Interior points in the core of two-sided matching markets (Q1110430)

A two-sided matching market is a specified list of data \(M:=(P,Q,f,g,r,s)\), where P and q are disjoint finite player sets, a member of P (of Q, resp.) being either matched with a member of Q (of P, resp.) or self-matched (i.e., unmatched), \(f:=\{f_{pq}\}_{p\in P,q\in Q}\) is a family of functions, \(f_{pq}:\) \({\mathfrak R}\to {\mathfrak R}\), \(f_{pq}(u)\) being interpreted as the monetary amount player p must receive in order to achieve utility u if he is matched with player q, \(g:=\{g_{pq}\}_{p\in P,q\in Q}\) is analogously defined and interpreted for players \(q\in Q\), \(r:=\{r_ p\}_{p\in P}\) is a set of numbers, \(r_ p\in {\mathfrak R}\), \(r_ p\) being interpreted as the utility of being unmatched for player p, and \(s:=\{s_ q\}_{q\in Q}\) is analogously defined and interpreted for players \(q\in Q\). A matching \(\mu\) is a bijection of \(P\cup Q\) onto itself such that (1) \(\mu\) \(\circ \mu\) \(=\) identity (i.e., p is matched with q iff q is matched with p), and (2) [\(\mu\) (p)\(\neq p\in P\) (\(\mu\) (q)\(\neq q\in Q\), resp.)] implies [\(\mu\) (p)\(\in Q\) (\(\mu\) (q)\(\in P\), resp.)] (i.e., if \(p\in P\) is not unmatched, then he is matched with somebody in Q). A feasible payoff (u,v) of M is a ({\#}P\(\cup Q)\)-tuple of utility levels, such that there exists a matching \(\mu\) with the properties, \[ u_ p=r_ p\quad (v_ q=s_ q,\quad resp.)\quad if\quad p\quad (q,\quad resp.)\quad is\quad self-matched;\quad and \] \[ f_{pq}(u_ p)+g_{pq}(v_ q)\leq 0\quad if\quad \mu (p)=q. \] The second condition requires that the compensation to the two members of a matched pair is in fact a transfer. A feasible payoff (u,v) is stable, if \[ u_ p\geq r_ p,\quad v_ q\geq s_ q;\quad and\quad f_{pq}(u_ p)+g_{pq}(v_ q)\geq 0\quad for\quad all\quad (p,q)\in P\times Q. \] The first condition is the individual rationality. The threats of \(p\in P\) and \(q\in Q\) at a stable payoff \(x:=(u,v)\) with the associated matching \(\mu\) are \[ t_ p(x,\mu)\quad:=\quad \max \{f^{-1}_{pq}(v_ q),\quad r_ p| \quad q\neq \mu (p)\}, \] \[ t_ q(x,\mu)\quad:=\quad \max \{-g^{- 1}_{pq}(u_ p),\quad s_ q| \quad p\neq \mu (q)\}. \] A symmetrically pairwise-bargained payoff (SPB) is a stable payoff \(x=(u,v)\) with the associated matching \(\mu\) such that for each matched pair (p,q), the utility allocation \((u_ p,v_ q)\) is obtained by dividing their monetary surplus over \((f_{pq}(t_ p(x,\mu))\), \(g_{pq}(t_ q(x,\mu)).\) Theorem. An SPB exists. Define a partial order \(\geq_ P\) on \({\mathfrak R}^{P\cup Q}\) by \((u,v)\geq_ P(u',v')\) iff \(u_ p\geq u_ p'\) for all \(p\in P.\) Theorem. The set of SPB's forms a complete lattice under the partial order \(\geq_ P\) whose sup is P-optimal and whose inf is Q-optimal. These results are tied together by the lattice structure of the set of stable outcomes.