In the 1990s, economists suddenly looked around and noticed that US income inequality had taken a sharp turn for the worse during the previous decade. After some fretful speculation, they hit upon an explanation: Technological change was responsible, and since technology occupies the status of demigod in neoclassical economic theory, it stood to reason that nothing much could be done about it. Too bad somebody forgot to tell the leftist Morena party in Mexico this gloomy news: It has implemented a massive improvement in income distribution of just the kind that mainstream economists claim is impossible.

To get at the reasoning behind America’s fatalist embrace of an economy producing less for the country’s workers, it’s necessary to revisit how the country’s economics profession processed the shock of the 1980s upturn in inequality. University of Michigan economist George Johnson published a useful article in 1997 surveying a decade of economists’ work on the subject. The supply of more-skilled workers (with some college or four-year degrees) had surged from 21 percent of the labor force in 1970 to 45 percent in 1993. This meant, other things being equal, that according to the iron logic of supply and demand employers ought to have been paying them less of a premium, thereby reducing income inequality. So why were they earning more of a premium, and worsening it? Studies had reached “virtually unanimous agreement,” Johnson wrote, “that during the 1980s relative demand increased for workers at the high end of the skill distribution and thus caused their relative wages to rise.”

Economists looked around for culprits that might inflate demand for high-skilled workers fast enough to account for their heightened earnings. Reaching into their tool kit, they found two main suspects: increased trade and technological change. As manufacturers outsourced goods production to low-wage countries, well-paid factory workers were laid off and reduced to taking low-paid service jobs. You’d rather make autos for GM than sell Chinese imports at Walmart. At the other end of the salary scale, industries such as finance seeking higher-skilled workers had swelled.

But there was a problem with this narrative: You can plausibly measure how such industrial shifts worsened income inequality, and they didn’t cut deep enough to explain the 1980s surge.

So the main culprit had to be the steamroller of “skill-biased” technological change: production methods demanding more skilled workers. A factory increasing automation of the production line requires fewer workers and more engineers to fix everything that goes wrong. At least Johnson honestly faced the implications of his argument: “Public policies to ‘do something’ about earnings inequality”—his scare quotes—such as significantly raising the minimum wage, restricting imports, or retraining workers “either present potentially undesirable side effects or may work only very slowly or not at all.” One could always recommend more education, but the structural shifts spiking demand for high-skilled labor were so far-reaching that this wouldn’t happen at a great enough scale to counteract them.

This argument presents countless problems. How did economists know that most of the demand shift was due to skill-biased technological change? They had no way to measure it directly. In reality, they were simply offering a loose hypothesis to try to account for what’s known as a statistical “residual”——the growth in inequality that wasn’t attributable to the baseline changes in the economy’s industrial structure. In other words, economists were straining to explain everything they couldn’t measure. One skeptical economist called the residual “the measure of our ignorance.”

Current Issue

October 2026 Issue

You could just as well label the main driver of inequality in the ’80s “politics” instead of “technology.” When economists talk about demand, they don’t mean the number of items customers want to buy. They mean a relationship specifying how many items consumers would buy at a range of prices. If retailers hold a sale, cutting the price, they know customers will buy more, and they hope the higher volume will help their bottom line. If they raise the price, they know customers will buy less, but they hope the larger markup will help their bottom line.

That story makes rough sense as a way to understand shopping. But it doesn’t work to explain production. Over decades, production methods obviously change, and new technologies—AI, for example—may allow a sudden leap. But new technologies don’t come along every time managers snap their fingers. A given state of the art requires, within narrow limits, a given complement of higher-skilled workers, lower-skilled workers, and capital equipment.

For example, when BMW opened a $1 billion auto plant in San Luis Potosi, Mexico, in 2019, it issued a press release bragging how it was employing the “latest industry 4.0 technologies” (whatever that means) and more automation than the company was using its older South Carolina plant. When it expanded the San Luis Potosi plant this year, it incorporated artificial intelligence to manage production, including its army of a thousand robots. Given that Mexican factory workers make only a little more per day than their US counterparts do per hour, why didn’t BMW managers use more of them? Wouldn’t less high-skilled labor, and perhaps less expensive capital equipment, enhance the company’s bottom line? Obviously, they had no idea how to and still make modern cars.

It isn’t just manufacturing that has relatively fixed methods. In the early 2000s, I investigated the Mexican government’s supermarket chain (yes, despite the many outraged attacks on New York Mayor Zohran Mamdani’s plan to introduce government-run supermarkets, such things are indeed common in the wider world). Competition from Walmart was driving large losses. So managers decided that if you can’t beat them, join them: They turned full-speed toward “the latest technology in all the stores, the same as Walmart,” one of them told me. Given how little workers were paid, I asked if there wasn’t some alternative method to use less automation and employ more of them. The answer: “It doesn’t exist.” The alternative to Walmart technology was bankruptcy. The chain adopted and prospered.

If neither manufacturers nor major retailers have any idea how to efficiently employ larger numbers of less-skilled workers, commercial vegetable producers have no idea how to use fewer. Looking at an export farm in Baja California, you can’t tell it isn’t north of the border in California—it operates with much the same ratio of workers to machinery. If Americans aren’t available to fill the jobs in California, the solution is to import workers from Mexico (many of the same laborers at Mexican farms move north with the harvest). That’s why, despite the anti-immigrant furor on the American right, the first Trump administration granted a rapidly rising number of H-2A visas for temporary agricultural workers, and that growth has continued under Trump’s second term.

