Top Trading Cycles (TTC) Mechanism

• Top Trading Cycles (TTC) is a mechanism in one-sided matching that reallocates indivisible objects using cycle detection based on agents’ top remaining choices and initial endowments.
• It extends to complex settings like school choice and multi-center allocation, incorporating capacities and priorities to achieve Pareto efficiency and strategy-proof outcomes.
• Recent research explores TTC’s computational methods, welfare implications, stability under farsighted behavior, and extensions to account for fairness and networked environments.

Searching arXiv for recent and foundational TTC papers to support the encyclopedia entry. arXiv search query: "Top Trading Cycles housing market school choice characterization large markets" Top Trading Cycles (TTC) is a canonical mechanism for reallocating indivisible objects without money in one-sided matching environments with initial entitlements. In its classical housing-market form, each agent points to her most preferred remaining object, each object points to its current owner, at least one directed cycle exists, and every agent in a cycle receives the object she points to before those agents and objects are removed and the procedure repeats. In priority-based school choice, the object-side step is modified so that each school points to its highest-priority remaining student; the resulting student-optimal top trading cycles mechanism (TTCM) is Pareto efficient and group strategy-proof, but not stable in general (Gogulapati et al., 13 Jul 2025, Chen et al., 2021).

1. Classical mechanism and formal environments

In the standard Shapley–Scarf housing market, there are as many agents as objects, each agent initially owns one object, and preferences are strict. TTC proceeds round by round on a directed graph in which every remaining agent points to her top remaining object and every remaining object points to its owner. Because every node has outdegree $1$, at least one cycle exists. Each cycle is executed simultaneously, the involved agents receive the objects they point to, and the reduced market is processed in the same way. In the strict-preference housing market, this procedure produces a unique allocation and is identified in several papers as the core allocation (Chen et al., 2021, Sandholtz et al., 23 Mar 2025).

The same cycle-removal logic extends to school choice, but with capacities and priorities. In the model studied in the recent axiomatic literature, there is a finite set of students $I$, a finite set of schools $O$, each school $a$ has quota $q_a$, each student has strict preferences over schools and the null school, and each school has a strict priority order over students. The student-proposing TTCM is generated by the sequence: each student points to her favorite remaining school, each school points to the highest-priority remaining student, cycles are assigned, capacities are updated, and the process repeats until all students are removed (Chen et al., 2021).

A common misconception is that TTC is a single model-specific algorithm. The recent literature instead treats TTC as a family of closely related cycle-exchange procedures defined over distinct environments: housing markets with endowments, school choice with quotas and priorities, bundle reallocation, multi-center allocation, and capacity-aware reassignment. What remains invariant is the elimination of trading cycles induced by locally optimal choices and institutional ownership or priority relations (Coreno et al., 2024, Cheng et al., 26 Sep 2025, Satpathy et al., 31 Jan 2026).

2. Efficiency, incentives, and axiomatic characterizations

The classical normative profile of TTC in housing markets is Pareto efficiency, individual rationality, and strategy-proofness. Recent probabilistic work restates Ma’s deterministic theorem as follows: if endowments are deterministic, then TTC is the unique deterministic assignment rule satisfying Pareto-efficiency, individual rationality, and strategy-proofness; moreover, a probabilistic assignment rule is SD-efficient, SD-individually rational, and SD-top-strategy-proof if and only if it coincides with the TTC rule (Gogulapati et al., 13 Jul 2025). A further generalization shows that on any Free Pair at the Top (FPT) domain, TTC is the unique probabilistic assignment rule satisfying SD-Pareto efficiency, SD-individual rationality, and SD-top-strategy-proofness, and that on any Free Triple at the Top (FTT) domain the same conclusion holds when SD-Pareto efficiency is weakened to SD-pair efficiency (Donthu et al., 6 Nov 2025).

Restricted-domain characterization has become a major theme. In object reallocation problems, TTC remains uniquely characterized by individual rationality, pair efficiency, and strategyproofness on domains satisfying the “top-two condition”; the paper further shows necessity in the special case of three objects (Goel et al., 26 Jan 2025). In housing markets with limited externalities and demand lexicographic preferences, TTC is characterized by individual rationality, pair efficiency, and strategy-proofness, and equivalent characterizations follow if pair efficiency is strengthened to Pareto efficiency or pairwise stability. The same paper proves an impossibility result: if the domain includes both demand lexicographic and supply lexicographic preferences and $|N|>3$, then no rule satisfies individual rationality, pair efficiency, and strategy-proofness (Klaus, 2024).

