Akerlof's Market for Singles

On a recent visit to Manhattan, I spoke at length with friends about dating. According to them, dating apps are broken. Leaving aside whether they ever worked that well, dating apps seem to have a problem of adverse selection. Put simply, people looking for a serious relationship are driven away from dating apps like Hinge, and microeconomic theory can help explain the phenomenon.¹

Economic Background

My favorite concept from my favorite grad school course was George Akerlof’s “Market for Lemons.” Akerlof talks about how markets respond to asymmetric information, or a situation where one side knows something the other side doesn’t. Specifically, he examines used car markets wherein the seller generally knows whether the car is a lemon (bad car) or a peach (good car), while the buyer doesn’t know. This leads to adverse selection, a fancy term for a suboptimal outcome resulting from asymmetric information.

To expand on Akerlof’s canonical example, imagine you’re trying to buy a used car. You don’t know whether a specific car is a lemon or a peach but you do know the market contains both lemons and peaches. How much are you willing to spend on a car that is possibly a lemon and possibly a peach? If you think there is an equal probability of each outcome, you would pay something like $\frac{P_\text{peach} + P_\text{lemon}}{2}$, or the average value of the two cars. However, sellers don’t want to sell a peach for that price; they’re worth more! So, the sellers stop selling peaches because buyers won’t pay peach prices. Now there are more lemons than peaches, and buyers are willing to pay even less than they were earlier because it is no longer 50/50. This feedback loop worls until the market is all lemons. Sellers don’t sell the good cars because buyers won’t pay for them because they have uncertainty about the quality of the car. This is obviously a bad market outcome! Some lemons in the mix poisoned the whole lot, and now there are only lemons. Fortunately, there is a solution.

One way to reverse the lemon feedback loop is to disclose the asymmetric information. For example, one seller may start offering certified inspections of his cars, so buyers can find a peach and be willing to pay peach prices. If one seller does this, it drives the prices for certified vehicles up, and lemons further down. In turn, this pressures all sellers to reveal the condition of their cars. Then, they sell lemons for lemon prices and peaches for peach prices, rather than the initial condition (selling all cars for an average of lemon and peach prices) or the adverse selection outcome (selling only lemons for lemon prices). It’s important to note that the disclosure of the asymmetric information must be easy and relatively cheap. For example, if getting a mechanic to certify a car was very expensive, it might not be worth it for the sellers to disclose.

Rise of the Lemon-Humans

It’s hard to find “the one.” People struggle despite making sincere efforts at it. Dating apps and their algorithms intend to simplify the problem by making it easier to connect with prospects. I will focus specifically on Hinge because it is designed to produce relationships rather than flings; it is self-proclaimed to be “designed to be deleted.” The real goal is to help users filter the millions of potential matches, find promising first dates and start successful relationships. Hinge does this primarily by revealing asymmetric information through short, self-reported dating profiles. I will expand on the lemon market analogy later, but Hinge is something like the mechanic doing the used car certification. However, the mechanic is lazy so he just asks the previous owner if it works and kicks the tires.

Matchmaking is inherently a difficult task. How can you predict whether two people will get married? You can’t! Duh! But that is not a reason to dismiss Hinge and other efforts. The reality is people have always relied on some sort of matchmaking service. While Hinge focuses on a few characteristics (height, political affiliation, religiosity, charm in the form of witty answers to silly questions), this is largely what other app-based and community-based matchmaking services do. Yente was appraising her community members by some arbitrary shtetl standard. Work, high school, college, and the local dive bar all bring people together based on salient features (geography, life interests, class, etc). While these institutions have other primary purposes, their byproduct is a social setting where people can find auspicious matches, at least when compared with a completely random selection from the population. Hinge’s imperfect quest to quantify compatibility is not in and of itself objectionable, but it misses out on what is perhaps the most important factor.

As I see it, the real problem with Hinge is its inability to screen for intention. Intention comes before all other characteristics in predicting the outcome of a given date: are you looking for a serious relationship? Hinge tries to select such people, but invariably the less serious serial daters enter the mix. A first date between people with mismatched intentions will likely go poorly, with each feeling like they wasted their time. Whether the intention mismatch is a result of duplicity or not, it is bad and could lead to adverse selection. In the Hinge market for singles, you have lemons and peaches just the same. In terms of intention, the lemons are those looking for hook ups and the peaches are looking for soulmates. The problem is it is costly to find out whether someone is a lemon or a peach. It takes a number of dates, heck it could take years, to determine whether someone is open to a serious relationship.

