Penn economists investigate how reputation rules on the ‘dark net’
In online retail markets, a seller’s rating plays a significant role in his or her business, indicating to buyers whether or not the seller is an honest merchant.
“There’s a long-running interest in economics in understanding how reputation works to enforce behavior and cooperation,” says Nick Janetos, a recent graduate of the Department of Economics in Penn’s School of Arts & Sciences. “When you go on Amazon and you buy from a third-party seller, there’s this rating that you can look at that tells you something about the quality of that seller. And there’s some research that’s starting to look at how that works.”
Janetos and Jan Tilly recently completed a study while doctoral candidates at Penn, investigating how these rating systems work on the “dark net,” an online market for illegal items such as drugs and firearms that operates free from government regulation and law enforcement.
Unlike Amazon and eBay, where defrauded buyers can go to the police, small claims court, or report the seller to the site, customers on the dark net have little recourse when deals go sour. This makes the dark net an ideal place for researchers to investigate how ratings systems work, since it’s a clean environment without extraneous factors.
“These markets exist outside the court system, so it seems like reputation really should be a thing that keeps them going,” Janetos says. “What we’ve done is we’ve collected a rich data set of sellers, behavior, prices charged, and who’s buying from them, and we formally investigated it using a new statistical model. We found that, in fact, reputation is extremely important: If it weren’t for this information that the market platform is providing in terms of ratings, they would completely collapse. They wouldn’t even exist.”
Janetos and Tilly wrote a computer program that visits these websites periodically and takes snapshots, then later goes through and tries to pull out everything from the snapshots it took and form it into complete histories for particular sellers.
They then came up with a new statistical model to analyze the data sets.
“There’s a unique problem when you think about statistically analyzing this environment, which is that it matters very much what beliefs people have about what ratings mean,” Janetos says. “In order to say something about how these ratings work, we need to be able to statistically estimate the beliefs that people have when they come to the market and they see a rating of 5 or 4. What we did is we wrote down an economic model of rational buyers coming to this market to draw sophisticated statistical inferences based on seller and buyer behavior from these ratings.”
Although what Janetos finds most exciting about this research is learning how reputation works on online markets, he says figuring out what keeps the dark net afloat could help law enforcement figure out how to dismantle it.
“If you’re trying to think about how you can shut these markets down, it’s very important to understand what it is that’s keeping them running,” he says. “You can imagine suggesting that a way to weaken these markets and reduce traffic on them is to come in and try to mess up the ratings system somehow by doing things like making fake listings or leaving good reviews for bad sellers.”
Janetos says that for this research, they only analyzed a very small part of a much larger data set so there’s a lot of interesting work that remains to be done.
“There’s lots of scope here for continuing to look at various aspects of how these markets work,” he says. “The sensible next step is to start looking at market platforms like Amazon and eBay and trying to apply what we’ve learned here to talk about reputation.”