It is very common today for consumers to use peer-to-peer reputational information. Online reviews are the latest incarnation of the old social customs of gossip and word-of-mouth, and they generate considerable excitement among policymakers and scholars. There is a growing sense that, with the advent of the sharing economy and mega reputational platforms such as Amazon,  we can increasingly do away with law and regulations in consumer markets. Why should the law set quality standards if consumers can simply avoid buying products that other consumers tell them that they are defective?

In a recent work-in-progress, titled Reputation Failure, (comments welcome!), I develop the idea that once we understand why people generate reputational information, we can see the limits and internal distortions of reputational information. Peer-to-peer reputational information is a public good par excellence; the people who produce it and the people who use it are not the same, and the producers get no financial reward from the users. The fact that there exists reputational information at all is far from obvious. In the paper, I review diverse literature that tries to track the motivations that lead people to share experiences with others. What emerges from the review is that these private motivations tend to bias and distort reputational information in systematic ways. As a result, reputational information is far less reliable than generally recognized. There are many latent dangers in relying on such information and consumers may suffer adverse consequences from using it.

The paper analyzes diverse evidence, but what I think is most interesting and familiar is the consistent finding that individuals share experiences when they are either upset or grateful. Venting feels good, doesn’t it? The literature calls it the ‘brag-and-moan’ model, although I prefer to call it the ‘spite and gratitude’ model. In any event, the consequence of this very familiar dynamic is quite harmful for the evolution of reputational information. Think what impression people would have of you if they only asked the opinions of former partners! More generally, it is useful to think of consumer experiences as a statistical sample fo the properties of the good or service. With enough datapoints, a prospective consumer can make informed predictions of her own experience from using the good, anticipating the likelihood of bad, bland, or good experiences. But when opinions cluster around the extremes–everything is 1 or 5 stars–then this selection bias comprises the statistical validity of inferences drawn from this sample.

Interestingly, the products on Amazon feature a J-shaped distribution which is consistent with the spite and gratitude model. Here is a nice graph prepared by Max Woolf at minimaxir.com:

Implications. On the first level, reputation failure adds to the list of standard market failures, like asymmetric information or externalities. So when such a failure is strong, we should pay more attention to legislation, regulation, standards and other market interventions. On the second and more ambitious level, we shouldn’t take these failures for granted. In fact, there are ways to mitigate these failures rather than dealing with their symptoms. For example, we could think of ways to compensate consumers for reviews in ways that will assure that the incentives have a content-neutral effect. We also need to pay more attention to how reputation platforms handle and disseminate reputational information. Perhaps most controversial, I believe it will be a good idea to shield consumers from lawsuits by disgruntled business owners. While I recognize the potential for bogus complaints, I am much more concerned with the chilling effects of these lawsuits. Whatever your take is on any individual measure, I think focusing on fixing reputation failure should be a high-priority for future legislators and policymakers who work in the area.