Market design is always political

El 16 de noviembre Alvin Roth, Nobel de Economía 2012, brindó en nuestra universidad la Conferencia exclusiva Who gets what and why: the new economics of matchmaking and market design, en el marco de las actividades de la Maestría en Economía UCEMA.
Una de las centrales contribuciones de Roth es haber colaborado con el desarrollo del sistema de intercambios de donaciones de riñones (en inglés, kidney paired donation), en los que parejas de donantes y receptores que son incompatibles buscan a otro par o pares de donantes y pacientes compatibles para realizar un intercambio, reduciendo el tiempo de espera.
Roth es un verdadero ingeniero de mercados. Es uno de los pocos economistas que tienen la capacidad de, además de ser fuerte en lo teórico, hacer desarrollos muy aplicables, que buscan mejorar el bienestar. A continuación, compartimos los pasajes centrales de este importante encuentro:

When people think about markets, they think about commodity markets. Commodity markets are markets in which prices alone determine who gets what. Think about the New York Stock Exchange (NYSE). When buying securities, you don’t have to worry about whether the seller will pick you. Likewise, the seller doesn’t have to pitch himself to you. The job of the market is price discovery, to find the prices at each moment that makes supply equal demand for each security in the market. On the NYSE, the price decides who gets what.
But not all markets are commodity markets. In matching markets, you care who you are dealing with, and the details of the deal, and the job of the market is more than price discovery. You can’t just inform Stanford University that you are enrolling, even if you can afford it, you have to be admitted. And, in the same way, labor markets don´t work by prices alone. You can´t just show up for work at Google, you have to be hired. Neither can Stanford or Google dictate who will come to them. So, these are matching markets, markets in which you can´t just choose what you want even if you can afford it, you also have to be chosen. A market involves matching whenever price isn’t the only determinant of who gets what.

Making Markets
Thicker, Faster and Safer
Economics is about the efficient allocation of scarce resources, and about making resources less scarce. Marketplaces help shape and satisfy people’s desires, bringing together buyers and sellers, students and teachers, workers and firms, and even those looking for love. Matching markets are markets where sellers and buyers must choose each other, and price is not the only factor determining who gets what.
Now we are starting to see new Internet marketplaces designed to help people match to each other. eHarmony, match.com, tinder are marketplaces that deal with the problem of matching. Upwork is a labor market. Uber and lyft match travelers to cars. Google, Bing and Facebook are markets to match advertisers and people who are looking for something.
Markets that involve offers and responses require easy two-way communication. Consider Airbnb as an example of a marketplace. Think what Airbnb have to accomplish before they could become comparable to hotels. The first task of a successful marketplace is bringing together many participants who want to transact, so they can seek out the best transaction. Having a lot of participants makes a market thick. And initially Airbnb did that by buying up their competitors, who were other websites, a lot like Airbnb. And today its main competitors aren´t other websites, their main competitors are big hotel chains. Markets can be congested, even on the Internet. And hotels have some advantages over Airbnb in terms of managing congestions.
Congestion is what happens when you are successful in making a thick market. Now there are lots of rooms you can get on Airbnb. However, hotels have a big advantage over Airbnb. If you want a room at the hotel, you can call up and say “I want a room for tomorrow night” and they can tell you if they have an available room, because they can look at all the rooms at the same time. But, that is not the case with Airbnb because each host might just have one apartment or room to rent. Think about how much less convenient the hotel would be, if you could only call to inquire about each room individually.
A big problem for Airbnb was to figure out how a market with many hosts offering one room at a time could compete more effectively with hotels. Price was important. But it was the spread of smartphones that helped Airbnb close the speed gap, and that may have mattered even more than price. The smartphones make the home hosting market work better not just because hosts can respond faster but also because they can update their bookings, which makes them more informative. In addition, now when you go on Airbnb, if you try to rent a room, that would immediately make the room disappear from everyone else´s app and browser. So the rooms that you are trying to rent are not simultaneously being competed for by other people, and as a result helps travelers search more efficiently, with fewer timewasting false leads. That makes it easier, it helps you deal with the congestion that comes from having a thick market.
So marketplaces have to make markets thick. If they succeed in making them thick they have to deal with the resulting congestion. They also have to make the market trustworthy and safe, if you are trying to rent an Airbnb apartment you have to have a reason to believe that the apartment will look like the photograph on the Airbnb site, and the person who is posting should have reason to believe that you will be a polite guest, who has successfully been a guest at other Airbnb sites. So they have developed different types of reputation systems.
These are three things that marketplaces have to do. They have to make markets thick, they have to deal with congestion and they have to make the marketplaces safe places to do business.

