I like to think that creative people think non-linearly. So, if you’re one of those people who were induced to debate the merits and demerits of George Akerlof’s “The Market for Lemons” (1970 [gated], [ungated]) because you read the Janet Yellet news,1 you can consider that a good thing. A not so good thing is to reject the lessons from Akerlof’s paper, because (a) he advocated intervention as a means of mitigating uncertainty-related welfare losses, and/or (b) for something like “he didn’t consider preference subjectivity and heterogeneity,” and/or (c) for whatever other reason. This is a good opportunity to stress some of the interesting things that come out of Akerlof’s model.
I think it safe to assume four groups of Akerlof readers.
- Group 1 are those who read the paper and favor interventionism predominately over private solutions.
- Group 2 are those who favor both interventionism and private solutions — there may be scope for both.
- Group 3 are those who don’t think much of government solutions, but accept the conclusions of the paper and focus attention on private solutions.
- Group 4 are those who look at Group 1 and 2, assume something must be wrong with the paper, and reject the conclusions altogether.
The point I hope readers take away is that you don’t want to be in Group 4, or even near it, and that, for the most part, it’s more interesting to be in Group 3 than in any other group. I’m somewhere between Group 2 and Group 3. I suspect that most economists are in Group 2, but that Austrians, Market Monetarists, and economists with similar (skeptical) priors on government are more-or-less where I am (maybe most are relatively closer to Group 3).
I won’t discuss (b), other than to suggest to read the article, because in his model — while still simplified — he does assume two different utility functions (for two “groups of traders”). Actually, the assumption of heterogeneity is important, because it helps explain why, at the extreme, a market might breakdown as a result of uncertainty over the quality of some good. Differences in people’s (subjective) utility functions helps us illustrate why, even if “objectively” (if there were perfect information) there is opportunity for gains from trade, two people may still not undertake the exchange.3 Also, I saw some discussion on whether or not we can really say what is an “average quality car,” or at least if we can say that one car is really of poorer quality than another. After all, one man’s trash is another man’s treasure. But, I think we can reasonably assume that the average person prefers a car with a working engine than a car with a non-working engine (assuming the same general end of, say, wanting to drive the car).
Before talking about (a), which will really be about talking about the interesting aspects of the paper, I will summarize Akerlof’s argument. He makes a good analogy to Gresham’s Law, although after explaining his theory, but I think it’s a good thing to think about Gresham’s Law beforehand, because it might make Akerlof’s argument clearer. Gresham’s Law is commonly summarized as “bad money drives out good money.” The assumption behind the theory is that “bad money” (say, a coin that is 50% gold and 50% copper) trades at par, meaning it trades at the same value as “good money” (say, as pure a gold coin as possible). Why would this happen? Assume the government enforces the at par value of “bad money.” People, who know that “good money,” despite the enforcement (by fiat, here), is really worth more than “bad money,” will get rid of the latter first. As a result, the average quality of circulating coin falls.
Akerlof’s theory differentiates itself from Gresham’s Law in that, while the latter assumes people can tell the difference in quality, the former is all about an inability to perfectly distinguish between goods by quality.4 In other words, it assumes imperfect information; more importantly, it assumes that different people know different sets of information, allowing one to take advantage of the other.5 Why does information asymmetry exist in the car market? Suppose someone has owned a car for n time, and that a buyer has no experience with that same car. It’s reasonable to suspect that the owner, who has n time experience with the car, will know more about it than the buyer, who has 0 time experience with the car. I am a real world example. A drunk driver recently hit my, now ex-, car, forcing me to buy another one. I first looked on Craigslist for another used car, but it was very difficult for me to know the history of the different options. One guy was selling a 2002 Honda Civic, but he has an incentive to withhold information I find crucial (accident history, et cetera). I opted not to buy the car, even if I would have actually gained from the exchange.
Because I can’t tell the difference between the various 2002 Honda Civics on the market — as far as I know, they are all of similar quality —, I have to accept a more-or-less uniform price for this good. At this price, p, sellers have an incentive of getting rid of their worst quality goods first. On average, this drives down the quality of the supply of that good. Furthermore, sellers of good quality goods, even if they don’t have bad quality goods, have an incentive to not put their goods for sale on the market. Let’s say that price, p, is a function of the expected average quality of the good, such that the poorer the quality — as people come to recognize it — the lower p will be (what Akerlof calls the price level). If a seller of high quality goods enters the market, raising the expected average quality and therefore p, she’s not guaranteed the benefits of providing a higher quality product. Because people can’t distinguish between goods (again, in reference to their quality), that seller of the good product will raise p not just for her goods, but also for other sellers, who are offering a lower quality product.6
Using Akerlof’s model, we can predict that, under these conditions, there could be no market at all! Akerlof illustrates this using some quick algebra, but I will try to summarize the argument without any math (the math is not difficult, so you will probably be able to get the clearest idea if you just read that part of the article — part II, section B, pp. 490–492). Remember that the argument is that at any given “price level,” or price p, the average quality of the good for sale will be lower than if buyers could differentiate by quality. The result is a feedback loop of sorts, where the average quality continues to fall, bringing down p with it. This is because as the expected quality of the average car falls, people will be willing to buy it only at a lower price. As this process takes place, at any given p the expected quality can be so low as to dissuade any purchase at all. This is true even if, if people knew better, there is a car out there that is worth price p (according to their subjective preference).
