In the gig economy, reputation counts
An online labor market yields important insights into worker reliance on employer reputation
Published February 6, 2019
Economists have devoted substantial effort to the problem of finding good workers and managing them effectively. Their research generally takes the bosses’ perspective: How can a firm hire employees who match its needs and ensure that these workers produce to their full potential? But the converse—how can workers find companies that will treat them well and pay them as promised?—has received relatively little attention.
“Can Reputation Discipline the Gig Economy?” from the Federal Reserve Bank of Minneapolis’ Opportunity & Inclusive Growth Institute sheds welcome light on the question, drawing broad lessons from a unique perspective—that of workers engaged in online gig work.
The ingenious study, by Institute visiting scholar Aaron Sojourner with University of Minnesota colleagues Alan Benson and Akhmed Umyarov, looks at a widely used online platform that connects and arranges payment for workers and employers with simple, short-term tasks.
These jobs, from grading tests to answering surveys to entering data, are performed entirely online. They thus offer flexibility and relative autonomy to workers, but leave them vulnerable to abuse—in particular, nonpayment for work performed. A 2015 government report suggested that online clearinghouses for ad hoc jobs “can lead to violations of worker protection laws.”
Not coincidentally, a variety of “reputation systems” have blossomed around these platforms to give workers a means of rating their employers: Do they pay on time? Do they communicate clearly? Workers—who contribute their views voluntarily and anonymously—clearly value them. And employers themselves have argued that such informal reputation systems help weed out bad apples.
Are they right? Have online ratings of employers by their workers been influential on workers, wages, and those who hire? Sojourner, Benson, and Umyarov find that, at least for the example they analyze, the answer is a strong yes. Online reputation systems are quite effective. Employers with good reputations find it easier to attract workers, workers earn more when they’re informed about employer reputations, and lesser-known companies (because they’re small or new) with good reputations have the most to gain from reputation systems.
“Reputation systems may have an important role to play in providing employers with incentives to treat workers well,” note the economists, “giving lesser-known employers direct access to workers, and ultimately expanding the scope of work that can be completed online.”
It’s a promising development in worker-employer relationships, since the casual nature of contingent jobs is likely to remain lightly regulated. This is particularly true in “electronically mediated employment,” as the Department of Labor refers to it.
But the research suggests that online reputation systems may well have broader application. “Wage theft and other forms of opportunism are also pervasive in other settings where legal enforcement is weak, including among independent contractors, undocumented immigrants, misclassified employees, and low-wage employees,” the economists write. If reputation platforms grow and improve, they can help workers find reliable employers. “Then the falling costs of information processing and diffusion may move labor markets closer to the competitive ideal.”
The economists develop a job search model in which workers face the risk of employer opportunism and have incomplete information about prospective employers. Workers search for job offers from random employers and accept offers if they believe the prospective employer will actually pay them after task completion. Workers obviously have an incentive to work only for companies they think will pay them, and so companies have an incentive to pay reliably.
However, information about past treatment of workers doesn’t flow perfectly. Worker beliefs depend on what they can learn from either an informal reputation system or the company’s public visibility. The model allows the economists to explore the importance of employer reputation systems and to develop three hypotheses:
- For a given job offer, employers with better reputations will attract workers more quickly, enabling them to operate more quickly, on a larger scale, and at higher quality.
- Workers will earn more on average when working for employers with good reputations.
- The reputation system most benefits employers with good reputations but low visibility outside the reputation system.
The economists focus on one online hiring platform, Amazon’s Mechanical Turk (MTurk), and one reputation system, Turkopticon. MTurk allows individuals and firms to post simple, quick, often repetitive online jobs that humans can perform but machines can’t (yet), like adding descriptive keywords to photos or documents, answering surveys, or cleaning data. Individuals can take on a task if the pay looks reasonable. MTurk settles payment between worker and employer when work is submitted.
There is no contract or formal dispute-resolution mechanism in MTurk. Employers can renege on promised payments even if work is satisfactory. Doing so is obviously bad for the employee, whose only recourse is reporting that behavior in an online forum like Turkopticon, an informal rating system developed by MTurk workers.
Employer reputation—through either the online system or public visibility—is the only “enforcer” in this marketplace. To test their hypotheses about the value of information about employers’ treatment of workers in this labor market, the economists perform two experiments and take advantage of a third “natural” experiment.
The importance of being honest
The first experiment tests their theory about employers with good reputations. They create 36 MTurk employer identities, generate Turkopticon reputations for these employers (12 good, 12 bad, 12 none), and examine how quickly these employers can recruit real workers. As hypothesized, reputation matters. Employers with good reputations recruited workers about 50 percent more quickly than otherwise identical employers with no rating and 100 percent more quickly than those with very bad reps. (See figure.)
A second experiment tests the validity of online reputation: Do workers earn more from “good” employers? The economists assign randomly selected tasks from real MTurk employers of varying reputations to research assistants not informed of those reputations and then measure wages paid (or not) after tasks are completed. They find that “effective wages while working for good-reputation employers are 40 percent greater” per hour than when working for no- or bad-reputation employers, indicating faster task completion times and payment reliability. This demonstrates that an online reputation system like Turkopticon, though completely voluntary and unverified, can provide valuable information to workers. Such institutions help workers navigate away from opportunistic employers.
The third experiment takes advantage of Turkopticon misfortunes: When their servers crash (seven times during the study) and ratings aren’t available to workers, who suffers? The data show that good but little-known employers feel it most: “Workers sharply withdraw their labor supply from less-visible, good-reputation employers, who presumably were benefiting from Turkopticon informing workers” of their merit.
The study, say the economists, “is the first clean field evidence of the effects of employer reputation in any labor market” and, as it pertains to a form of labor still in infancy, serves as a benchmark for future research. Given the likelihood that informal worker relationships will remain a small but evolving slice of the employment world, employer reputation systems will bear continued scrutiny, particularly as “it seems implausible that [opportunistic employers] will ever be fully eliminated.”