The US Market for New Doctors
The US market for new doctors is particularly instructive, since at various points in history, it suffered from many of the failures that are common in matching markets. The first job American doctors take after graduation from medical school is called an internship or residency. Before 1900, when people graduated from American medical schools, they started to look for their first job after medical school near the end of their final year. But by 1930 hospitals began to try to hire interns a little earlier than competing hospitals. This unraveling of the market moved the date of appointment earlier, first slowly and then quickly. By around 1940, the standard time at which American doctors could be hired was in the summer after their second year. So they were being hired two full years in advance, and this caused lots of mismatches.
The way American medical schools work is: for the first two years you take classes and then you start meeting patients in your third and fourth year. So, at the end of the second year, after getting an A in anatomy, a student might have wanted to be a surgeon, only to discover in his third year, when he finally could watch surgeries up close, that he fainted at the sight of blood and that he didn´t want to be a surgeon at all. But it was too late because the good jobs were all gone. The student and the surgeon who hires him would both have made a bad match.
They understood that there were bad things happening in the market for new doctors. In 1945, the medical schools intervened and they agreed not to release any information about students before a certain date. They wouldn´t even confirm that the student was enrolled in their medical school, and that solved the problem of going early. From 1945 to 1950 they moved the date of appointment back from two years in advance to about four-five months in advance, where it is now.
Once the date of appointments was controlled, a new problem emerged: hospitals started to make exploding offers to students. By 1950 it was customary for a medical student to get an offer of a job by telephone, and to be told that he has to say `yes´ or `no´ while on the phone. Candidates had to reply immediately, even before they could learn what other offers might be available. Hospitals had learned from experience that if they let a candidate think about it, and he later decided not to take the job, all the other people they wanted to offer the jobs to would have taken other jobs. So, even though they had a thick market, even though there was a time in which all the jobs could be offered, the fact that they were making exploding offers meant that job candidates couldn´t consider many offers. They could only consider one, the one that they had on the phone, and it took some courage to turn down a pretty good job in the hope that when you put down the phone it would ring again with a better job. That wasn’t a very satisfactory state of affairs either.
After having suffered from this kind of congestion for five years, in the early 1950s the doctors reorganized the market in a big way. Instead of a completely decentralized market, they proposed to organize it through a centralized clearinghouse. This clearinghouse, this way of organizing the labor market was voluntary, no one had to participate in it. However, it got very high rates of participation. Instead of making offers by telephones, the hospitals would rank order their candidates. Here´s our first choice, our second choice, third choice. And the medical students would rank order their jobs. After some trial and error; they developed an algorithm in a way that got very high rates of voluntary participation. More than 95% of people were going through the whole process voluntarily. Graduating doctors and residency programs both submitted preference lists and then they signed contracts with those to whom they were matched.
The Medical Match worked smoothly for decades. One of the secrets of their success, as in other successful labor market clearinghouses, was that they produced outcomes that were stable, in the sense that no applicant and residency program not matched with each other preferred each other to their assigned matches. But this started to break down in the 1970s when women started to enter medical school in significant numbers.
In the 1950s almost 100% of American doctors were men, but by the 1970s about 10% of the medical graduates were women, and today half of the American medical graduates are women. One thing that medical students can do, even though they are very busy, is marry each other. So last year there were about seventeen thousand graduates of American medical schools and about two thousand of them were married to each other. That is, there were about a thousand couples who went on the market together and wanted to find two jobs, near enough to each other so they could keep living together.
Starting in 1970, this created a new problem for the Medical Match, as these couples would sometimes decline the offers suggested by the Match. It wasn’t long before some couples declined to participate in the Match at all, but instead communicated directly with hospitals that might hire them. The Medical Match was no longer producing stable outcomes.
Even when the process produced two jobs in the same city, couples sometimes didn’t take those jobs because one member of the couple was a offered a good job while the other member a not so good job. Instead, they worked the phones to find jobs they liked better, often with success. This problem is a good illustration of what I like to call the Iron Law of Marriage, which says that you can’t be happier than your spouse.
The stability of the Match produced an outcome that was stable for a doctor and a job, but it couldn’t generate stability for two doctors and two jobs if those two doctors were a couple. It turns out that there’s no way the clearinghouse could produce an stable outcome with respect to couples and jobs if it didn’t allow those couples to express their preferences for pairs of jobs. But this is a very complicated problem. One of the things that I showed in a 1984 paper was that even if you allow couples to indicate their preferences for pairs of jobs, for some preferences there may not exist any stable matching.
I still remember in 1995 when I was sitting at the University of Pittsburgh, and my phone rang, and the Director of the National Residential Matching Program was on the line and he asked me whether I would redesign the medical match. I remember the feeling I had on the phone which was: “Oh, I´m sorry I picked up the phone” because I knew it was a hard problem. But when you agree to redesign the market, it becomes your problem. That was really the moment which I became a market designer, when I decided to accept responsibility for the hard problems.
I set only one important condition for taking on the task: I wanted to collaborate with Elliot Peranson, who had been providing technical support for the Match for years. We eventually developed a hybrid algorithm that has come to be called the Roth-Peranson algorithm. It finds a preliminary matching of doctors to residency programs starting with a deferred acceptance algorithm, which yields an outcome containing some blocking pairs. Then it tries to fix them one by one.
We knew we wouldn’t be able to find a stable outcome if none existed. But to our happy surprise, when we look at the data, it turned out it was almost always possible to find a stable matching. That the Match has been able to help hundreds of thousands of doctors, including couples, find jobs in thousands of residency programs is a testimony to the flexibility of markets as a tool for coordinating complicated human endeavors.

