Measuring Employment Demand Using Internet Search Data

2018 ACM Conference on Human Factors in Computing Systems (CHI) |

Published by ACM

We are in a transitional economic period emphasizing automation of physical jobs and the shift towards intellectual labor. How can we measure and understand human behaviors of job search, and how communities are adapting to these changes? We use internet search data to estimate employment demand in the United States. Starting with 225 million raw job search queries in 2015 and 2016 from a popular search engine, we classify queries into one of 15 fields of employment with accuracy and F-1 of 97\%, and use the resulting query volumes to estimate per-sector employment demand. We validate against Bureau of Labor Statistics measures, and then demonstrate benefits for communities, showing significant differences in the types of jobs searched for across socio-economic dimensions like poverty and education level. We discuss implications for macroeconomic measurement, as well as how community leaders, policy makers, and the field of HCI benefit from this information.