by Robert Litan
By the 1950s, hospitals and medical organizations addressed this problem by establishing rules concerning when offers could be made. That soon led to congestion because hospitals were not giving medical students enough time to make decisions. To solve this problem, medical groups organized a centralized clearinghouse to better match supply with demand called the National Resident Matching Program (NRMP), which is still in use today and is a version that Alvin Roth redesigned in the 1990s to improve its efficiency (see box on Alvin Roth later).
As the NRMP example demonstrates, introducing a centralized clearinghouse can improve a market that previously, in the words of market designers, could unravel, and fail to produce an efficient outcome. This solution raises a question, however: Can all decentralized markets be made more efficient by introducing a centralized clearinghouse to match supply with demand?
According to Clayton Featherstone, a former post-doctoral fellow who was mentored by Roth at Harvard Business School and who now conducts research in market design and matching theory at the Wharton School of the University of Pennsylvania, the potential gains of having a centralized clearinghouse must be weighed carefully against the potential gains of having numerous smaller exchanges compete with each other for market participants.15 As always with market design, details matter; so it depends on the market in question.
For example, as we will see below in the dating market, there are currently several dozen viable dating websites available to individuals, each offering a different interface and user experience, with some catering to very specialized audiences. This is a case where the net benefits of a single clearinghouse are not sufficient to justify the costs.
There is a good counterexample, however, and it is the market for kidneys. This is a market that Roth has thought a lot about and has helped to greatly improve in the past few years. In the United States, as in many other countries, commercial trade in human organs is illegal, mostly for ethical reasons, which is why the market for kidneys is sometimes described as a repugnant market. Kidneys therefore are allocated in the United States through a donation system that matches patients with donors. There are currently about 70,000 patients in need of a kidney transplant, yet only about 11,000 a year receive them.16
Since people are born with two kidneys and can live healthy lives with just one, a person can donate a kidney to someone in need. However, since the patient and donor may be incompatible (because of differences in their blood types, for example), they would have to locate another such pair in which the donor in one pair could donate a kidney to the patient in the other pair, and vice versa.17 Up until the early 2000s, these so-called paired exchanges were rare because it was difficult for people to locate compatible patient-donor pairs.18
In this situation, then, it is clear that the market lacks thickness and would benefit from having a centralized kidney exchange to bring together more compatible donor pairs and thereby permit the transplantation of many more kidneys. Roth and his colleagues M. Utku Unver and Tayfun Sonmez argued exactly this point in a 2004 article.19 Kidney surgeons in New England were receptive to the idea and, with the help of Roth and his colleagues, the 14 kidney transplant centers in that area were brought together to create the New England Program for Kidney Exchange. Roth believes there is a substantial business opportunity for someone to organize a single, national kidney (and other organ) clearinghouse. In fact, Congress has established a national pilot program for kidney exchanges that is being operated as a nonprofit organization (like the New England organization). Kidney transplants not only offer vastly greater quality of life for patients than dialysis, but also save money; by one good estimate, $60,000 per Medicare beneficiary.20
Where there are cost advantages to a single clearinghouse, such as in the market for kidneys, economists cite this as evidence of economies of scale. With kidneys, what matters is that the number of patient-donor matches is maximized. However, as we will see in the example of dating websites, there is more room in that setting for product differentiation and, indeed, that is partly what has contributed to that market’s decentralization, which ends up benefiting consumers because greater competition leads to the differentiation that consumers want.
Alvin Roth
Alvin Roth is one of the pioneers of market design and has made significant contributions to matching theory, game theory, and experimental economics. Born in New York City in 1951, Roth went to high school in Queens, where he dropped out in his junior year and began taking weekend engineering classes at Columbia University.21 After a professor there suggested that he apply to college, Roth was admitted formally to Columbia and graduated with a degree in engineering in 1971. He went to Stanford University to complete his graduate work, receiving both his masters and PhD in a branch of engineering called operations research, which uses mathematics to improve decision making, often in the face of uncertainty (see Chapter 4, which discusses this field in greater detail).
Roth has taught at multiple universities—the University of Illinois, the University of Pittsburgh, Harvard, and Stanford—applying his training in operations research to explore topics in game theory, including market design and matching theory.
In the early 1980s, Roth recognized the real-world relevance of the Gale-Shapley algorithm (explained in the box on Lloyd Shapley earlier). In particular, in 1984 Roth studied the algorithm used by the NRMP clearinghouse to match new doctors with hospitals and discovered that it was closely related to the Gale-Shapley algorithm proposed decades earlier.22 By the 1990s, there were signs that the NRMP clearinghouse was encountering problems. As the number of female medical students grew over the preceding decades, dual-doctor couples were often looking for work in the same region, and this presented a challenge to the system: Many new doctors were no longer using it, a sign that it was not producing stable matches. In 1995, Roth was asked to redesign the system and modernize it to produce stable matches and improve its efficiency. The new algorithm, which he designed with Elliott Peranson, was adopted by NRMP in 1997 and is still in use today.