The recent literature also weakens incentive requirements while retaining TTC uniqueness. One paper introduces truncation-invariance and proves three characterizations: individual rationality, pair-efficiency, and truncation-invariance; individual rationality, Pareto efficiency, and truncation-invariance; and individual rationality, endowments-swapping-proofness, and truncation-invariance all characterize TTC (Chen et al., 2021). In multi-object reallocation with bundle endowments, generalized TTC is uniquely characterized on the lexicographic domain by balancedness, Pareto efficiency, the weak endowment lower bound, and either truncation-proofness or drop strategy-proofness; in the Shapley–Scarf special case this reduces to Pareto efficiency, individual rationality, and truncation-proofness (Coreno et al., 2024).

These results collectively suggest that TTC is not merely one rule with attractive properties. Across deterministic, probabilistic, and restricted-domain formulations, it repeatedly appears as the unique rule surviving combinations of participation, local or global efficiency, and weakened but still structured nonmanipulability.

3. School choice, mutual best groups, and farsighted stability

In school choice, TTCM is re-described through the notion of a mutual best group (MBG). For a subset of students $I'\subseteq I$, the set $B(\epsilon,I')$ consists of schools that are the top choice of at least one student in $I'$, and $T(\epsilon,O')$ is the set of students who are highest priority at some school in $I$0. A subset $I$1 is a mutual best group if it is the largest subset satisfying

$I$2

Operationally, all students in cycles of the TTC graph form the mutual best group, and repeated removal generates a sequence $I$3 that re-describes TTCM as a sequentially prioritized mechanism with endogenous rather than exogenous priority over groups (Chen et al., 2021).

This MBG perspective supports new axiomatic foundations. The paper introduces MBG-quota-rationality, a weakening of stability, and MBG-robust efficiency, a weakening of robust efficiency based on MBG-collusion-proofness. Its central theorem states that a mechanism satisfies MBG-quota-rationality and MBG-robust efficiency if and only if it is TTCM. The same analysis emphasizes a point often obscured in policy discussions: TTCM is Pareto efficient and group strategy-proof, but unlike the student-optimal stable mechanism it is not stable in general (Chen et al., 2021).

A separate line of work studies TTC under farsighted behavior. In school choice with farsighted students, the singleton set consisting of the TTC outcome is a farsighted stable set, whereas the Deferred Acceptance outcome may fail to belong to any farsighted stable set (Atay et al., 2022). Under bounded farsightedness, looking forward three steps ahead is already sufficient for stabilizing the TTC matching: for $I$4, $I$5 is a horizon-$I$6 farsighted stable set (Atay et al., 2022). In the related priority-based matching model with limited farsightedness, if $I$7 denotes the size of the largest TTC cycle ever generated, then $I$8 is a horizon-$I$9 vNM stable set for $O$0, with a tighter sufficient bound $O$1 noted as requiring more coordination (Atay et al., 2022).

These papers shift the interpretation of TTC in school choice. Under myopic evaluation, TTC trades stability for efficiency; under sufficiently farsighted dynamics, TTC becomes stable in a behavioral sense that DA may fail to satisfy.

4. Computation, inversion, and manipulation

Recent computational work on TTC addresses both forward execution and inverse structure. “Fast TTC Computation” does not alter TTC’s economic logic, but reinterprets it through a preference-weighted matrix $O$2, column normalization, and dominant singular-vector extraction. The proposed procedure maps ordinal rankings into linearly decreasing weights, normalizes columns so that each sums to $O$3, and interprets the first singular vector as revealing the TTC elimination order. The paper claims that the method retains Pareto-efficiency, individual rationality, and strategy-proofness because it reproduces the classical TTC allocation, and it presents the speed claim as $O$4 in an engineering sense linked to constant-time eigenvector inference, not as a standard dimension-independent complexity claim in the classical theoretical sense (Aldridge, 2024).

A different computational contribution studies the non-injectivity of TTC. In one-sided matching, many distinct initial endowments can converge to the same Pareto-optimal outcome, so focusing on a single TTC output obscures the structure of the entire Pareto frontier. The Inverse Top Trading Cycles Enumeration Algorithm (ITEA) reconstructs all initial allocations that map to a given TTC outcome and, by iterating over equivalence classes, computes the full set of Pareto-optimal allocations. The paper proves soundness and completeness of the inverse traversal and gives the complexity bound

$O$5

while emphasizing that ITEA reduces the number of forward TTC invocations from $O$6 to $O$7 (Dodda et al., 23 Apr 2026).

Manipulation has also been revisited from a parameterized-complexity viewpoint. In generalized housing markets with bundle endowments, the beneficial-misreport problem for TTC becomes polynomial-time solvable when the manipulator’s endowment size $O$8 is fixed; the explicit search algorithm has running time $O$9. At the same time, the parameterized problem is $a$0-hard, even on the lexicographic domain, so it is unlikely to be fixed-parameter tractable (Phan et al., 2018).

