Just as employers face uncertainty when hiring workers, workers also face uncertainty when accepting employment, and bad employers may opportunistically depart from expectations, norms, and laws. However, prior research in economics and information sciences has focused sharply on the employer’s problem of identifying good workers rather than vice versa. This issue is especially pronounced in markets for gig work, including online labor markets, where platforms are developing strategies to help workers identify good employers. We build a theoretical model for the value of such reputation systems and test its predictions in on Amazon Mechanical Turk, where employers may decline to pay workers while keeping their work product and workers protect themselves using third-party reputation systems, such as Turkopticon. We find that: (1) in an experiment on worker arrival, a good reputation allows employers to operate more quickly and on a larger scale without loss to quality; (2) in an experimental audit of employers, working for good-reputation employers pays 40 percent higher effective wages due to faster completion times and lower likelihoods of rejection; and (3) exploiting reputation system crashes, the reputation system is particularly important to small, good-reputation employers, which rely on the reputation system to compete for workers against more established employers. This is the first clean field evidence of the effects of employer reputation in any labor market and is suggestive of the special role that reputation-diffusing technologies can play in promoting gig work, where conventional labor and contract laws are weak.