d4-journals-jet-2013-0090

Theorem 1 gives an explicit representation of a computationally meaningful equilibrium. But its inputs and outputs are already type masses, and normalizing the population vector reproduces the proposed distributional question exactly; this is a collision with the paper's existing model, not a discrete-to-continuous mirror.

The objection that survived

The paper already solves for real-valued type masses and fractional marriage masses; scaling and normalizing its population vector leaves the same equilibrium equations, while the proposed question adds no distinct finite-to-continuum computation.

What the mirror covers

The proposed computation restates Theorem 1's inverse equilibrium problem and leaves Theorem 2, Corollary 3, and the transfer comparative statics unexamined.

Open questions for a prover

The case FOR (proponent)

The strongest positive mirror is Choo–Siow Mass-Equilibrium Computation. Consider a large marriage market with a finite support of male and female types. A type records the observable characteristics that determine systematic gains from each possible match, together with the common distribution of that type’s idiosyncratic preferences. Agents of the same type are interchangeable for the expected-equilibrium question. Their mass is their share of the market; the number of agents can be very large while the number of distinct types stays comparatively small—for example, a regional or national market with millions of people but a much shorter list of demographic profiles.

Given rational type masses \(m_i,f_j>0\) and nonnegative rational gains \(\Pi_{ij}\), the problem is to compute the equilibrium masses of marriages and singles. A solution consists of nonnegative masses \(\mu_{ij}\), \(\mu_{i0}\), and \(\mu_{0j}\) satisfying
\[ \mu_{i0}+\sum_j\mu_{ij}=m_i,\qquad \mu_{0j}+\sum_i\mu_{ij}=f_j,\qquad \mu_{ij}=\Pi_{ij}\sqrt{\mu_{i0}\mu_{0j}}. \]
The output is the unique equilibrium mass distribution, or its approximation to a requested precision. There is no planner’s objective: the object being computed is the market-clearing equilibrium. Equivalently, compute its positive single masses \(\beta_i,\beta_{I+j}\), where \(\mu_{i0}=\beta_i^2\), \(\mu_{0j}=\beta_{I+j}^2\), and \(\mu_{ij}=\Pi_{ij}\beta_i\beta_{I+j}\).

My anchor is Theorem 1 (Existence, uniqueness, and explicit representation of a real positive solution), proved in this paper. It says that for every nonnegative gains matrix \(\Pi\) and strictly positive population vector \(\nu=[m\mid f]\), there is exactly one positive vector \(\beta\) solving the paper’s quadratic system (13). Writing \(b_k=\log\beta_k\), the solution is \(b=DH^*(\nu)\), where \(H^*\) is the Legendre transform of
\[ H(b)=\frac12\sum_k e^{2b_k}+\sum_{i,j}\Pi_{ij}e^{b_i+b_{I+j}}. \]
The resulting marriage masses are \(\mu_{ij}=\Pi_{ij}\beta_i\beta_{I+j}\). Existence had been addressed in earlier work cited by the authors; this paper proves uniqueness and gives the variational representation. For the mirror, the expected direction is Class A: the equilibrium is the minimizer of the strictly convex function \(H(b)-\langle\nu,b\rangle\). Turning that into a polynomial-time bit-complexity result for rational input and a requested output precision is a further algorithmic question; the theorem itself does not establish such a bound.

This mirror covers the paper’s inverse equilibrium problem, not its full set of comparative statics or its discussion of transfers. The authors should recognize it: the type masses, gains, market-clearing conditions, and output are precisely those of their model. The main qualification is also the main weakness of the positive case. The paper already allows real-valued population parameters and fractional marriage masses, and its equilibrium is defined by balancing expected demand and supply. So the continuous mass formulation is extremely close to the paper’s own model. The case for a mirror rests on treating computation and complexity as a distinct question, and on distinguishing an exact nonatomic allocation from finite-market realized counts. The paper does not analyze those realized counts or give finite-size error bounds. If a referee regards the expected mass equilibrium itself as already the continuous-population problem, this mirror risks being a restatement rather than a new one.

The case AGAINST (opponent, writing after the proponent)

Theorem 1 is the only anchor, and its strongest vulnerability is that the population axis is already expressed in masses. The theorem allows arbitrary positive real population entries \(\nu\); its singles \(\beta_k^2\) and marriage quantities \(\mu_{ij}\) are real-valued. Scaling \(\nu\) by \(a>0\) scales \(\beta\) by \(\sqrt a\) and every marriage and single mass by \(a\). Normalizing the population vector therefore gives the finite-support type distribution and mass equilibrium proposed as the mirror. The expected-supply and expected-demand equations are already the paper’s object; the authors even note that their equilibrium need not occur in any particular realization of the random preferences. The proposed question repeats that mass-equilibrium problem in normalized units. This is a collision in the model, not an objection based on prior high-multiplicity work.

A stronger alternative would start with named individuals, draw their random preferences, and ask for an actual one-to-one matching or a finite-market approximation guarantee. That requires specifying a realization-level matching process or mechanism the paper does not define. If the alternative keeps expected flows, it returns to equations (9)–(13); if it adds realized assignments, it poses a different stochastic matching problem. The law-of-large-numbers observation around equation (3) also already points from individual random choices to aggregate type frequencies, though a sharper finite-size bound could be a separate result.

There is a real weakness in this negative case: Theorem 1 does not give bit-complexity bounds or a precision-controlled algorithm for rational inputs. Those could support a worthwhile algorithmic study. But that would analyze computation for a model already formulated in population masses; it would not supply a new continuous mirror of a discrete problem in the paper. The case against is strongest on that distinction, and weaker if “continuous mirror” is taken broadly enough to include computational analysis of an already mass-based model.

The adversarial triple: the proponent anchors on up to three named results; the opponent sees that case and must defeat every anchor; the judge decides which case convinced it. These are the pipeline’s own outputs, generated by tools/triple_run.py — no human edited them. The paper’s own text is not reproduced here beyond the quoted statement above.