Microsoft Research Interactive Science
All
Education
Population
Income
of all job searches in
bottom counties
of all job searches in
top counties
Statistical difference (t = ) is a effect size (d = )

Measuring Employment Demand with Search Data

Scott Counts, Justin Cranshaw, Stevie Chancellor

The United States, along with much of the world, is in the midst of an economic transition from manual to intellectual labor. The changing nature of work, including the automation of labor, is an important issue facing society, with implications not only for our standard of living, but also for socio-economic and political divisions.

With this backdrop in mind, at Microsoft Research we hope to augment the study of employment in the U.S. Traditionally, agencies like the Census Bureau and the Bureau of Labor Statistics use a variety of measurement techniques to monitor the number of new jobs, either positive or negative, across an exhaustive taxonomy of employment. These methods are accurate, but often lag actual trends, as they are predominately based on surveys done after the fact.

Given that a majority of people use internet search as part of their job search, in work presented at the CHI 2018 conference we explored whether we could use Bing searches, anonymized and aggregated, to augment these traditional approaches to surveying the shifting employment landscape. The use of internet search data provides several advantages, including the ability to update employment measures in near real-time, to measure at significant scale, both in terms of population size and geographies covered, and finally to understand intent to change jobs that may or may not be realized.

For the CHI research paper, we started by classifying almost a quarter of a billion job search queries from 2015-2016 into employment sectors, such as retail, transportation, technology, education, healthcare and so on. To do so, we first created very carefully curated dictionaries of job titles to roughly categorize searches that contain those job titles (for example, “nursing jobs in Seattle”) into one of the employment sectors (healthcare). We then used Word2Vec to find new job searches not in our dictionaries (for example, “rn jobs in Seattle”). The overall performance of the classifier was quite high, and some representative search query keywords can be found in the table below.

Employment Sector Example Job Search Query Keywords
Architecture / Engineering entry level biomedical engineering jobs, auto-cad jobs
Art voice acting careers, museum curator job
Business vp of operations jobs, hr career
Construction construction laborer jobs, welder jobs
Education community college professor jobs, school district careers
Finance financial banking jobs, mortgage lender careers
Food bartender jobs, craigslist dishwasher jobs
Healthcare surgical tech jobs, clinic jobs rn
Leisure casino jobs, laundry jobs in hotel
Manufacturing machine operator jobs, jobs in shopfitting
Retail clothing store job applications online, retail career outlets
Science psychology research associate jobs, jobs in r\&d
Technology software architect career, sql dba jobs
Transportation airport runway jobs, cdl jobs

Exploring the Data

Building on the CHI paper, and using similar methods, this site provides an interactive look at job searches across the U.S. The map above (click to reset map) shows summaries of job searches from all U.S. counties in 2017 in the Engineering job sector. The intensity of the color indicates the percent of job queries from the county relative to searches for jobs in other sectors. Scanning the list of job sectors at the top right shows that healthcare and education are the two most common sectors for job searching, followed by business, transportation, and finance.

Selecting an individual county will pull up its job sector search profile. For instance, selecting New York County (click to navigate map) in New York shows that Manhattan’s county is similar to the national average in terms of what jobs are searched for, but lower on education jobs, particularly with respect to finance job (12% compared to 8% nationally) searches, and is substantially higher than the national average in searches for business jobs (14% compared to 8% nationally). Nearby Bronx County is more dominated by job searches for healthcare (31% compared to 28% nationally) with relatively fewer searches for architecture/engineering jobs (1% compared to 2% nationally).

High population counties like Seattle’s King county tend to have a distribution of searches over job sectors that are closer to that of the national average. Lower population counties are more likely to skew toward an individual job sector. Looking at job searches for manufacturing jobs, for instance, we see a broad trend toward the more of such searches in the Midwest and rustbelt, but also a few standout counties with smaller populations, such as Menominee County, Michigan (21% compared to 4% nationally) and Allamakee County, IA (20% compared to 4% nationally).

We see similar patterns in other employment sectors. In the hospitality sector, for instance, we see high numbers of job searches in Summit County, CO (21% compared to 4% nationally) and Douglas County, NV (14% comapred to 4% nationally), both homes to major ski resorts.

How Do Job Searches Relate to Demographics?

Importantly, we wanted to understand how job searches in the different employment sectors relate to various demographics.

What is the role of income and education level in job searching?

Jobs in sectors like business and finance and technology often pay more than retail or manufacturing jobs, for example. If we split counties into quartiles (top 25% and bottom 25% of counties) based on percent of population below the poverty line, we first notice that the majority of higher income counties (shown in green) contain urban centers. We also see that people in higher income counties do in fact search for significantly more finance jobs. However, in terms of what people search in the top and bottom counties in terms of income, while most of the differences were in the expected directions in wealthy counties, we do not see more searches for business, technology, and science jobs compared to poorer counties.

We do see more job sector differences are in counties with higher versus lower levels of education, with people in higher education counties significantly more likely to search for business and finance jobs, and those in lower education counties much more likely to search for construction, manufacturing, and transportation jobs. Transportation jobs searches from counties lower in educational attainment, for instance, constitute about 1.5% more of all job searches than do transportation job searches from counties higher in educational attainment.

Is there an urban advantage for job prospects?

Broadly speaking, high population counties correspond to urban environments that may offer more options for employment, particularly in “new economy” areas such as the technology sector. Our findings show significantly more searches for business, but we do not see significantly more searches for technology jobs in counties in the top quartile of population.

Somewhat surprisingly, we observe an urban advantage to searches in the food sector. This also corresponds with simillar trends across both income and education demographic cuts, perahps reflecting a rise in food culture in urban environments. However, given higher percent of searches for high-paying job areas such as business, this could also be a marker of rising inequality in cities.

More populated areas likely offer a wider variety of job opportunities. Indeed in the CHI 2018 paper we show that increases in population correlate with increases in the diversity of job types searched for. For example, looking at highly populated Kings County, NY (Brooklyn), we see a profile of job searches that includes a healthy percentage of searches from all employment sectors. On the other hand, smaller population counties often skew toward a particular industry. For example, in Rooks County, KS, nearly a third of all searches were for transportation jobs.

Summary

In summary, we utilized jobs search queries on Bing to explore demand for employment at county scale in the United States. We show here that in 2017 people in higher education counties tended to search more for higher paying (finance) jobs. While those in lower income counties did search more for traditionally lower paying jobs, such as in transportation. Encouragingly we did see equivalent levels of searches for “new economy” jobs (technology) in those same lower income counties.

Please refer to our CHI 2018 paper for additional analyses, including a comparison of the jobs people are currently working according to BLS versus the jobs people are searching for, broken down by low and high poverty areas.

If you have comments, questions, or ideas about these data, contact us at interactivescience@microsoft.com.

Education
Population
Income
percentile
percentile
percentile
% w/ college
% in poverty
Job Searches by Sector