In a research note published Monday, Goldman Sachs economists Pierfrancesco Mei and Jessica Rindels drew on four decades of individual-level data to assess what they call the “scarring” effects of technological displacement on U.S. workers. Their verdict is sobering. Workers whose jobs are eliminated by technology don’t just struggle in the short term—they can spend the better part of a decade fighting to recover.

“Over the 10 years following a job loss, real earnings for technology-displaced workers grow nearly 10 percentage points less than for never-displaced workers,” the report found, “and 5 percentage points less than for other displaced workers.”

The research team tracked more than 20,000 individuals across two cohorts—one born in the 1950s and ’60s, and another in the 1980s—using the National Longitudinal Surveys sponsored by the Bureau of Labor Statistics. By identifying which occupations faced the steepest technology-driven employment declines in each decade since 1980, they were able to map the full career arcs of workers caught in automation’s path.

The immediate pain is real

The short-run picture is rough. Workers displaced from technology-disrupted occupations take approximately one month longer to find a new job and suffer real earnings losses more than 3% larger upon reemployment compared with workers let go from more stable fields. The core culprit, Goldman found, is occupational downgrading: Displaced workers tend to slide into roles that are more routine and require fewer analytical and interpersonal skills, not less, because the same technological forces that eliminated their old jobs also eroded the market value of their existing skills.

The scarring doesn’t stop at paychecks. Goldman found that workers displaced early in their careers—between ages 25 and 35—accumulate less wealth over time, largely because they delay buying homes. They’re also less likely to be married at any given age compared with never-displaced peers, suggesting the economic shock ripples into their personal lives as well.

Recessions make everything worse

Goldman’s most urgent warning may be about timing. Firms disproportionately shed routine jobs during economic downturns, when efficiency pressure peaks. For workers, a recession-era technology displacement widens the already painful gap versus other displaced workers by roughly three additional weeks of unemployment and five percentage points each for the risk of returning to unemployment and exiting the labor force entirely. With AI adoption accelerating at a moment of unusual macroeconomic uncertainty, that compounding risk is hard to ignore.

The Gen Z twist

Here’s where the report defies the prevailing narrative. Much of the public anxiety about AI-driven job losses has centered on young workers—particularly new graduates entering a market increasingly shaped by automation. Goldman’s data tells a different story. Younger, college-educated, and urban workers experience cumulative earnings losses roughly half as large as other technology-displaced workers over the decade following a job loss. Their advantage comes from flexibility: They switch occupations more readily and migrate up the skills ladder into roles with higher analytical content that complement, rather than compete with, new technology.

“Contrary to current concerns that the costs of AI will fall especially hard on new graduates,” the report states, “younger workers have actually been able to adjust more flexibly through occupational mobility and skill upgrading in the past.”

Retraining also helps cushion the blow. Workers who participated in vocational or technical programs within three years of displacement saw roughly two percentage points more cumulative wage growth over the following decade and a 10-percentage-point lower probability of returning to unemployment.

Goldman has been estimating for several years that AI could displace 6% to 7% of U.S. workers over the next decade. This 40-year sweep of data suggests the workers who should be most worried aren’t the youngest ones in the room—they’re the older, less mobile workers with deeply occupation-specific skills and no recession-proof timing on their side.

For this story, Fortune journalists used generative AI as a research tool. An editor verified the accuracy of the information before publishing.