Why Women’s Jobs Are Disproportionately Threatened by Automation
The AI debate has tended to center on men, but research shows that women make up 58 percent of the workers at the highest risk of job loss due to automation.
If the automation revolution is as bad as some researchers believe, almost half of all occupations in the United States are at risk of replacement by 2026. Truck drivers will be swapped out for self-driving AI. Manufacturers will use smarter machines instead of hands. Supermarkets will go cashier-free. Even more conservative projections acknowledge that some kind of transition is coming: The Bureau of Labor Statistics has projected that the overall number of jobs of the future will grow, but that 1.4 million current ones could soon become “redundant.”
As it rages, the automation debate has tended to center on men because they dominate in many of the low-wage occupations that are at a high (and high-profile) risk of replacement, like driving and factory work. But according to a report released Wednesday by the Institute for Women’s Policy Research—the first known comprehensive analysis of how automation will affect U.S. workers differently based on their gender—women could have even more at stake.
While women make up only half of the labor force, researchers found that they make up 58 percent of the workers at the highest risk of automation.
In the U.S. labor market, women are overrepresented in other high-risk occupations that involve routine data work easily disrupted by AI, like secretaries, administrative assistants, receptionists, and information clerks. They’re also overrepresented in jobs like child care, elder care, and education—ones that are comparatively “safe” from automation, but that may have lower salaries and few benefits.
While women make up only half of the labor force, researchers found that they make up 58 percent of the workers at the highest risk of automation. Among them, it’s Hispanic women who could be most affected, due to the professions where they’re most highly concentrated. Meanwhile, the share of women working as computer scientists and systems analysts, software developers, and computer support specialists has declined since 2000, leaving the task of shaping the proverbial Future of Work mostly to men.
Some of IWPR’s figures square with a World Economic Forum report released in January of 2018, which used Bureau of Labor Statistics data to find that women will hold 57 percent of the jobs likely to be disrupted by automation. But the IWPR went further, creating two new databases that together analyzed historical changes in employment by race and gender, and created future projections for occupations with the lowest and highest risks of automation and digitalization. To balance both conservative and dramatic estimates, the researchers used 2018 data from the BLS, probability of automation scores developed by Carl Frey and Michael Osborne in 2013, digitalization scores from a 2017 study out of the Brookings Institution’s Metropolitan Policy Program, and employment and earnings data from the American Community Survey to make their case.
“From a point of view of developing policies that reduce the threat of economic destruction of technological change and maximize the potential benefits, if you don’t look at gender or don’t have a gender-specific analysis you’re likely to come up with policy that doesn’t work very well,” says Ariane Hegewisch, who co-authored the report with Chandra Childers and Heidi Hartmann.
In conducting this analysis, IWPR wanted to “chart whether we are building on gains toward greater equality or cementing inequality for generations of workers to come,” said economist and IWPR president Hartmann in a press release. “What we found were warning signs.”
Examining trends in the U.S. labor market through this binary is valuable, the IWPR argues, because many occupations are already extremely gender-segregated: Of the top occupations for men, women only represent 25 percent of the workers; and vice versa.
Still, few past automation studies have used gender as a lens. Take Frey and Osborne’s famous study, which found that almost half (47 percent) of all jobs could be computerized, but did not disaggregate results specifically by gender. The IWPR found that, when broken down, the data showed job losses of 18.1 million for women in the 20 largest occupations for women; compared to 17.4 million in the 20 largest occupations for men.
When focusing only on the jobs with the highest risk of automation, the IWPR found that 20.2 million women work them, compared to only 14.4 million men. And of all women workers, 28.9 percent work in high-risk occupations, compared to only 19 percent of men.
“Women of each race and ethnicity are more likely to work in high-risk occupations than men of the same race or ethnicity,” the study found. But technological change will likely have more skewed effects on women of color. Hispanic women, for example, are more likely to work in manufacturing jobs like food and electronic processing, and transportation industries. Almost three quarters of the top 10 most common occupations for Hispanic women are high-risk—including the occupation category that earns Hispanic women the highest annual median salary, secretaries and administrative assistants. White men, meanwhile, are the least likely to work high-risk jobs.
Automation could also put a bigger dent in women’s paychecks than men’s. While the high-risk jobs men occupy are predominately lower wage—again, truckers and factory workers—the high-risk occupations that are dominated by women workers swing more along a spectrum of high and low compensation.
