Ashutosh Thakur is using matching and game theory to study institutional design in some unusual places.
Thakur, a doctoral candidate in political economics at Stanford GSB, is trying to determine if systems designed for matching, say, medical students with residency posts can also be used to improve the effectiveness of public administration. He is specifically interested in the allocation of civil servants to diverse regions across India and in the party-specific procedures used in the U.S. Senate to assign politicians to committees.
“These projects take the theoretical tools and apply them to real-world problems and systems in practice,” says the Princeton graduate. “I’m hoping to improve government administration and development, trying to not just write academic papers but also use the findings to inform policymaking.”
Thakur is also an expert chess player and an accomplished musician — pastimes that connect in their own way to his work.
You’d better explain the concept of matching theory before we go any further.
When we think of economics we generally think of allocations in markets being determined by prices equating supply and demand. That’s a classical economic notion. Matching theory deals with markets that don’t have prices. For example, what do you do if you want to allocate civil servants to different states in India? Everyone’s price is the same, because everyone gets the same government wage. So if you don’t have prices, then how do you figure out how to make allocations in the market? My co-advisor [the economist and Nobel laureate] Alvin Roth has used matching theory to study and design systems for assigning kids to schools, medical students to hospitals for residency, and donors to recipients in kidney exchanges.
Why apply matching theory to government systems?
When people think of politics, they think of voting. But it’s not always about voting. A lot of things in government are allocation or assignment problems. I wondered if we could use matching theory to learn more about what’s happening and how to design better systems.
India changed its civil service system in 2008, and you found that change had some widespread ripple effects. What was happening?
To understand where this is all coming from, you have to understand the history of the institution itself. These civil services originated during British colonial rule, which developed a very efficient method for collecting taxes and implementing social programs. The bureaucrats in charge were in it for life and among the most highly qualified people in the country. India wanted to maintain this bureaucracy, but it also wanted the system to be part of a national integration effort — to move people around the country. It’s a unique civil service in which people are shuffled around and given lifelong assignments to a particular state.
And the new system created some unforeseen problems, right?
The old system didn’t take into account civil servants’ preferences about where they wanted to work. But the new system takes those preferences very seriously. You get to pick your first choice for assignment, if available, and then your second choice, and so on. And everyone chooses in order of their exam rank. So basically, those with the highest scores on the civil service exam tend to end up in the nicer, highly sought-after areas. Those with lower scores end up in underdeveloped regions, or areas with foreign conflict or internal conflict — places where people typically don’t want to go. It systematically sends lower quality civil servants to certain regions, which, I find, then do worse in terms of tax revenue collection, bureaucratic efficiency, and so forth. So I asked: Why is that happening? What design portion of the system is doing that? And how does it affect outcomes?
So matching theory takes into account factors beyond individuals’ preferences?
You’re still taking into account where the civil servants want to go, but there are also other factors such as affirmative action and quality balances. So there’s a bunch of preferences and constraints, and you have to design an algorithm to make allocations in a given market environment. If a system is designed in a way that enables people to game it — say, if they are incentivized to misreport their preferences and be not truthful — then you might want to change the design so that becomes impossible.
What sort of reaction did you get from the Indian government?
Reaction is split in the Indian Administrative Service among the civil servants being assigned and those in charge of designing the system of assignment. But civil servants assigned under both the old system and new system find it troubling that they were getting this division in the country in terms of bureaucratic quality. One of the things that’s happening is that people tend to stay closer to their home state now so they don’t have to learn a new language or culture. That defeats the national integration purpose of the bureaucracy. I’m hoping the government will be receptive to understanding what the systems are doing and the effects they have on outcomes. But working with the bureaucracy is challenging. It’s a long process, for sure.
What have you concluded about the committee assignment processes in the U.S. Senate?
It’s interesting. The two major parties have very different systems for committee assignment. On the Democratic side, the system incentivizes people to misreport their preferences. For example, say I really want to be on Appropriations because it’s a very powerful committee. But it’s also very popular and hard to get, given this competition. So, based on how that system currently operates, I might have an incentive to not truthfully report that it’s my first choice. Because if I fail to get my first choice in a tie-breaker, by the time that’s decided my second and third choices may already be filled up and I might get assigned to a committee I like much less.
So you pretend that, say, your second preference is your first choice, because it’s less popular and thus easier to get, and that works as a hedge against potentially getting a much worse assignment.
Yes. This is the type of gaming that goes on. Democratic senators strategically misreport their preferences to get a better assignment. And where there’s more demand than supply, it gives party leaders, who make the tie-breaking decisions, a lot of power. They can use committee assignments to reward certain people or punish those who don’t vote the party line.
And the Republican system?
It’s ordered by seniority, and it’s very formulaic. The party doesn’t really have any say in it. You have an incentive to truthfully report your preference. They use a mechanism called serial dictatorship, and because it’s based on seniority, it’s strategy-proof. So you have one system where people are truthfully revealing their preference and another where there’s all types of gaming going on. These committee assignments are at the heart of political decision making, so it’s important for people doing empirical research related to politics and Congress to understand how the system really works.
You were co-president of Princeton’s varsity chess team and have a national expert rating. What has the game taught you about institutional design and problem-solving?
The funny thing about chess is that it tells you a lot about strategic thinking but doesn’t use much game theory at all. It’s very complex and requires a lot of calculation — on what we call game trees — but at the end of the day it’s a computational exercise.
Your other hobby is playing tabla (Indian drums). Is music connected in any way to the complicated problems you’re trying to solve?
Any creative field is a mixture of art and science, right? Even with chess, when you see an elegant pattern that ends the game or wins you a piece, it’s not just science and calculation but visually it has some artistic element to it. Indian music is very similar. It’s very complex. You’re trying to improvise within a framework that has to be followed. So there’s this underlying scientific structure, but inside of that is the artistic part of creating something different and new.
Have you had any memorable Stanford GSB experiences or encounters?
One of the primary reasons the Stanford market design group is so tightly knit is due to Alvin Roth’s efforts in bringing the group together. He institutionalized this 45-minute coffee every Wednesday morning. It’s an informal, interdisciplinary gathering of faculty and students from the economics department, business school, engineering, and computer science, along with occasional guests and practitioners who might be visiting the area. It’s just a bunch of people getting together to talk about things they find interesting and what they’re working on. It’s been great at helping to build a community across all these different disciplines and getting exposure to a broad range of topics and applications.