In April 2017, we launched the Economic Graph Research Program (EGRP) to encourage researchers, academics, and data-driven thinkers to solve some of the world’s most challenging economic problems. Today, I’m happy to share the latest details on the evolution of the EGRP as well as some of the interesting findings based on the work of the research teams.
Unique data, compelling findings
LinkedIn has a unique dataset. Our highly-structured Economic Graph maps members, skills, jobs, and companies to provide insights into the functioning of the global economy. We work with governments and organizations like the World Economic Forum and the World Bank Group to help identify macroeconomic trends and opportunities to benefit populations around the world. The outcome of our work includes compelling milestones that range from WEF’s Global Gender Gap Report to our monthly LinkedIn Workforce Report.
The Economic Graph Research Program is an initiative where we work with select members of the global research community to investigate fundamental questions about economic theory, data science, and related fields, using LinkedIn’s Economic Graph data. We believe that together, we can effectively remove barriers to economic growth and create opportunity for every member of the global workforce by surfacing new understandings of complex labor market dynamics.
Big economic repercussions for non-compete agreements
This program grew out of the original Economic Graph Challenge (EGC), a Grand-Challenge-style initiative for a small, select group of researchers that would help encourage long-term public-private partnerships on compelling research topics.
One interesting output from EGC was research by Professor Jessica Jeffers at the University of Chicago. She examined costs and benefits of restricting worker mobility using noncompete agreements. Such clauses can be a double-edged sword: while they potentially limit workers’ ability to seek out better employee-job matches in the short term, they may also encourage capital investment at existing firms by helping firms retain talent over time. Using data on employment history and court decisions on the enforceability of non-competes, she concluded that both of these effects indeed occurred. Her work found that non-competes lead to a 18% decrease in the entry of new firms (in knowledge-intensive sectors such as technology and professional services) and that departures to entrepreneurship also declined by an average of 8.4%. She also found that noncompetes lead to greater internal investments by large, existing companies. This sheds important light on the economic implications of noncompete agreements that, in the U.S., can have ramifications for federal and state economic policy.
Relationships between human capital investment, skills, and firm success
Another EGRP team is looking at the longstanding debate about the extent to which employers can benefit from the human capital investments of their employees. MIT researchers Erik Brynjolfsson and Daniel Rock and Wharton School professor Prasanna Tambe (along with Jacqueline Barrett from the Economic Graph team) are using the educational data provided on LinkedIn by tens of millions of U.S. employees to evaluate the relationship between firm market value and educational investment. This work explores fundamental assumptions made by economists about the global labor market.
Prior to the EGRP, Professor Tambe used LinkedIn skills data to identify the importance of geography and access to talent pools to explain differences in productivity growth rates across labor markets when new technologies (in this case, Hadoop expertise) are in the early phases of adoption. (He also wrote about what it’s like to work with corporate data as an academic researcher here).
Demystifying gender bias on professional networks
Other research has had ramifications for a range of product decisions at LinkedIn, providing insights into issues like gender bias or gendered self-handicapping in response to certain job ads.
One team of EGC researchers analyzed whether male and female graduates of top-ten MBA programs from 2011-2016 promoted themselves equally on their LinkedIn profiles. The team discovered that women and men were comparable in the number of skills and honors they included on their profiles, but that women were less likely than men to include job descriptions or summaries. We don’t know why women in top MBA programs are less inclined to include these more descriptive parts of their LinkedIn profiles, but this could be a fertile avenue for future research. It also provides us with useful insight to better understand what not to use as signals in, for instance, job recommendation models.
Dr. Laura Gee from Tufts University recently published research based on a 2012 internal A/B test that was done on the construction of job listings on the LinkedIn platform. The research concluded that for job listings that were viewed by more male, rather than female applicants, “female candidates were especially more likely to apply for positions” when they had additional information about the current number of other applicants. Work like Dr. Gee’s informs how we display jobs for those that we recommend to our members and shows how job boards can overcome societal bias or expectations in job application behavior.
This kind of research not only shows how our data can answer questions about the world but also allows us to take advantage of outside expertise to gain more insights into how LinkedIn functions.
Coming soon: Call for proposals
I’m excited to announce that we will be accepting a new round of Economic Graph Research proposals next month. Starting August 20th, interested researchers from universities, think tanks, and other non-profit entities can submit their proposals for Economic Graph Research projects to the team at LinkedIn using the downloadable proposal template.
At LinkedIn, we have a strict policy of “members first” when it comes to data privacy and security. Accordingly, researchers’ proposals will be vetted by LinkedIn’s legal and security teams for privacy and security concerns before the proposals are peer-reviewed by experts in fields including, but not limited to, AI, analytics, and economics. As proposals are evaluated, we look at a few key factors, such as the novelty of the research idea, the data availability and accessibility, the team’s commitment, and the proposed duration of the research. When a researcher is selected, either their organization (generally a university) or, in some cases, the individual themselves, must enter into an agreement with LinkedIn. EGR teams are limited to a total of five team members.
Full details about the submission process, selection criteria, and how we protect member privacy can all be found on the new Economic Graph Research website, and you can watch this space for more details as well. Please consider joining us in our endeavor to answer fundamental questions about data science and the world economy.