The transplant of kidneys
When we see a long line of people waiting to buy a scarce good, we suspect that demand exceeds supply and we may conclude that this shortage is occurring because the price is too low to generate more supply. In the US and in Argentina there is a real shortage of kidney transplants. This morning in the United States about one hundred thousand people are waiting for deceased donor kidney transplants, but we only get about twelve thousand a year. Many of those waiting for kidneys are on dialysis, and life expectancy while on dialysis isn’t long. Thousands of people die each year while waiting. Meanwhile, the price of kidneys is zero, since it is illegal in almost every country to buy or sell kidneys for transplants. The kidney itself must be a gift.
But kidneys are unusual. Healthy people have two kidneys and can remain healthy with just one. So if you love someone who is dying of kidney failure, you might be able to save his/her life by donating a kidney. There are close to six thousand living donations each year in the US. That means that in the US there are almost as many living donors as deceased donors (deceased donors donate two kidneys). But sometimes, even though you are healthy enough to give a kidney, you can’t give it to the person you love because kidneys have to be matched, they have to be compatible. And this is what opens up the possibility of kidney exchange and brings economists into the market design process.
The simplest type of kidney exchange is between two patient-donor pairs. In more complex cases, additional pairs can be included generating cycles with more than 2 kidney transplants. Notice that only kidneys are exchanged here, not money, no kidney is sold and that´s because it´s illegal to buy a kidney for transplant. Section 301 of the US National Organ Transplant Act said: “it shall be unlawful for any person to knowingly acquire, receive or otherwise transfer any human organ for valuable consideration for use in human transplantation”. We started doing kidney exchange in the early 2000s. It took until 2007 before Congress finally amended the National Organ Transplant act to say that in fact kidney exchange is legal in the United States (but using the polite phrase “human organ paired donation,” which avoids use of the word “exchange”).
To make sure that both parts of the transaction go through, in a kidney exchange doctors perform the surgeries at the same time. You can´t enforce a contract on a kidney (since the law forbids “valuable consideration”). If for some reason one pair donated a kidney but then didn’t receive one, then this donor and her patient would really have suffered. They would have had a surgery that didn´t help them and more importantly, they would no longer have a kidney to take part in a kidney exchange. So, to make sure that never happens, we always do these simple two way exchanges simultaneously. But that means that you need four simultaneous operating rooms and four simultaneous surgical teams to be able to perform the two nephrectomies (kidney removals) and two transplants at the same time. So it´s hard to do, it´s a congested marketplace.
An exchange that begins with a a non-directed donor is a chain rather than a cycle, since it doesn’t have to return to its beginning: the non-directed donor is an altruistic person who arrives without a patient and is willing to give a kidney without receiving one in return. In the past, non-directed donors were directed to people on the deceased-donor waiting list, generating only one transplant. But kidney exchange made it possible for a non-directed donor to generate more than one transplant, since a chain could start with a non-directed donor, include some patient donor pairs, and end with a donation to someone on the waiting list.
In a non directed donor chain it´s not as costly if a link is broken as it is in an exchange entirely between pairs. Suppose a link is broken, e.g. on day one the non-directed donor gives his kidney to recipient one and on day two, for whatever reason, the donor who is the partner of recipient one fails to give to recipient two. That would be very disappointing, but it wouldn´t be a tragedy for the second pair because they were scheduled to get a kidney before they gave one. They still have a kidney even though the transplant didn´t take place. So the following week, they can take part in kidney exchange again. Therefore, the cost of a broken link in a chain is much lower than in a cycle and that gave us room to explore its benefits.
