Applying for a job can feel like shouting into an abyss. And for good reason: Unless you know someone in the organization, you’re at a major statistical disadvantage. While this can feel unfair, new research shows employers are justified in this bias.
With referrals, there’s a ready-made reference and employers tend to believe they’re better workers, because current employees should know who’s likely to be a good fit. As a result, they get faster reviews and responses, and vastly better hire rates.
Confirming whether they’re actually better workers is difficult, because a non-referral who manages to get hired is very much an exception. That makes them a bad proxy for an average blind applicant to compare against referred candidates.
But now there’s even more bad news for the networking-averse. In anew NBER working paper (pdf)—amazingly titled “Why the Referential Treatment?”—researchers from Harvard and Coursera used a clever experiment to conclude referrals really do turn out to be better employees in a number of significant ways.
Referred workers cost less to recruit, performed better, and had lower turnover across three experiments conducted on the online freelancing platform oDesk (now Upwork), which allowed the authors to get around the problems with comparing hired employees.
The researchers hired freelancers for a series of week-long jobs, with tasks common to the platform.
The study found that the performance gap grew over time on the job. In addition, referrals from high performing people and from people particularly close to the recommended employee performed even better.
The first experiment tested whether peer pressure was a factor. That is, do referred employees work extra hard to make the person who referred them look good? It turns out that’s not really the case.
In the “observed” group, people who made recommendations got information about their referral’s performance, and were told their own evaluation depended partly on how their referrals did, giving them a strong incentive to check up. The second group of referrals was told their performance data would go only to managers. Both were compared to non-referred hires doing the same tasks.
Both groups of referred workers outperformed the non-referred group. And although the “observed” group did slightly better than the other referrals, the difference wasn’t statistically significant.
A second experiment tried to find if a referral revealed something about a candidate that’s not visible on a resume, like diligence or cultural fit. Four months after the initial job experiment, the referred candidates got a new job offer from a different fake company for another task. Referrers were not used this time.
The candidates referred in the initial experiment once again performed better than the non-referrals. Some of the outperformance was predictable based on resume differences, but the vast majority couldn’t be so easily explained.
Referrals did even better when paired up with the person who recommended them on a third job, which required teamwork.
The experiment isn’t perfect, but it does help overcome some of the shortcomings of only examining hired workers. And it does appear to confirm the assumption among companies that referred candidates have some special unseen quality.
Freelancing, of course, is different from full time employment. So, it’s not clear that these effects would last.
And the study doesn’t address one of the biggest issues with this kind of hiring behavior: People tend to refer others like them. The result, particularly in rapidly growing tech companies, can be a shockingly homogenous workforce.