Who Gets What — and Why: The New Economics of Matchmaking and Market Design

by Alvin E. Roth

AS WE SAW in the previous chapter, markets can be dramatically improved when their design encourages people to communicate essential information they might otherwise have kept to themselves. But sometimes markets suffer from too much communication. It is a paradox of market design that as communication gets easier and cheaper, it sometimes also gets less informative. We are seeing this ever more clearly as communication becomes more electronic. Email and social media are good examples. As the volume of messages grows, sorting the real ones from the spam, and the ones that deserve thoughtful replies from those that can be quickly acknowledged or ignored, becomes harder. And as entire markets move to the Internet, this overload of messages can make markets congested.

Notice that I’ve just spoken about signals for two quite different kinds of information. First, is the candidate qualified enough for the college, the romantic partner, the employer, to be worth further investigation? And second, is she interested enough in them to repay that effort? Both kinds of information are especially valuable in congested markets, as there isn’t time to explore every possibility. So signals, and how to send

You might ask, Why do so many colleges use the Common App instead of requiring individual applications? Individual applications would make it harder for students to apply, so each application received would be a stronger signal of interest. The answer is that while the very top schools can do what they like, all the others must face the fact that the Common App has become a very thick marketplace. It attracts so many applicants that a college that now insists on individual applications might not get enough applicants, since many students are drawn to the convenience of using the Common App.

*

As colleges pay more attention to these signals in order to help them sort through increasing numbers of applications, the signals themselves may take on different meanings. That is, signals sent deliberately may convey different information than signals sent in passing.

The women in turn reply to a smaller and smaller percentage of the messages they get, and the men respond by sending even more, and even more superficial, messages. Economists call such superficial messages “cheap talk.” When talk is cheap, it doesn’t reliably signal anything.

Good grades signal good study skills and hard work, high intelligence and aptitude, or all of the above, which are all attributes that help students succeed in college. Signals such as high grades—or a peacock’s tail or big bank building—don’t indicate interest; they are marks of desirability. That is, the gaudy tail doesn’t signal how much the peacock is interested in a particular peahen; it signals how much the hens should be attracted to that particular cock.

Asking a person out on a date in person offers lots of opportunities to send both kinds of signals. By comparison, arranging a date on the Internet, which makes the dating market thicker by making initial contacts easier, also makes it harder to send credible signals to cut through the congestion.

That’s why mechanisms that limit the number of signals you send can accomplish the same thing over the Internet. When other signals may be cheap talk, these signals indicate that you are interested enough to use scarce resources that you can’t just send to everyone.

So a scarce signal isn’t cheap talk; it comes with an opportunity cost—you could have sent that signal to someone else instead.

There are lots of ways to run an auction; auction design is one of the most active parts of market design, as well as one of the oldest.

Notice that while a second-price auction makes it safe for bidders to bid the true value to them, it doesn’t necessarily impose a cost on the seller, even though the seller receives only the amount of the second-highest bid. That’s because in a first-price sealed bid auction, for example, it isn’t safe for bidders to bid their true value; they have to bid less than that if they are going to make any profit, since if they win the auction, they will have to pay the full amount of their bid. So the seller in a first-price auction receives the amount of the highest bid, which is, however, less than the true value of the highest bidder. By comparison, in a second-price auction, the seller receives only the second-highest bid, but the bids are higher.

One version of a first-price auction is used to sell cut flowers in bulk, in a “descending bid” auction. The auctioneer sets up a “clock” that has the current bid on it, starting with a very high bid and quickly descending, until some bidder stops the clock by offering the price it currently shows, which is higher than any of the other bidders have offered to pay, as they haven’t already stopped the clock. Since the first bid stops the clock, these auctions can be very fast—a good thing, since time is of the essence when you’re buying cut flowers. A big international marketplace operates this way in the Netherlands, right near the Amsterdam airport, from which flowers can be shipped around the world. As a result, this kind of descending bid auction is often called a Dutch auction.