When managers are pretty well restricted to using a given state-of-the-art method, they can’t fall back on the standard market saga of supply and demand to determine wages. Unlike retailers, they can’t experiment with paying workers more and using fewer of them—or paying them less and using more of them. They’re just stuck. Instead, the broader labor relations system, and the influence of collective bargaining decide how total income is divided among more-skilled labor, less-skilled labor, and owners of capital.

The MIT economists Frank Levy and Peter Temin tell this politics story well. The “Treaty of Detroit,” as Fortune dubbed it—initiated in a 1949 agreement between General Motors and the United Auto Workers—created a system where unionized firms raised wages in line with productivity gains plus inflation, and other firms fell into line with this broad accord in order to forestall unionization. When Jimmy Carter abandoned labor in the late 1970s, the treaty collapsed, and wages for less-skilled workers began eroding. Ronald Reagan fired striking air-traffic controllers and launched an onslaught against labor. The rest is history. This story explains what “technology” doesn’t: why wage inequality worsened sharply in the 1980s—the era before computers came into widespread use and when economists ritually lamented poor productivity growth.

More recently, economists have noticed the income shift from wage earners to profits. Total wages fell from 59 percent of national income in 1970 to 49 percent today, while profits escalated from 9 percent to 17 percent. This time out, the economics didn’t have to cast about for the main reason behind the shift: It had to be technology. “The machines will keep getting better and are likely to claim ever more of the nation’s income,” writes Brent Neiman in The New York Times. “Anyone who has used ChatGPT can see how much work now done by people could soon be done by technology.”

Well, not everyone. A July 2025 MIT study of firms adopting generative AI found that a few realized impressive gains, but 95 percent got “zero return.” Inequality hardly seems an inevitable outgrowth of technological innovation, but the appeal of the technology-first account remained its simplicity and social complacency. We can fight technology “and end up worse off,” Neiman writes.

Data suggests that wages have been rising faster than the cost of capital—computers, software, machinery. In other words, the relative cost of labor is rising. In this situation, the usual market parable says, firms will rely more on capital outlays and less on labor. But to explain why wages are a declining share of national income, firms’ use of capital doesn’t just have to rise; it has to rise really fast. If wages should rise 10 percent, the use of capital would rise 12.5 percent, Neiman says. Others say that’s not the case: The use of capital would rise less than wages. If the others are right, labor should receive an increasing share of national income. But even assuming their underlying never-never-land model holds, economists can’t be sure what to predict.

Meanwhile, the political explanation has logged a striking new data point in its favor. It might astonish economists if they noticed, but in eight years Mexico’s Morena party has raised real minimum wages 153 percent—not ordering them by fiat but negotiating them with representatives of workers and business in the National Minimum Wages Commission (Conasami). As a share of national income, wages grew a monumental seven percentage points while profits sank eight points—a shift in favor of labor in eight years only a little smaller than the shift American bosses created in the opposite direction over half a century.

After saying you can’t fight technology, mainstream economics turns around and adds that, should a government be so rash as to engineer such a project, it would cause disaster. Tampering with he forces of technological determinism in conditions of economic volatility is to court massive unemployment, and to invite an accompanying exodus of workers from better jobs into really low-paid gigs.

Evidently, Mexico never got that memo—and strikingly it has suffered no such dire results. It’s true that the country’s official economic data is rather lenient about what counts as a job—if you’re selling boom-boxes at a street stand, you’re a full-time worker. But even with such irregularities, job numbers have ticked up since Morena took office, as unemployment declined from 3.3 to 2.6 percent of workers. Meanwhile, average real wages rose a stunning 44 percent, lifting 11 million workers above the poverty line.

It’s true that eroding the profit share past a sustainable threshold could discourage investment and dampen growth. That’s been the recent trend in Mexico; annualized GDP growth fell from a poor 1.6 percent annually under neoliberal administrations (2000 through 2018) to a worse 0.7 percent under Morena. There are lots of causes, including the destructive influence of Trump’s erratic tariff regime, but even if wage policies caused the slowdown, it seems a small price to pay for so much better income distribution.

Going forward, Morena would be wise to exercise caution. Sheinbaum, who has been called a “leftist with an Excel file” and holds a doctorate in environmental science from the University California, seems to understand this. Having brilliantly kept Trump at bay, she is implementing a serious “Plan Mexico,” which seeks, among other things, to curtail imports from China, and declining to raise business taxes though fighting tax evasion.

American progressives might take note. Medicare for All and state-financed childcare are hardly trivial policies, but they’re still patching up a deeply inequitable economy. A longer-term goal should be ratcheting the income distribution back toward the more equitable levels the country saw in the 1950s and ’60s. In adopting this mission, the example set in Mexico—together with a key lesson from the launch of the New Deal make the case that “technology” does not block this goal. Though the 1930s saw some of the most rapid technological advances in US history, Franklin D. Roosevelt raised wages from the starkly inequitable levels of the Gilded Age—and the economy began recovering from the Depression. The bulwark behind this achievement was the 1935 National Labor Relations Act (NLRB), which supported union organizing, and the establishment of a minimum wage.

These tools are still available. Gallup finds that 71 percent of Americans approve of unions today, just one point short of their popularity when FDR passed the NLRB. At the height of the labor movement in 1968, the minimum wage was $1.60 an hour, or $14.80 in current dollars. Today a $15 national minimum wage is as popular as unions, but that goal seems too small. In the years since 1968, the minimum wage sank to $7.25, but labor productivity—the dollar value of output the average worker produces—tripled. A Democratic trifecta could at least aim to accomplish something along the lines of what Morena has done.