Credence Goods Markets and the Informational Value of New Media: A Natural Field Experiment

Abstract

Credence goods markets are characterized by pronounced informational asymmetries between consumers and expert sellers. As a consequence, consumers are often exploited and market efficiency is threatened. However, in the digital age, it has become easy and cheap for consumers to self-diagnose their needs using specialized webpages or to access other consumers' reviews on social media platforms in search for trustworthy sellers. We present a natural field experiment that examines the causal effect of information acquisition from new media on the level of sellers' price charges for computer repairs. We find that even a correct self-diagnosis of a consumer about the appropriate repair does not reduce prices, and that an incorrect diagnosis more than doubles them. Internet ratings of repair shops are a good predictor of prices. However, the predictive valued of reviews depends on whether they are judged as reliable or not. For reviews recommended by the platform Yelp we find that good ratings are associated with lower prices and bad ratings with higher prices, while non-recommended reviews have a clearly misleading effect, because non-recommended positive ratings increase the price.

Reveal It or Conceal It: On the Value of Second Opinions in a Low-Entry-Barriers Credence Goods Market

“…The remaining four shops (12.12%) recommended to the undercover shopper to visit another shop for a second opinion. A successful repair rate of only 75% is very low in comparison to other computer repair studies (Kerschbamer et al, 2016(Kerschbamer et al, , 2019 with rates above 95%, showing that our manipulation required an expert's effort to solve the problem.…”

“…This is much harder to detect as it is not visible when the computer is opened and the error message is ambiguous. As a consequence of these subtle differences in the manipulation, there are large differences in the results: While the frequency of successful repairs is about 95% in Kerschbamer et al (2016) and about 98% in Kerschbamer et al (2019), it is only about 75% in the present study. Of course, our paper is also completely different since Kerschbamer et al (2016Kerschbamer et al ( , 2019 do not study consider second opinions.…”

“…On the surface it might appear that the experimental designs in Kerschbamer et al (2016) and Kerschbamer et al (2019) are very similar to the one employed here: All these papers work with a manipulation that involves the RAM modules. However, there are subtle differences in the manipulation: In Kerschbamer et al (2016) only one of the two RAM modules is destroyed and this leads to an easily interpretable error message that appears on the screen.…”

“…Turning to field experiments in credence goods markets, most of the existing work is intentionally designed in such a way that it is easy and cheap for experts to diagnose the problem correctly, without the need to invest noteworthy effort. 5 This is done because the main focus of large parts of previous field work is on intentional misbehavior on the provision and charging stage -as, e.g., in Balafoutas et al (2013), Kerschbamer et al (2016), Balafoutas et al 2017, Kerschbamer et al (2019), or Gottschalk et al (2020). While it is important to understand the drivers of intentional mistreatment and mischarging, and ways to decrease the probability to get defrauded in such markets, the question how to induce experts to perform a serious diagnosis is obviously important as well.…”

“…However, there are subtle differences in the manipulation: In Kerschbamer et al (2016) only one of the two RAM modules is destroyed and this leads to an easily interpretable error message that appears on the screen. In Kerschbamer et al (2019) both RAM modules are loosened, which makes the diagnosis easy as the loosened RAM modules catch the eye of the expert immediately when she opens the computer. In the present study both RAM modules are destroyed.…”