“It’s partly because women’s earnings distribution is more bunched up,” says Hegewisch, meaning women have fewer high-paid jobs in their most common occupations than men do—three of the top 20 occupations held by men pay a median of $80,000 a year, while none of women’s top 20 jobs do. But it’s also because many of women’s top jobs ripe for automation via sophisticated software or AI “have traditionally served as a bridge to the middle class”: they’re knowledge-based jobs like accountants, human resource managers, medical secretaries, bookkeepers, and paralegals.
“For men, you have good middle-skill jobs in construction, and as technicians,” Hegewisch says. Many of those jobs have median annual earnings of at least $40,000. They don’t require much digital literacy, and “they don’t go away—at least, they aren’t predicted to be impacted quite as strongly.”
Women who don’t gain digital skills don’t have as many options. Even those who do face a large earning gap, the study found. Though “[w]omen are more likely than men to work with computers and digital media,” which does enhance their earning potential, “the returns are significantly higher for men than for women.” For every $436 increase women get in annual earnings for doing more high-tech jobs, the study found, men get $740. That’s a gap of 41 percent.
There’s been some progress in the tech sphere: While the share of women in the three highest-paid tech jobs has gone down, the number of women of color has increased. Still, Hispanic women are 76 percent less likely to work in the digital space.
“I.T. jobs or STEM jobs are only 5 to 10 percent of the workforce,” Hegewisch says. “But they have a huge impact on how we design anything: from how self-driving cars are designed, to how we design new medical equipment, to how we design kitchens, whatever.” They’re the people developing AI to screen resumes—and adjusting the algorithms when they spit out biased results. And they’re the people that are doing the hiring. Only 22 percent of women working in AI were women as of 2018.
“If you don’t have diverse voices at the table or on the design team you come up with skewed results,” Hegewisch says. Of course, a certain degree of automation has already changed the nature of many women’s jobs. During the “IT revolution” that’s spanned the past few decades, it became easier to take certain manufacturing and low-wage labor jobs overseas. As a result, women were booted from textile and light manufacturing plants and electronics processing companies, and were replaced in call centers and as transcribers. “Women’s jobs from manufacturing in a way have been automated already,” Hegewisch says. “Men’s [manufacturing] jobs haven’t gone to the same extent yet.”
And technological change has already opened up other options even as it takes some away. The gig economy still represents a relatively small portion of the labor force, and it’s usually a precarious, benefit-free endeavor. But both men and women have turned to these gigs for supplemental income, and achieved relative parity in their likelihood to participate. “It opens new opportunities, and you can do work from home, and it can overcome bias because it’s a more universal sales platform,” Hegewisch says. “With Uber and Lyft it does make it possible—more possible—to work around childcare or caregiving or college responsibilities, and it does provide more flexibility particularly if it’s a supplementary earning for you.”
Like their roles in the “traditional” labor force, however, women and men tend to be segregated in the kinds of gig work they do: Men are still more likely to be ride-hail drivers, and women are more likely to participate in online entrepreneurship and platform-based domestic work. And for women, this creates another uneven workplace dynamic, where they’re again subject to pay disparities; higher penalties for cancellation (turning off the Uber app an hour early doesn’t have the same potential repercussions as canceling a promised house cleaning gig); and more frequent online harassment.
It’s not that policymakers and employers are blind to these realities: All the worry about automation has driven many to think about how to create a more resilient workforce, Hegewisch says. It’s just that much of the attention has been paid to men.
As “[w]omen fall further behind in terms of gender parity and upward economic mobility, and we design policies that try to buffer the fall for men and improve their ability to do jobs for the future,” Molly Kinder, a researcher at New America, told Axios. “But we don’t do the same for women.”
Many of the proposed solutions involve “life-long learning” programs, or workforce development initiatives. But women are also caregivers. “The problem is people don’t have the time … or the information,” or, crucially, the time to seek out the information, Hegewisch says. The report suggests helping carve out that time by expanding paid family leave policies; and bringing information to workers, by increasing access to on-the-job training.
Looking beyond any gender binaries, Hegewisch says, preparing for automation needs to include the voices of workers. Marriott hotel workers went on strike to ensure staff would have a seat at the table before new technology was implemented in their buildings. Las Vegas’ Culinary Workers Union members have done the same. “We need to make sure that innovation is people centered,” Hegewisch says. “That it’s not just top-down, and implemented as one way to cut labor as much as possible. [That will] make conditions worse for everybody in the process.”