| paper | d4-journals-mss-2019-0051 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
other→NP-hard discrete result → its continuous mirror
Answer: NP-hard and undecidable, even for \(m=\tau=1\), fixed logarithmic utilities, and constant program-encoded market coefficients. · discrete side: other · adjudication: proved
full proof & adversarial review →
Theorem 4.4
Given a complete-market, finite-horizon environment, total wealth \(x\), a finite support \(T\) of client types with rational masses \(\mu_t>0\) summing to \(1\), type-specific utility and discount parameters satisfying the paper's assumptions, and a common terminal-wealth utility, find the optimal per-client allocations \(a_t\ge 0\) and common terminal allocation \(a_3\ge 0\), along with their admissible consumption and investment plans, maximizing \(\sum_{t\in T}\mu_t V_t(a_t)+V_3(a_3)\) subject to \(\sum_{t\in T}\mu_t a_t+a_3=x\).
The society is a distribution \(\mu\) over complete client preference types \(t\in T\), each with per-client allocation \(a_t\) and value function \(V_t\); the fund also chooses a shared terminal-wealth allocation \(a_3\) with value \(V_3\). The decision variables are these allocations and the corresponding admissible plans, with weighted welfare \(\sum_t\mu_tV_t(a_t)+V_3(a_3)\) and total resource use \(\sum_t\mu_ta_t+a_3=x\). Under the paper's assumptions, the sharing rule uses a common marginal value \(y\), with \(a_t=H_t(y)\), \(a_3=H_3(y)\), and \(\sum_t\mu_tH_t(y)+H_3(y)=x\).
The mirror extends Theorem 4.4 from two consumption components and one shared terminal-wealth component to a finite support of client types. It leaves aside the paper's specific market dynamics, risk-aversion results, and numerical illustrations.
The strongest defensible case is a narrow one: continuize the population of clients sharing a managed portfolio. The paper explicitly mentions a heterogeneous client pool as an application and says its method extends beyond two consumption streams. A plausible regime is a large fund whose clients fall into relatively few complete preference types: members of a type share the same utility and discount parameters, while type \(t\) has population mass \(\mu_t\). The paper’s continuous time and Brownian uncertainty do not themselves continuize the population; there are still only two household members in its main model.
My anchor is Theorem 4.4, proved in this paper. It gives the optimal initial sharing rule for total wealth \(x>0\): \(V(x)=V_1(x_1)+V_2(x_2)+V_3(x_3)\), with \(x_i=H_i(Y(x))\) and \(Y(x)=(H_1+H_2+H_3)^{-1}(x)\). Here \(H_i(y)\) is the initial capital required by subproblem \(i\) when its marginal value is \(y\): for a consumption subproblem it is the discounted, risk-adjusted expected cost of consumption \(I_i(y\zeta^i_t)\); for the terminal-wealth subproblem it is the corresponding expected cost of \(I_3(y\zeta^3_T)\). Thus the theorem allocates wealth by choosing the common marginal value \(y\) that makes the three capital requirements sum to \(x\). It supplies an exact solution to the paper’s optimization problem, though it does not claim a standard complexity class. The paper has no named hardness result to anchor a second question.
The mirror I would put to a prover is Type-Continuum Household Sharing. An instance consists of the paper’s complete-market, finite-horizon environment; total normalized wealth \(x\); a finite support \(T\) of client types with binary-rational masses \(\mu_t>0\), \(\sum_t\mu_t=1\); each type’s utility and discount parameters satisfying the paper’s assumptions; and the common terminal-wealth utility. Choose per-client initial allocations \(a_t\ge0\), a common terminal-wealth allocation \(a_3\ge0\), and admissible type-specific consumption and investment plans to maximize
\[
\sum_{t\in T}\mu_t V_t(a_t)+V_3(a_3)
\quad\text{subject to}\quad
\sum_{t\in T}\mu_t a_t+a_3=x.
\]
A solution is an optimal allocation and its associated consumption paths and aggregate portfolio strategy. With \(H_t\) denoting the type’s capital-demand function and \(H_3\) the common terminal-wealth demand function, the mirrored sharing rule is
\[
a_t=H_t(y),\qquad a_3=H_3(y),\qquad
\sum_{t\in T}\mu_t H_t(y)+H_3(y)=x.
\]
The type masses weight both aggregate welfare and resource use; agents within a type remain indistinguishable in the respects the model uses.
I expect this mirror to be tractable in the paper’s analytic sense: the allocation reduces to finding one common marginal value \(y\) through a strictly monotone scalar equation, then applying each type’s demand function. For a genuine complexity claim, the stochastic primitives need a finite encoding or a specified evaluation model; the paper’s general stochastic processes do not provide one. A numerical version could restrict to the CRRA and mean-reverting-market setting of Section 5 and ask for an \(\varepsilon\)-approximate allocation. Further questions include the bit complexity of evaluating the demand functions, and whether separability survives when types differ in market access or face incomplete markets.
The authors have a plausible reason to recognize this as their problem: the type classes preserve heterogeneous client preferences while replacing named clients with population shares, and the paper itself points to a pool of heterogeneous clients and larger agent groups. The soft spot is that the paper proves Theorem 4.4 for three subproblems, not a continuum of types, and it gives no complexity classification. So this is a credible mirror of its optimal sharing question, but a modest computational-social-choice case rather than a claimed complexity result from the paper.
I would not count Theorem 4.4 as a computational result. It gives an exact economic characterization of an optimal allocation in a continuous-time stochastic-control model. It states no complexity bound, algorithm, or approximation guarantee, and its assumptions permit general utility functions and stochastic processes without a finite input encoding. So the paper supplies no computational result for a population mirror to continue.
The strongest proposed mirror is still coherent: a large managed fund with many clients in a finite number of complete preference types. Type masses and per-client allocations make sense, and the paper itself points to heterogeneous client pools and extensions to more consumption streams. I can’t defeat that scenario by arguing that type grouping distorts the model or that the population lacks multiplicity. To pose a computational problem, though, one must add an input model for the market and preferences—perhaps the CRRA, mean-reverting setting—and specify an approximation target. That would be a new, specialized algorithmic finance problem, rather than a complexity question already present in the paper.
That is the strongest negative case I can make: this is a poor fit as a mirror of a computational result, because the paper has no such result and the proposed mirror needs a new computational formulation. The case is weak as a universal objection. The client-pool model appears to be a plausible high-multiplicity problem the authors would recognize, and the material here gives no principled reason it could never be worth studying.
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.