I presented this paper by Abdulkadiroglu and Sonmez for a microeconomics course, and it's one of my favorites because it takes what sounds like abstract math and shows how it directly changed real policy.
The Problem
How do you assign people to things (dorm rooms, schools, kidneys) when everyone has different preferences and some people have existing rights? You can't use prices, nobody's auctioning off kidneys, and you can't just let people grab what they want. This is the matching problem, and it comes up in way more places than you'd expect.
University housing, public school enrollment, and organ donation are systems that seem totally unrelated, but they can all be solved with the same small set of models.
The Two Core Models
Two-Sided Matching (Gale-Shapley): Both sides rank each other. Hospitals rank residents, residents rank hospitals. The Deferred Acceptance algorithm finds a stable match where no pair would rather ditch their assignment and match with each other instead. This is how the National Resident Matching Program works.
One-Sided Matching (Shapley-Scarf): Only one side cares. You have a dorm room, I have a dorm room, the rooms don't have opinions. The Top Trading Cycles mechanism lets everyone trade until nobody can do better. It's also strategy-proof, so you can never game it by lying about your preferences.
Where It Gets Messy
Real life doesn't fit neatly into either model. Schools don't have "preferences," they have priority rules like siblings and proximity. Not everyone starts with a house (what about new students?). Kidney donors don't pick recipients because compatibility is biological. The paper's whole contribution is showing how to adapt those foundational tools to handle all of this.
The Applications
Housing with existing tenants. Some students already have rooms they have the right to keep. New students need rooms too. A modified TTC handles both: existing tenants point to their top choice, occupied rooms point back to their tenant, and when trading cycles form, those swaps happen. Nobody's forced to downgrade, and you still can't game it.
School choice. The old systems in Boston and NYC actually punished families for being honest. Listing your true first choice could land you at your worst option. The fix was applying Deferred Acceptance or TTC. Neither is perfect since you can't satisfy every desirable property at once, but both are way better than what they replaced. Both cities ended up adopting these.
Kidney exchange. When a donor-patient pair is biologically incompatible, that donor might match a different patient whose donor matches someone else. These form trading cycles, just like housing. Altruistic donors with no patient attached can kick off long chains. The mechanism's incentives make sure hospitals don't benefit from hiding pairs, which means more transplants actually happen.
The Tradeoffs
Every application runs into the same tensions: stability vs. strategy-proofness, priority respect vs. efficiency, chain length vs. simplicity. No mechanism gets everything. But the point is making those tradeoffs explicit so policymakers can choose deliberately instead of stumbling into bad designs.
Why I Like This
Matching theory started as pure math in 1962 and now directly determines how cities assign students to schools, how hospitals match with residents, and how kidney exchanges save lives. It's one of the clearest examples of economic theory actually making it out of the classroom and into policy that affects millions of people.