That’s why one of the ancient uses of auctions is called price discovery: letting the market tell you what price you can get for what you are selling, and to whom you should sell it to get that price. Auctions are matching markets that match sellers with the buyers who most value what is being sold.

The FCC and the economists advising them and potential bidders quickly realized that selling spectrum licenses one at a time would be bad market design, as it would make it too risky for bidders to assemble the packages they needed. That is, if licenses were auctioned one at a time, bidders would have to bid very cautiously for fear of being left with only part of the package they wanted, which would be less valuable to them than what they had paid. To put it another way, it wouldn’t be safe for bidders to fearlessly bid for each license the amount it would be worth to them as part of the right package, since they would have to pay their bids even if they couldn’t ultimately put that package together. To address this problem, the FCC adopted an auction design in which many licenses were sold at the same time, in “simultaneous ascending” auctions, with the rule that no auction would close until they all closed.

Bidders reluctant to share their intentions would want to wait until near the end of the auction before bidding, as we saw in chapter 7 when we considered sniping in eBay auctions. But if everyone waited, the information needed to produce an efficient allocation wouldn’t be transmitted. To avoid this, the design for the spectrum auction included activity rules, proposed by my colleagues Paul Milgrom and Bob Wilson, to prevent bidders from making late bids unless they had made bids on equivalent numbers of licenses (measured in terms of population served) earlier in the auction. Thus big bidders had to make their bids known early, and all bidders could adjust their bids in light of the competition. Simultaneous ascending auctions with activity rules enabled many bidders to compete simultaneously for many licenses, creating a thick market in which price discovery could take place. The activity rules also kept the auctions from dragging on interminably—another possible side effect of thick markets having to cope with the congestion of many possible transactions.

That’s because when you use the Internet, your eyeballs are for sale to the highest bidder, in some of the fastest auctions ever.

all this new technology impinges on our privacy, that we’ll want legal restrictions on some kinds of transactions involving our private data. Property rights—who owns what, and what they can do with it—are an important part of market design, and I predict that we’ll be seeing some new efforts to define the property rights to our transaction data.

*

Economists say transactions have “negative externalities” if they harm people who aren’t party to the transactions. If your neighbor opens a nightclub and throws loud parties at 2:00 a.m. and the noise wakes you up, that’s a negative externality.

But the criminal organizations that had grown rich under Prohibition moved into other businesses and were a long-lasting reminder of how prohibiting a market is a ham-handed form of market design that doesn’t necessarily achieve even its main objectives.

Adam Smith, in his book An Inquiry into the Nature and Causes of the Wealth of Nations (1776), famously observed, “It is not from the benevolence of the butcher, the brewer, or the baker, that we expect our dinner, but from their regard to their own interest.”

And once you observe that matching is one of the major things that markets do, you realize that matching markets—markets in which prices don’t do all the work, and in which you care about whom you deal with—are everywhere, and at many of the most important junctures of our lives.

*

But the reason natural languages aren’t really under our control is because they emerge from the interaction of millions of users. English speakers certainly know that it’s hard to change the spelling of a word that could be spelled more simply and in a way that is closer to how it is pronounced.

As we start to understand better how markets and marketplaces work, we realize that we can intervene in them, redesign them, fix them when they’re broken, and start new ones where they will be useful. The growing ability in recent years of economists to be engineers is a bit like the epochal transformations that farming or medicine have experienced over the millennia.

Because markets and languages are tools that we use collectively, they may be hard to redesign even when they are working badly. As a result, we have to limp along with some bad designs, like awkward spellings.

This is more important than he puts an emphasis on, for example, how American Democracy is intrasigent...it doesn't rally change

But sometimes we do get to redesign markets that are working badly. And sometimes we even get to design entirely new markets. These are opportunities to treasure, to study, to approach with humility, and to monitor with care. Markets are human artifacts, not natural phenomena. Market design gives us a chance to maintain and improve some of humanity’s most ancient, essential inventions.