NO NEED TO APPLY

Algorithms are recruiting for companies desperate to hire a workforce that looks like their customers

Restless Bandit scours millions of resumes to help major brands find qualified candidates. In 2016, the human resources-technology company said almost every one of its customers indicated they wanted new employees to better reflect the racial diversity of their customers.

“I’ve been in this space for 20 years, and only in the last three years has it started to come up and only in the last year has it been operationalized,” says Restless Bandit CEO Steve Goodman.

This sudden demand for diversity is not altruism, says Goodman. Shareholder return and profit are driving these decisions rather than social pressure. Last spring, an executive at one global apparel brand, a Restless Bandit client, told Goodman’s team that “for us to be a successful company, employee diversity must demographically mirror our customer base.”

If so, most companies have a long way to go.

Restless Bandit sifted through their customers’ resumes, as well as job boards and professional social media partners, to compare demographics of white-collar workers (based on their resume skills) in six industries with the larger workforce using Bureau of Labor Statistics data. Out of 30 million resumes scanned, they were able to identify ethnicity for about 19 million workers using academic models assessing ethnicity by first and last names.

The gaps they found between how many black and Hispanic workers are out there and how many hold white-collar jobs was huge across the board. Blacks, despite comprising 12.3% of the total workforce, held about 1.75% of white-collar positions. Hispanics, representing 16.8% of the total workforce, were only 5% of the professional positions workforce.

Few industries have expended more money and words on diversity with less to show for it than tech. Google spent more than $100 million in 2014 and 2015 on diversity initiatives but its numbers have hardly budged: In 2014, Hispanic and black workers comprised 3% or less of the workforce each; in 2016, the numbers were exactly the same (pdf). Of the six industries in Restless Bandits’ study, technology companies had the lowest relative number of black and Hispanic white-collar employees.

Critics say insular social networks and a hiring bias towards candidates with elite schools and brand-name companies on their resumes has hampered the diversification process, and the sheer size of some big tech companies—Google’s parent company Alphabet employs 62,000 people—means substantial changes will take years, if not decades.

This gap is set to grow as the US diversifies. To prepare, US companies are now asking firms like Restless Bandit to use algorithms that increase the probability new hires will reflect the diversity of their customers. By starting now, companies hope to head off an even greater gap in the future when the US is expected to have no single racial or ethnic majority by 2043.

Identifying these candidates remains difficult. Resumes are inadequate to the task (since they lack explicit ethnic or gender identifiers, although inferences can be made, as in this study), and it is illegal to make race an explicit condition of employment under Title VII of the US Civil Rights Act. But companies may follow the path of universities’ affirmative actions programs that consider a constellation of traits, some associated with race or ethnicity, to overcome racial biases. Companies applying machine learning to the task found they increased hiring of blacks and Hispanics by 26% looking for high performers using behavioral survey data, even without racial identifiers.

But companies know they’re venturing into risky territory as ethnicity becomes an explicit consideration for new hires. That’s complicating sales at Restless Bandit. “It’s hard to sell software when you’re having a political conversion inside the company,” said Goodman.

home our picks popular latest obsessions search