Uber, Airbnb, TaskRabbit, and other online platforms have drastically reduced the price of micro-contracting and grown a “gig” economy, where workers must more frequently decide which potential employers to trust. Traditionally, workers have used labor unions and professional associations as a venue for exchanging information about working conditions and coordinating collective withdrawal of trade in order to discipline employers.

The rise of new institutions that facilitate information sharing may take up some of this role, a recent IZA discussion paper by Alan Benson, Aaron Sojourner and Akhmed Umyarov (all University of Minnesota) suggests. In two experiments in an online labor market (Amazon’s Mechanical Turk or M-Turk), the authors show that a public, employer-reputation system has value for:

  1. workers who use it to screen potential employers on otherwise unobservable differences, and
  2. employers who can benefit when a better reputation makes it easier to attract more workers of any given quality, basically shifting out the labor supply curve they face.

The experiments provide the first estimates of the value of employer reputation measured outside the lab based on any design more credible than a control function and highlight the under-appreciated struggles workers face in navigating the labor market.

The first experiment tests the validity of the reputations from the perspective of a worker. The researchers act as a worker to assess whether other workers’ public ratings reflect real variation in employer quality. One research assistant (RA) randomly selects tasks from employers who have good reputations, bad reputations, or no reputation and sends them to a second RA who is blind to employers’ reputations. Benson, Sojourner and Umyarov  find that, holding work effort fixed, effective wages while working for good-reputation employers is 40 percent greater than effective wages while working for bad-reputation employers.

Employers with bad reputation must post 200% higher wages

The second experiment measures how employers’ reputations affects their ability to recruit workers. We create multiple employers on M-Turk with varying reputations and, holding all else fixed, measure differences in the rate at which they attract workers to posted jobs. Good-reputation employers attract workers about 50 percent more quickly than our otherwise-identical no-reputation employers and 100 percent more quickly than bad-reputation employers. Average quality of work does not differ. This implies employers with better reputations can operate at a faster pace, a larger scale, or be more selective in hiring. Existing estimates of M-Turk wage elasticities imply that posted wages would need to be almost 200 percent greater for bad-reputation employers to attract workers at the same rate as good-reputation employers do.

Most models of the labor market assume workers have perfect information about employer differences and gloss over the difficulty workers face in navigating these matters. The authors of the IZA Discussion Paper propose a simple, equilibrium-search model consistent with their results. Informed-type workers screen employers with bad reputations, and the threat of losing a good reputation and thus losing informed workers discourages employers from engaging in wage theft and other forms of opportunism. In this way, employers’ reputation serves as collateral against wage theft, effectively substituting for the role that formal contracts normally play in the labor market.

Overcoming the information problem

What relevance does this have for labor markets broadly? All workers strive to distinguish which employers will treat them well or ill. Two prospective employers that offer identical employment contracts may actually differ widely in the criteria they apply for raises, promotions, terminations, scheduling, bonuses, task assignment, and many other working and payment conditions. In contingent, undocumented, and low-wage labor markets, concerns are as basic as whether employers will pay for all hours worked or pay at all.

Workers have always made decisions with partial information about employer quality and, so, these forces have always shaped labor markets. Falling communications costs have made it easier for workers to share information and several websites now enable this exchange, including Glassdoor, Turkopticon, Contratados, Kununu, JobeeHive, TheJobCrowd, and the Freelancers Union’s Client Scorecard. Attention to the worker’s information problem suggests innovative directions for policy and institution-building.

Download the complete paper (IZA DP No. 9510):