It's a model. A pretty good one, since the mechanism at play here can occur in a very general sense. The car market is an example where this effect may occur, a

It's a model. A pretty good one, since the mechanism at play here can occur in a very general sense. The car market is an example where this effect may occur, as is insurance and finance. However, in reality, there are many more things going on that this model abstracts from. Nevertheless, the basic issue of asymmetric information shown in this model exists, to some degree, among other effects. You should not understand this paper as a statement about the used car market. So for example: You say the owner does not know whether it is a lemon or peach. Well, the model is about asymmetry. Everything goes through if the owner (on average) knows MORE than the buyer. I think that's pretty fair, and if it is not the case, we are not dealing with an information asymmetry and the model is silent on the matter. Similarly, clearly cars are not just two types, lemon and peach. But again, results will hold if you make introduce more and more classes. It just makes the math less simple. You also say that some people do not sell their car because it is a lemon. True. But again, on average, the rate of lemons for a given price will be higher than it would be if you could identify the quality of a car exactly - because it is sufficient that some people sell their lemons as peaches. You could introduce all this in the model, you just probably wouldn't learn anything new. It's a good model, because its simple and even if you throw more stuff at it, the mechanism persists to a degree. Clearly, in reality, more things are going on than just this. But reading the paper we learn about an issue that exists within the complex system, and that is valuable.