I agree, but it begs the question: is there a point at which the damage you can inflict with a single bad transaction makes some central authority necessary? e.g. if every 20th car I sell on eBay is a lemon, I still have a 95% rating. If I'm a driver on Uber/Lyft/Sidecar and every 20th person I leave in a ditch I still have 4.75 stars.
Obviously regulation isn't meant to stop outright crime like this, but I think it's meant to guarantee a minimum level of service so that the distribution of outcomes has a reasonable lower bound.
I think it can be. Obviously you wouldn't use a system like this to decide who gets to own nuclear weapons.
There are a couple of factors: cost per "incident" and volume of transactions. P2P trust systems work where there is a large volume, the cost-per-incident is relatively low, and, most importantly, the provider looks forward to continued transactions. To me cabs seem to easily fall under this rubric.
You know, it's possible to get a medical license and be so inept as to kill people for decades until you get caught. It's possible to get a hoteliers license and spy on your customers and cheat them out of services -- happens all the time. So we have to be clear that we are not comparing absolutes. There will always be problems with either solution. The question, to me, is "which system is more adaptive?" Some systems adapt quickly as conditions change. Most legal systems do not.
I'd still want legal recourse for bad actors in many situations. But things like taxicab drivers robbing you? We already have legal recourse, yes? So in reality with the taxi situation we're only talking about the cab fare, not a FUD list of possible really bad things (tm) that could happen in a taxi. Geesh. That'd be a huge list.
Obviously regulation isn't meant to stop outright crime like this, but I think it's meant to guarantee a minimum level of service so that the distribution of outcomes has a reasonable lower bound.