2ndOpinion.FYI中文Log in
genius.wiki

#10 1996 · eBay · E-commerce

eBay let strangers grade each other and the market policed itself

the problem

Wiring money to an anonymous seller

background

Pierre Omidyar's AuctionWeb, launched in September 1995, matched anonymous strangers to trade sight-unseen over the mail — the exact setup economist George Akerlof had described decades earlier as a "market for lemons," where buyers who can't verify quality assume the worst and honest sellers get driven out. Online, none of the usual signals of a trustworthy trader — a shopfront, years in the same location, a reputation in town — existed at all.

The conventional fixes were institutional: eBay itself could vet sellers, insure transactions, or require identity verification before anyone could list an item — the way a bank vets a borrower. Omidyar instead gave that job to the users themselves. When the site was about six months old, he launched the Feedback Forum, letting any buyer or seller leave a public, cumulative rating on the other party's trading identity.

what everyone would do

Have the platform itself vet sellers before they can list — identity checks, insurance, an approval process, the way a bank vets a borrower before lending. It fails at the scale and speed a mass anonymous marketplace needs: centrally verifying every stranger who wants to sell an item makes the platform the bottleneck on every listing, and the openness that let anyone start selling instantly is exactly what a vetting gate would remove.

what they saw

Omidyar saw that trust doesn't have to be guaranteed by an institution up front — it can be produced after the fact, by making a trading identity's history visible and permanent. Instead of the platform judging whether a seller could be trusted, he let every buyer who'd already dealt with that seller do the judging, publicly and cumulatively, which meant the market itself could price the risk without eBay ever having to certify anyone.

the move

The Feedback Forum: cumulative public ratings attached to every trading identity.

why it works

A single anonymous trade between strangers is a one-shot game with an obvious incentive to cheat, which is exactly the 'market for lemons' failure Akerlof described. Attaching a permanent, cumulative public score to a trading identity turns that one-shot game into a repeated one: a seller who cheats today taxes every transaction they'll ever make under that identity afterward, which makes honest dealing the seller's own rational self-interest rather than something the platform has to enforce. Because buyers can observe the score before trading, they can price the risk directly — the field experiment that later isolated the effect found buyers paid 8.1% more for an identical item from an established identity than a fresh one, reputation alone, nothing else changed.

the payoff

Made million-strong stranger markets self-policing; the template for every rating system since.

where it breaks

The mechanism depends on reputation being hard to fake or discard — a seller who can cheaply abandon a bad identity and start a new one resets the game back to one-shot, which is why platforms that let identities be recreated freely see the deterrent weaken. It also needs enough transaction volume for a score to mean anything; a brand-new seller has no history to be judged by, the cold-start problem every reputation system inherits and has to solve some other way. And it only works if buyers actually see and act on the score — review manipulation, rating fatigue, or an opaque scoring algorithm can leave a rating system nominally in place while quietly breaking the self-policing effect it exists to produce.

what came after

A later randomized field experiment measured the mechanism directly: a single high-reputation eBay dealer sold matched pairs of postcard lots under his established identity and under freshly created ones, and buyers paid 8.1% more for the identical item from the established identity — reputation alone, isolated from everything else. eBay's public-rating template became the reference model for every stranger-to-stranger platform that followed — Amazon reviews, Uber and Airbnb's two-way ratings, Etsy, Upwork — to the point that the mechanism itself is now unremarkable, even though solving the lemons problem for a mass anonymous market was not, in 1996.

references

  1. [1]The value of reputation on eBay: A controlled experimentExperimental Economics (via EconPapers/RePEc), 2006econpapers.repec.org
  2. [2]Pierre OmidyarEBSCO Research Starters, 2024ebsco.com
  3. [3]eBayWikipedia, 2024en.wikipedia.org

keep it

same kind of clever

Back to the archive