rooms can usually do it with different mixes of people and equipment. A dishwashing machine substitutes for some hours of hand washing. A self-checkout lane substitutes for some cashier shifts. A conveyor substitutes for some walking.

Which mix a firm chooses depends on what each input costs. When the price of an hour of low-wage labor goes up by law, and the price of the machine does not, the machine becomes relatively cheaper. Some investments that did not pay for themselves at the old wage now do. Economists call this the substitution effect, and it operates alongside the more familiar scale effect, in which higher costs mean higher prices, fewer customers, and less output overall.

Two features of this logic matter for the policy debate. First, it targets a specific kind of job: work made of predictable, repeatable tasks that a machine can already do or nearly do. Second, it takes time. A restaurant does not rip out its counter the week a new wage takes effect. It folds the new cost into the next remodel, the next location, the next equipment cycle. Economists such as Isaac Sorkin have formalized this with "putty-clay" models, in which capital is flexible when it is chosen but fixed once installed. The prediction is that short-run employment effects of a wage increase can look small even when long-run effects are meaningfully larger, because the adjustment happens as old equipment is replaced.

That timing point cuts against a common style of evidence. Many minimum-wage studies look at employment in the months or few years after an increase. If the substitution margin works mainly through slow capital turnover, those windows may miss much of it.

What the research on automatable jobs finds

The most direct test is a 2018 paper by Grace Lordan and David Neumark, "People versus machines: The impact of minimum wages on automatable jobs," published in Labour Economics. Rather than asking whether total employment fell, they asked a sharper question: among low-skilled workers, did higher state minimum wages reduce the share of employment in jobs that are easy to automate?

Using U.S. data covering several decades of state and federal minimum-wage changes, they found that it did. Increases in the minimum wage were associated with a lower share of low-skilled workers in automatable jobs, and with a higher likelihood that low-skilled workers in those jobs ended up unemployed or in worse jobs. The effects were not uniform. They were especially pronounced in manufacturing and among some groups of older low-skilled workers, a population that rarely features in a debate usually framed around teenagers.

A related study by Daniel Aaronson and Brian Phelan, published in the Economic Journal in 2019, looked at the task content of low-wage jobs. Their finding was more specific: minimum-wage increases appeared to reduce employment in low-wage jobs heavy in cognitively routine tasks, the sort of rule-following, information-handling work that cashiers do, while jobs built on manual or non-routine tasks were less affected. That pattern is what you would expect if the mechanism were technological substitution rather than simple demand loss.

That distinction matters. An automation effect that reshuffles who holds which jobs is less visible in headline employment counts than a wave of layoffs, but it is not less real for the person whose job was the one that went away.

The kiosk in the room

Self-order kiosks are the image everyone reaches for, so it is worth being careful about what they do and do not prove.

What is clear is that large fast-food chains invested heavily in ordering technology over the last decade. McDonald's, for instance, made self-order kiosks a centerpiece of a broad U.S. restaurant modernization in the late 2010s. Mobile-app ordering and delivery integrations spread at the same time, and the pandemic accelerated contactless ordering of every kind.

What is much less clear is how much of that was driven by wage floors. The chains rolling out kiosks operate in high-wage cities and in states where the federal minimum of $7.25, unchanged since 2009, is still the legal floor. Companies have also given reasons that have little to do with labor costs: larger average orders when customers browse a menu on a screen, fewer order errors, easier customization, and the ability to move staff from the register to the kitchen during busy periods. A kiosk that shifts a worker from taking orders to assembling them has changed the job, not necessarily eliminated it.

So the kiosk is a fair illustration of the substitution mechanism. It is a weak piece of evidence about the minimum wage's magnitude. Anyone who points at a touchscreen and says "that's what $15 did" is claiming more than the photograph can carry. The cleaner evidence comes from studies like Lordan and Neumark's, which compare places and periods with different wage floors rather than pointing at a technology that spread everywhere.

The case that wages are a sideshow

The strongest counterargument deserves to be stated plainly. Automation is driven mostly by the falling cost and rising capability of technology, not by wage policy. Computing power, sensors, payment systems, and software have become dramatically cheaper over decades. Labor economists such as David Autor have documented how computerization hollowed out routine middle-skill work in offices and factories across countries with very different wage institutions. If the price of the machine is falling fast enough, it will eventually undercut labor at almost any plausible wage, and a few dollars on the minimum wage only nudges the date.

There is also a version of the argument from the other direction: that automation is often complementary to workers rather than a replacement for them. A restaurant that takes orders on a kiosk may still need the same number of people to cook, clean, bag, and handle the drive-through, and may add staff if the technology increases sales.

We take this seriously. It is probably right that technology costs are the main driver of the long-run trend. But "main driver" is not "only driver." Relative prices determine when an investment crosses the line from not-worth-it to worth-it, and in which locations it crosses first. Shifting that date matters to the workers who would have held the job in the meantime, and to the teenagers who would have learned to work there. A policy that pulls automation forward by a few years in low-wage, routine jobs is imposing its cost on exactly the people it is meant to help, even if the robots were eventually coming.

The question is not whether the machine was coming. It is who loses the years in between.

The margins nobody photographs

Machines get the headlines, but many of the ways employers respond to a higher floor are invisible from the dining room. Total compensation is a bundle, and when the law fixes one part of it, the other parts are still free to move.

• Hours. The University of Washington team studying Seattle's increase toward $13 an hour (Jardim and coauthors) found that hours worked in low-wage jobs fell enough to offset much of the hourly wage gain for affected workers. The study has been contested, particularly over how it handled multi-site employers, but it is a reminder that "jobs" and "hours" are not the same measure.
• Scheduling and effort. In a study of fast-food restaurants in Georgia and Alabama after the 2007–2009 federal increases, Barry Hirsch, Bruce Kaufman, and Tetyana Zelenska found that managers responded less through layoffs than through other channels: modest price increases, tighter performance standards, and changes in how work was organized. That is a softer adjustment than automation, but workers still feel it.
• Benefits. Jeffrey Clemens, Lisa Kahn, and Jonathan Meer have found evidence that minimum-wage increases are associated with reductions in employer-provided health coverage for some affected workers, consistent with firms trimming non-wage compensation to offset mandated wage costs.
• Training. If a firm cannot pay a new worker less while it teaches them, it has less reason to hire someone who needs teaching. Masanori Hashimoto made this argument in 1982, and Neumark and William Wascher later reported evidence that minimum wages reduce formal training for young workers. Others, including Daron Acemoglu and Jörn-Steffen Pischke, found little such effect. This is genuinely unresolved.

These responses help explain why some studies find small employment effects. If employers can adjust hours, effort, benefits, training, and prices, they do not need to cut headcount much. That is good news in one sense: the disemployment story is not the whole story. But it also means the worker's gain is smaller than the wage figure implies, and some of the cost lands in places that rarely show up in a press release.

What to take from all this

The automation argument against high wage floors is neither a slam dunk nor a myth. The substitution logic is basic economics. The best direct studies find that higher minimum wages shift low-skilled employment away from automatable, routine work, with costs concentrated on particular workers rather than spread evenly. The kiosk boom is a vivid illustration but a poor measuring stick. And the strongest objection, that technology costs drive the trend, is mostly right about the long run while missing who bears the transition.

For policy, the implication is the same one that runs through the rest of this site. If the goal is to raise the incomes of low-wage workers, it is better to do it in a way that does not make their labor more expensive relative to a machine. A wage subsidy or a larger Earned Income Tax Credit raises take-home pay without changing the price an employer compares against the equipment catalog. That will not stop automation, nor should it. It just stops the law from tipping the scale against the workers with the least room to absorb it.