Publications des chercheurs de PSE

We study a large market model of dynamic matching with no monetary transfers and a continuum of agents who have to be assigned items at each date. When the social planner can only elicit ordinal agents' preferences, we prove that under a mild regularity assumption, incentive compatible and ordinally efficient allocation rules coincide with spot mechanisms. The latter specify “virtual prices” for items at each date and, for each agent, randomly select a budget of virtual money at the beginning of time. When the social planner can elicit cardinal preferences, we prove that under a similar regularity assumption, incentive compatible and Pareto efficient mechanisms coincide with spot menu of random budgets mechanisms. These are similar to spot mechanisms except that, at the beginning of time, each agent chooses within a menu, a distribution over budget of virtual money.

Voters have strong incentives to increase their influence by trading votes, acquiring others' votes when preferences are strong in exchange for giving votes away when preferences are weak. But is vote trading welfare-improving or welfare-decreasing? For a practice long believed to be central to collective decisions, the lack of a clear answer is surprising. We review the theoretical literature and, when available, its related experimental tests. We begin with the analysis of logrolling – the exchange of votes for votes. We then focus on vote markets, where votes can be traded against a numeraire. We conclude with procedures allowing voters to shift votes across decisions – to trade votes with oneself only. We find that vote trading and vote markets are typically inefficient; more encouraging results are obtained by allowing voters to allocate votes across decisions.

We propose novel approaches to estimating student preferences with data from matching mechanisms, especially the Gale-Shapley deferred acceptance. Even if the mechanism is strategy-proof, assuming that students truthfully rank schools in applications may be restrictive. We show that when students are ranked strictly by some ex ante known priority index (e.g., test scores), stability is a plausible and weaker assumption, implying that every student is matched with her favorite school/college among those she qualifies for ex post. The methods are illustrated in simulations and applied to school choice in Paris. We discuss when each approach is more appropriate in real-life settings.

This dissertation introduces bounded rationality on matching markets,by way of imperfect cognition in social learning. On a two-sided agent-agent one-to-one matching market with Beckerian match utilities though forbidding transfers, a Live-Polarised-Unidimensional-Valuation (LPUV) rational agent m observes the surplus that a potential partner f is currently generating with her own match and uses it as an estimate of the surplus they (m and f) would jointly generate. The agent plugs this incorrect estimate of the surplus into the correct splitting rule, hence a coarse belief on match utilities. I compare the implicitly defined LPUV-stability to the usual(Gale-Shapley) stability concept, under odd or specific splitting rules, exogenous or endogenous surpluses. Quite remarkably, there exist splitting systems for which LPUV-stability is robust to the specification of surpluses. In addition, LPUV-stable matchings are disproportionately assortative with respects to GS-stable ones.

We document quasi-experimental evidence against the common assumption in the matching literature that agents have full information on their own preferences. In Germany’s university admissions, the first stages of the Gale-Shapley algorithm are implemented in real time, allowing for multiple offers per student. We demonstrate that nonexploding early offers are accepted more often than later offers, despite not being more desirable. These results, together with survey evidence and a theoretical model, are consistent with students’ costly discovery of preferences. A novel dynamic multioffer mechanism that batches early offers improves matching efficiency by informing students of offer availability before preference discovery.

Ce chapitre propose un tour d’horizon des enjeux théoriques et pratiques de l’utilisation des algorithmes dans la régulation des inscriptions scolaires et universitaires. Il présente le principe et les enjeux de l’appariement ainsi que les principaux algorithmes qui ont été proposés dans la littérature académique pour affecter les élèves et les étudiants aux établissements d’enseignement. Une attention particulière est portée à l’algorithme d’acceptation différée de Gale et Shapley (1962), qui s’est imposé comme le plus à même de résoudre les arbitrages complexes auxquels sont confrontées les procédures d’affectation.

Les procédures d’affectation centralisées et automatisées ont été déployées dans le système éducatif français au début des années 2000 pour faciliter la gestion de deux transitions majeures dans le parcours des élèves : le passage entre le collège et le lycée, d’une part, l’entrée dans l’enseignement supérieur, d’autre part. À la lumière de la théorie de l’appariement et des travaux empiriques réalisés dans son sillage, ce chapitre propose une analyse critique des procédures qui sont utilisées en France pour réguler les inscriptions dans l’enseignement secondaire (procédure Affelnet au lycée et expérimentation des secteurs multi-collèges) et dans l’enseignement supérieur (plateformes APB et Parcoursup).