As predicted by Akerlof, my friends (peaches) are leaving the market for pre-Internet alternatives. It’s not worth the risk of wasting time, money, and effort on a potential lemon and there’s no good way to “certify” peaches. But what can deliver an accurate appraisal of someone’s intentions? Can we learn anything from the used car markets’ techniques for information disclosure? Here’s an incomplete list of ways out of the adverse selection trap adapted from used car dealerships:

• Warranties and guarantees: Used car salesmen will guarantee something about the car to assuage your concerns, like offering powertrain warranties. It would be challenging to implement something like this because the dating app is merely the entry into courtship, which is necessarily unpredictable. You are owed nothing by a first date, although you would prefer that there was an ex ante chance of something materializing out of it. Unfortunately, it would be impossible to guarantee such an ex ante chance.
• Certifications and Inspections: You can’t check the CarFax on a dating prospect. You could plausibly verify the objective claims on the profile, but how do you validate intention? It’s ungainsayable short of an MK Ultra breakthrough.
• Insurance: See above.
• Signaling: Used-car sellers can create signals of value, like having a nice office, classy ads, etc. People try to do this on their dating profiles, but it’s easily insincere or faked. How many self-professed 6’ tall men are 5’11?
• Regulations: While lemon laws may work, it’s not practicable for the government to ban or punish Hinge sharks. Some norm that involves shaming lemons until they leave the Hinge platform could have some legs.
• Reviews: Similar to how you might check the Google Reviews before bringing your car somewhere or wince at a driver with a 4.2 Uber rating, you could probably implement something like this for dating. Examine the last few dates and see how it went. This could get really ugly, but if you restricted the post-date survey to questions about intention, it could work. Indeed, reviews seem to have the most promise, and it’s no wonder they have a prominent history in courtship. More on this later.

In general, intention is difficult to disclose because it is easily faked; it is difficult to catch someone lying about intention, and there are few consequences if caught. Indeed, the principal intention issue is closely related to the pseudonymity of dating apps. As you expand your network of potential matches, you lose the local knowledge that was present in the various relationship breeding grounds of yore. You see a profile, but know nothing of the person’s reputation, status, family history, etc. This information is very informative in finding a good match and it’s why many people meet others at work or school or elsewhere in their community. Moreover, the reputational risk associated with dating within a community encourages good behavior (not misrepresenting yourself) by way of social consequences (ostracism). Perhaps I am too Seeing Like a State pilled, but dating apps could be construed as a high modernist project to make the dating landscape more “legible.” It would be impossible to learn and concisely relay every individual’s role and reputation in their community. The app overlords try to simplify this by reducing people to a comprehensible set of facts. This simplification allows you to screen dates at scale, but it doesn’t work that well because it ignores many of the most important aspects of a person’s long-term desirability. It has no sense of “memory” and does not discourage bad behavior because your next match knows nothing of your last.

When switching from traditional, community-based methods to apps, you trade accuracy for scale. Ideally, Hinge would “set you up” like a friend might; that is, using intimate knowledge about you, it would arrange a date with an auspicious match. When a friend does this, they’ve intuitively screened each person’s intentions, personality, and many other relevant characteristics at an accuracy far beyond that possible with a dating app’s self-reporting. But a friend only knows so many people; with this approach, you miss out on most of the world. The dating app opens your horizons but has less fidelity. Instead of getting one person who shares your values, you get several who are of the requisite height, political affiliation, and wittiness.

To recap, in our Hinge dating market we have both lemons and peaches. We have incredible difficulty in discerning between the two and learning the other’s intentions. Accordingly, we would expect the peaches to leave the market. As more peaches exit the Hinge market, whether out of frustration or because they find success, this increases the ratio of lemons and decreases the satisfaction of Hinge first dates. In turn, more will turn to their community and local knowledge to find more promising candidates, at the expense of the infinite options on the Internet.

There will likely be more entrants into the apps market with the goal of creating lasting matches, but all will suffer some version of this problem. People want different things, and you can try to cream-skim the peaches from Hinge, but you’re going to get lemons too and the market will suffer from adverse selection. This can happen with local matchmaking solutions as well. The dating market of the past likely depended on strong cultural norms towards family formation, monogamy and chastity, which kept many lemons from the market. We may not have those same norms today, but culture will evolve and new lemon protection norms will emerge.