The employment landscape in the United States is changing dramatically: the COVID-19 pandemic has redefined essential work and moved workers out of the office; new technologies are transforming the nature of many professions; globalization continues to move jobs to new locations; and climate change concerns are creating jobs in the alternative energy sector while reducing them in the fossil fuel industry.

Amid this workplace turmoil, workers, as well as employers and policymakers, could benefit from understanding the characteristics of employment that lead to higher wages and mobility, says fellow Sarah Bana postdoctoral at Stanford Digital Economics Labpart of the Stanford Institute for Human-Centered Artificial Intelligence. And, she notes, there is now a large data set that could help provide that understanding: the text of millions of online job postings.

“Online data gives us a tremendous opportunity to measure what matters,” she says.

Indeed, thanks to machine learning, Bana recently showed that the words used in a dataset of more than a million online job postings explain 87% of the variation in wages across much of the labor market. This is the first work to use such a large set of assignment data and examine the relationship between assignments and salaries.

Bana also experimented with injecting new text — adding a certificate of competency, for example — into relevant job postings to see how those words changed the salary forecast.

“It turns out that we can use the text of job postings to assess the salary characteristics of jobs in near real time,” Bana says. “This information could make job applications more transparent and improve our approach to workforce education and training.”

Words in job ads matter

To analyze the link between the text of online job postings and wages, Bana obtained more than one million pre-pandemic job postings from Greenwich.HR, which aggregates millions of job postings. employment from online job board platforms.

She then used BERT, one of the most advanced natural language processing (NLP) models available, to train an NLP model using the text of over 800,000 job postings and their associated salary data. When she tested the model using the remaining 200,000 job postings, it accurately predicted associated salaries 87% of the time. In comparison, using only job titles and geographic locations of job postings only yielded accurate predictions 69% of the time.

In follow-up work, Bana will attempt to characterize the contribution of various words to wage prediction. “Ideally, we’ll color the words in posts from red to green, where darker red words are linked to lower pay and darker green are linked to higher pay,” she says.

The Value of Skill Upgrading – A Text Injection Experiment

To identify skills important to predicting salaries, Bana used a text injection approach: to some relevant job postings, she added short sentences indicating that the job requires a particular career certification, such as those listed on Indeed.com. 10 In-Demand Career Certifications (and How to Get Them). Obtaining these certifications can be expensive, with prices ranging from around $225 to around $2,000. But, until now, there was no way to determine whether the investment was worth it from a salary perspective.

Bana’s experience found that some certifications (such as the IIBA Agile Analysis certification) produce significant salary gains quickly while others (such as Cisco Certified Internetwork Expert) do so more slowly – valuable insights for workers. who want better information about how an investment in skills training will affect their salaries and prospects, says Bana.

Employees aren’t the only ones who benefit from this information, notes Bana. Employers can use these findings to better invest in human capital, she says. If, for example, machine learning models reveal a gradual shift from certain tasks to others, employers would be notified in advance and could retrain some employees.

And policy makers considering promoting vocational training programs would similarly benefit from understanding which skills increase or decrease in economic value.

To that end, Bana and her colleagues are currently working on a companion document that identifies tasks that disappear from job listings over time and new tasks that appear.

In the future, Bana hopes that text-based analysis of job postings could produce a web-based application where workers or companies could seek added value by upgrading their skills or moving to a new geographic location.

“Right now there’s not a lot of clarity around a path to higher incomes,” Bana says. “Tools like these could help job seekers improve their job prospects, employers grow their workforce, and policymakers respond to immediate changes in the economy.”

Stanford HAI’s mission is to advance AI research, education, policy, and practice to improve the human condition. Learn more.

Previous

Maine Catholic Churches Continue to Attend Mass, Ecumenical Group Urges Services Online | National Catholic Register

Next

Report: Hawaii Voters Support State's Online Service Upgrades, But Not With Tax Hikes

Check Also