Over time we have formed many nonsimultaneous chains and those chains can sometimes be very long. As experience accumulates, the evidence grows that potentially long nonsimultaneous chains are good for kidney patients, specially for the hardest-to-match patients. Today almost half of the kidney exchanges in the US come through nonsimultaneous chains. And kidney exchange has grown to be about 14 percent of the living donor transplant in the US. It started to spread around the world. There´s been a kidney exchange in Argentina at the Fundación Favaloro in Buenos Aires. But it´s still complicated in Argentina: It requires permission of a judge. Argentine transplant law is a little complex. Lots of transplants are done in Argentina but they are almost all transplants from deceased donors. There are not so many living donor transplants. So, in Argentina there is important room for growth in living donor transplants.
There are parts of the world where very few transplants are done. For example, in Philippines where it´s possible to get a kidney transplant but very few people can do it because the Philippine health insurance doesn’t cover it. It’s not affordable for many people who could otherwise have a kidney transplant.
Now we´re starting to develop Global Kidney Exchange, a way to bring patient donor pairs from abroad into US kidney exchange. Every time you transplant an American, who is otherwise on dialysis, that saves the American healthcare system about a quarter of a million dollars in the first five years. If you can bring a patient donor pair from the Philippines that could have received the transplant but it’s unavailable to them for financial reasons, then you can finance their transplantation in the US.
In January 2015, the kidney exchange organization called the Alliance for Paired Donation did the first Global Kidney Exchange, with a pair from the Philippines. Jose Mamaril received a kidney from a non-directed American donor in Georgia. His wife, Kristine, donated one of her kidneys to an American recipient in Minnesota, whose donor continued the chain by donating to a patient in Seattle.
In Argentina, where many patients from neighbor countries come to receive a transplant through the INCUCAI system, global kidney exchange could be a good possibility to increase living transplantation, by allowing foreign patients to bring living donors into kidney exchange.

REFUGEES
There are plenty of matching markets that are not working as well as we would like. One of them is refugee resettlement. That´s a matching market because you can´t just decide where you’re going to live, you have to get a visa and countries can’t just decide where people are going to live. Refugees are precisely the people who don´t stay where they don’t want to be.
Refugee resettlement is going to be a different kind of matching market, it’s going to be very political, but we will have to learn to do it better because in the next century, if the sea level rises, there will be lot of human migration.

TO SUM UP
Market design is always political, even when you are talking about kidneys or labor. There’s a lot of politics to be done. Market design is the engineering part of microeconomics and it’s an ancient part of human activity. People have been designing markets since before the invention of agriculture. And markets are very human things. Economist have recently got into the design process and we try to understand marketplaces rules well enough to fix them when they´re broken.
We can learn a lot about economics by looking around us and seeing what works well and what doesn’t, and thinking about how that might be fixed.

Agradecemos la gestión de la Asociación Argentina de Economía Política (AAEP), la Facultad de Ciencias Económicas de la Universidad Nacional de Tucumán y el Ente Provincial Bicentenario Tucumán 2016, que permitieron concretar la visita de Alvin Roth al país. También al Banco Ciudad por el acompañamiento para el desarrollo de la actividad.