The result is what economists call a market failure, but is probably better called simply a market imperfection. There will be trades that are pareto optimal, but won’t take place because the traders in question are uncertain as to whether they will really gain from trade. We can use this theory to help understand why certain markets — e.g. health insurance,7 cars, labor markets, et cetera — take the form they do. Then we can take two routes: (a) give a policy recommendation, or (b) use this information to explain private solutions. Those in the aforementioned group 4 of Akerlof (1970) readers reject the article because they assume the only option is (a). But, there is also option (b), which is the more interesting one. And, contrary to the claims of some,8 Akerlof does discuss (b), mentioning (a) only in passing, really (see section IV, pp. 499–500).
Humans, being extremely clever animals, have come up with a number of ways of mitigating the problem of quality uncertainty. This is why, in real life, we see functioning markets selling goods that are subject to the conditions of Akerlof’s theory. A seller, for example, can offer a guarantee for her product. If the quality of a car is higher than the expected average quality, the seller might command a higher price if she offers a guarantee to refund the customer if the quality is less than that promised. Those selling worse quality cars won’t offer the same guarantee, because their expected probability of pay-out is higher. Brand names also help mitigate this type of uncertainty, since brands can act as a proxy for the quality of a good, and consumers can discriminate between brands. Licensing also serves as remedy. These are all institutions that arise to help mitigate market imperfections, and they can all be private solutions.
What makes Akerlof’s theory interesting, then, is that we can explain why certain markets look the way they do, including the institutions that arise. These institutions would otherwise look very strange to us, because we couldn’t explain them under different assumptions (e.g. under the assumption of perfect information). This allows us to more accurately describe the world around us. For those of us skeptical of intervention, it also allows us to propose alternatives to policy recommendations. If the net benefit of a private solution (multiplied by the probability of it occurring) is greater than the net benefit of collective action, the latter is not an attractive option. Here is where we distinguish group 2 from group 3. If we think that we can accurately make these types of computations, we will belong to group 2 — whether we opt for private solutions or collective action depends on their comparative merits and demerits. If we’re skeptical of the accuracy of these computations, and inclined to think that it’s safer to avoid collective action altogether, we will belong to group 3. Like I said, I’m somewhere in between those two groups. But, for no reason should you be in group 4.
If you haven’t read the news, Yellen is the candidate with the highest probability of succeeding Ben Bernanke as chairman of the Federal Reserve. Edit: She is for sure going to be the next chairman (chairwoman) of the Federal Reserve.
2. Although, on the other hand, maybe that’s not such good news, since Yellen has made plenty of interesting contributions to economic science that deserve to be discussed more so than those of her husband, since she’s in the news and not her husband.
3. Not to mention, it would be absurd to talk about gains from trade without assuming subjective and heterogeneous utility functions.
4. Although, I’m not sure why we can’t upgrade Gresham’s Law to consider the impact of uncertainty on money — whether empirically relevant or not (and, my guess is that it is relevant, since some of the institutions discussed later in this post emerged in competitive money markets).
5. To be clearer, there is a difference between imperfect information and asymmetric information. Say that a set of perfect information looks like (x, y, z), and that there are two people in the market. These two people are imperfectly, but symmetrically, informed if both know the knowledge set (x, y). They are imperfectly, and asymmetrically, informed if one knows (x, y) and the other (y, z).
6. If people could distinguish goods by quality, then the seller of the higher quality product will command a higher price, while competitors will still be selling their goods at a lower price.
7. Kenneth Arrow actually made a similar argument to Akerlof’s, about seven years earlier, in “Uncertainty and the Welfare Economics of Medical Care” ([gated], [ungated]). Arrow, however, explains something slightly different (medical care v. healthcare).
8. Specifically, this is untrue,
According to Akerlof and others, market participants, facing the realities of imperfect information, have little or no incentive to gain more information for themselves. They are “stuck” in a disequilibrium trap with no way out unless Uncle Sam comes to the rescue. Yet, we know just from simple observation that Akerlof is wrong.