In recent years, Roth has applied the Gale-Shapley algorithm in a variety of other settings. In 2003, for example, Roth helped to design New York City’s public high school matching system. Before 2003, the city’s allocated spots through a complicated process that involved repeated rounds of applications and rejections, resulting in tens of thousands of unstable matches every year. The new algorithm that Roth implemented, based on the theoretical work of Gale and Shapley, resulted in a 90 percent reduction in the number of unstable matches. Today a growing number of school systems, including Boston’s, have begun to apply Roth’s techniques to improve their matching mechanisms.23
Another significant contribution by Roth to market design and matching theory is the market for kidney donors and patients, as discussed in the text. Roth’s most recent projects include helping redesign the job markets for gastroenterologists and economists, and setting up a nationwide kidney exchange, which would provide far more thickness and stable matches than any of the existing regional clearinghouses.
As is evident from the multitude of diverse applications of his work, Roth is a pioneer of applying sophisticated economic theories to solve practical problems. For his contributions to market design and matching theory, Roth was awarded the Nobel Prize in Economics in 2012 (along with Lloyd Shapley) and remains active in field experiments to this day as a professor of economics at Stanford University and emeritus professor of business administration at Harvard Business School.
Matchmaking in the Labor Market
In recent years, another market that has benefited directly from the research of Shapley, Roth, and other market designers is the labor market—perhaps the most important matching market of all. As discussed, Roth’s work in the medical labor market greatly improved a system that was in danger of falling apart, or unraveling. That is, there were signs that the centralized clearinghouse was not producing stable matches, and so medical students were often avoiding the mechanism altoge
ther and finding matches on their own.
In this section, I describe situations where the insights of market designers have produced positive business impacts in the labor market in general, helping to better match supply (job seekers) with demand (employers). In particular, algorithms are increasingly playing matchmaker to improve the efficiency of the labor market and the likelihood that the matches that take place are stable in the first place—often a significant problem in many companies.
Let’s start with the fundamentals. At a basic level, job hunting and hiring are a lot like dating. Both involve information asymmetries, costly and drawn-out search processes, and complex sets of criteria.24 And though money can play a role, it’s not necessarily the central factor in hiring, and less so in dating (and where it is, one or both sides are interested not only in wages paid today but in the future earning power of the other). Instead, the primary consideration is about the particular qualities of people that are thought to lead to enduring matches. In the labor market, much is made of concepts like fitting the company culture and compatibility of skills.
Yet we all know from personal experience that many employer–employee relationships do not work out, or in the language of the literature on matching theory, a substantial share of the matches between employees and employers are not stable. Put differently, there are probably millions of employee–employer matches that, if broken up, could lead to new matches that make all parties involved better off. Of course, it is not realistic to expect a perfect equilibrium in which every employee and employer is completely satisfied with their matches. But the insights of the market designers suggest there is room for improvement.
Some businesses have begun to make this possible. AfterCollege, Inc., for example, has developed software that thinks and acts like a human recruiter, examining the characteristics of stable matches so a computer can recommend vacancies to job seekers and candidates to hiring managers.25 Another popular online job matching service is provided by LinkedIn, the company that perhaps more than any other has upended corporate recruiting in the past decade. The company has taken advantage of its vast database of about 200 million career profiles to create an algorithm that notifies recruiters about “People You May Want to Hire.”26 LinkedIn’s growth is a testament of its growing popularity among employers and potential employees alike.
LinkedIn’s algorithm is a microcosm of what market design is all about: understanding the institutional details of a given market in order to improve the way it matches supply and demand. For example, LinkedIn discovered from mining its own data that computer engineers migrate back and forth between New York and San Francisco frequently; accordingly, its algorithm is designed to show specific companies in California a few candidates from New York, but it wouldn’t show them candidates from, say, Reno, because there is very little mobility between that market and Silicon Valley.27 The same logic can be applied to match job seekers and companies in finance in New York and London, for example.
The algorithmic approach to the matching of job seekers with employers differs from traditional hiring in at least two important respects. First, firms using algorithms focus on a broader set of employee traits than just education and experience to help employers find talent by adding cognitive ability, personality traits, and cultural fit to produce better candidates.28
Second, the new job matchmaking companies have harnessed the power of technology, and in particular the benefits of data analytics, to create algorithms that better match supply with demand and produce more stable matches.