Together, these results show that TTC’s algorithmic profile is more intricate than the standard cycle-detection narrative suggests. There are now forward shortcuts, inverse enumeration methods, and refined complexity analyses that separate practical tractability from worst-case or parameterized hardness.

5. Welfare, fairness, and large-market critiques

Several papers place TTC under direct welfare and fairness scrutiny. In school choice, TTC is treated as the canonical strategy-proof and Pareto optimal mechanism, but asymptotic random-market analysis reveals substantial costs relative to the rank-minimizing mechanism (RM). Under iid uniform random preferences and priorities in large one-to-one markets,

$a$1

so TTC’s expected average rank grows like $a$2. For the worst-off student,

$a$3

and simulations are reported to suggest a limit around $a$4. For justified envy,

$a$5

In Budapest secondary school admissions data, TTC yields mean rank $a$6, maximum rank $a$7, and share of students with justified envy $a$8, compared with RM’s $a$9, $q_a$0, and $q_a$1, respectively (Ortega et al., 2022).

A more recent large-market theorem argues that TTC’s priority advantage vanishes asymptotically. In a balanced one-to-one market with iid uniform preferences and priorities, the joint distribution of preference ranks and realized priority ranks under TTC converges to the same limit as under Random Serial Dictatorship (RSD). The mechanism’s superiority over RSD is traced to short cycles, especially $q_a$2-cycles, but the fraction of assignments made through short cycles vanishes in large markets. The paper shows that the ratio of expected justified-envy incidences to expected envy incidences converges to $q_a$3 under TTC, matching RSD, and that the expected fraction of agents who experience at least one justified envy converges to $q_a$4 under both mechanisms (Che et al., 12 Jul 2026).

Long-run equilibrium analysis extends the fairness critique beyond the assignment stage. In a model where families first choose neighborhoods and then participate in school assignment, TTC increases the option value of living in desirable zones because local residents can often secure out-of-zone schools through trading cycles. The paper proves that neighborhood segregation is ranked

$q_a$5

and that housing prices in overdemanded zones rise in the same order; it also states that TTC is generally the most segregating mechanism at the school level in the long run (Artemov et al., 13 Nov 2025).

These studies do not refute TTC’s classical efficiency claims. They instead sharpen the boundary of those claims: Pareto efficiency and strategy-proofness do not imply favorable rank distributions, strong priority respect in large random markets, or benign long-run distributional effects.

6. Extensions, domain boundaries, and applied variants

A broad recent literature generalizes TTC while also identifying its failure modes. In markets with objective indifferences, where agents are always and only indifferent between copies of the same object, TTC with fixed tie-breaking maintains Pareto efficiency, group strategy-proofness, and core selection; the paper further characterizes objective indifferences as the most general setting in which fixed tie-breaking TTC preserves these properties (Sandholtz et al., 23 Mar 2025). In probabilistic exchange with fractional endowments and weak preferences, Fractional Top Trading Cycle (FTTC) extends TTC’s trading logic through balanced fractional exchange, preserves individual rationality and SD-efficiency on the full preference domain, and yields an extension of Probabilistic Serial in the house allocation model (Yu et al., 2020).

Other papers modify TTC to fit new institutional structures. In a networked housing market with invitations, an impossibility result shows that no mechanism can achieve IR, SP, and PE simultaneously without restricting the preference domain, and the paper identifies competition between inviters and invitees as the reason TTC “cannot work properly” in that environment. Its proposed Top Trading Cycle with Diffusion (TTCD) is IR, SP, and core-for-paths, but not Pareto efficient (Zhang et al., 2023). In multi-center allocation with center-specific priorities, a natural extension of TTC is characterized either by strategy-proofness, pair efficiency, internal fairness, and external fairness, or by strategy-proofness and procedural fairness alone; the same paper proves that the ultimate core is the singleton $q_a$6 (Cheng et al., 26 Sep 2025).

Applied systems work has also adapted TTC beyond classical market design. ReACT-TTC is a capacity-aware post-deviation reassignment framework for shared cyber-physical systems. It extends TTC to many-to-one capacities and unassigned-resource conditions through co-ownership, virtual owners for empty slots, and capacity-aware cycle detection. The paper proves termination and preservation of Pareto efficiency, individual rationality, and strategy-proofness, while using a Prospect-Theoretic satisfaction model to prioritize overlapping cycles and path resolutions (Satpathy et al., 31 Jan 2026).

A final boundary result concerns what TTC does not provide by itself. In one-sided matching, multiple Pareto-optimal allocations may exist, and many distinct initial endowments can converge to the same TTC outcome, so a single TTC allocation does not reveal the full Pareto frontier (Dodda et al., 23 Apr 2026). This suggests that TTC is best understood not as an all-purpose resolution of welfare and fairness tradeoffs, but as a specific cycle-based rule whose normative force depends sharply on domain restrictions, institutional primitives, and the comparison criterion being used.