Companies like AfterCollege, CareerBuilder, Burning Glass, and oDesk are taking advantage of their rich databases to mine information and identify traits that are better predictors of successful matches in the workplace. oDesk, for example, is the nation’s largest online market connecting businesses to remote contractors. In 2011, it introduced an algorithm that generates recommendations of best-match candidates to businesses, resulting in more filled vacancies and more stable matches.29
The new job matchmakers still face an important challenge, however. The best predictors of stable matches—characteristics like creativity, personality, and cultural fit—are the ones that are the most difficult to quantify and therefore the hardest to incorporate into an algorithm. That is partly the reason why traditional measures like education and experience continue to be widely used despite their tenuous connection to efficient matches.
So what, if anything, can matchmaking firms do to fix this problem? The answers may come from academia. Amy Kristof-Brown, an expert on person–environment fit who earned her PhD in organizational behavior and now teaches at the Tippie College of Business at the University of Iowa, has found that optimal fits based on company culture reduce turnover and create stable matches.30 She has advised RoundPegg, a firm that seeks to quantify the elusive concept of company culture by measuring a variety of indicators with surveys to get at the attitudes and beliefs of individual employees.31 In this way, companies’ cultures can be quantified and then matched to the specific attributes of each job seeker in order to create more stable matches.
Another example is a startup called Good.co, which specializes in improving labor market matches involving millenials—those born in the 1980s and 1990s—many of whom are now entering the labor force in large numbers.32 Good.co uses surveys to assess job seekers’ personalities and make recommendations about workplace fits based on qualities that have proven to yield successful matches. Like Good.co, dozens of other startups have launched in recent years with the goal of improving labor markets by using surveys and other data-driven tools to explore the characteristics that lead to enduring matches.
Matchmaking in the labor market is turning out to be a win-win-win-win business. Job seekers and employers both are better off from more stable matches. Matchmakers profit from improving existing markets or creating new ones from scratch when they’re missing. And society benefits from better matches in the workplace, which should lead to higher productivity, greater innovation and, over time, a higher standard of living. In this way, market designers and the matchmakers in the labor market who use their insights help achieve a fundamental goal of economics: to improve the general welfare and increase the wealth of nations.
Matchmaking and Online Dating
As we just discovered with matching in labor markets, hiring and dating have a lot in common. In fact, many market designers have for decades considered them to be so similar as to be synonymous.33 Like hiring workers, dating has problems associated with information asymmetries, long search processes, and complex criteria, so it is not surprising that the Internet has facilitated matchmaking in both markets. In addition, as you will see shortly, economists have developed a number of insights that have been applied to improve the online dating world and thus create more stable matches, sometimes in ways that differ from the solutions being implemented by matchmakers in the labor market.
Online dating sites offer two things that neither traditional matchmakers nor chance encounters at, say, bars, shopping malls, or workplaces entail. One is the vastly greater choice available to market participants online.34 Using the framework of market design outlined earlier, online dating sites provide much more thickness than traditional settings. The other potentially large benefit of online dating is that, as with matchmaking in the labor market, the companies in this space tend to claim they have scientifically proven ways of finding the perfect matches for their users.
It will take much time and study to determine which, if any, of these claims for scientific validity have merit. In the meantime, it’s instructive to see how economists have helped some of the online sites overcome one of their initial obstacles.
Unlike the markets for medical doctors, kidneys, and schools in which there were plenty of potential customers on both the demand and supply sides, the online dating markets struggle with attracting as many women as they do men. In particular, with little refereeing by the online sites, men would flood the inboxes of attractive w
omen, who in turn would ignore most of the messages because it is difficult and very time-consuming to separate spammers from good prospects.35 Women thus faced a needle-in-the-haystack problem. Or, to borrow a term from market design, many of these dating sites had encountered congestion problems, preventing markets from creating efficient and stable matches. So how was this problem solved?
One website, Cupid.com, hired Muriel Niederle, an expert market design economist from Stanford, as an adviser to the firm. In 2005, at the suggestion of Niederle and economist Dan Ariely (then at MIT, now at Duke), Cupid.com began to allocate electronic roses to its male members.36 The logic of this change was based on the fact that the marginal cost of sending another message to a new member was essentially zero, which encouraged too many men to send messages to women’s inboxes. What if, the economists asked, communications were made more costly? If the number of messages people could send were limited—in this case to two electronic roses per month—then men would have an incentive to reveal their true preferences and signal their interest to the women that they wanted to impress. In short, the economists argued that the scarcity of roses would motivate suitors to be selective, and so members who weren’t serious would essentially be weeded out of the process.
According to Eric Straus, CEO of Cupid.com, the results were “a wonderful thing.” The electronic roses increased a suitor’s chances of getting a reply by about 35 percent.37 The majority of online dating sites have followed suit and implemented various mechanisms to increase the cost of communication between members. In fact, one site, whatsyourprice.com, has literally put a price tag on not just the cost of sending a message to another member, but on the dates themselves. Men bid for dates with women, and if a proposed price is accepted for a date request, the suitor is supposed to pay the woman the amount they agreed upon when they meet on their first date.
