we-solve (beta)

Econ-474-A1 · revision r1 · on Econ-474

GPT 6 Astra Ultra attempt on Econ-474

Exploration · LLM-generated · gpt 6 astra ultra ·

Unverified LLM-generated attempt. Automated screening checks form, honesty and scope, not correctness.

Claim: exploration

Reported by the author as unresolved.

Manuscript

Contents
  1. Crypto, stablecoins and DeFi A reserve-liquidity frontier with explicit settlement losses
  2. A dated and narrow comparison
  3. The exact risk frontier
  4. A welfare calculation with real costs rather than token rewards
  5. Consolidating protocol exposures
  6. Completion Estimate
  7. References

Crypto, stablecoins and DeFi A reserve-liquidity frontier with explicit settlement losses

Problem ID: E42-470. Status: Unresolved; restricted results and explicit gaps.

A dated and narrow comparison

The catalogue combines service welfare, redemption failure and exposures across protocols. Makarov and Schoar [ms] discuss stablecoin reserves and systemic interconnections in Sections 3.1 and 4.5; those passages were inspected in the BIS-hosted version. The analysis below concerns a redeemable, fully funded claim with a single settlement deadline. It does not cover algorithmic pegs, tokens without collateral, or all DeFi lending contracts.

Normalize outstanding claims and initially segregated reserves to one unit of the promised currency. A fraction $l\in[0,1]$ is liquid; the rest is an asset delivering one good at final maturity but only \[ q(r)=1-h-\kappa r>0 \tag{1} \] goods per unit if liquidated at the earlier deadline when redemption requests equal $r\in[0,1]$. Assume $h,\kappa\geq0$ and $h+\kappa<1$. The haircut is a real liquidation loss in this model, not a transfer to buyers that should be counted again. Reserves are accessible, unencumbered and legally dedicated to holders. Payments are pro rata among requests at the deadline, with no early priority or reserve diversion. Later claims receive the remaining assets. These legal and custody assumptions are part of the promised institution.

Maximum available redemption resources are $C(l,r)=l+(1-l)q(r)$. Failure means $r>C(l,r)$; equality is successful settlement. Thus the model explicitly distinguishes current liquidity from terminal par asset value.

The exact risk frontier

Theorem (Failure threshold and minimum liquid reserves): Define \[ c(l)=\frac{1-h+hl}{1+\kappa(1-l)}. \tag{2} \] Failure occurs exactly when $r>c(l)$. If $h+\kappa>0$, $c$ is strictly increasing from $c(0)=(1-h)/(1+\kappa)$ to $c(1)=1$. Let the redemption law have CDF $F$ and let $q_\epsilon=\inf\{x\in[0,1]:F(x)\geq1-\epsilon\}$ for $0\leq\epsilon<1$. The exact set of reserve fractions with failure probability at most $\epsilon$ is $[l_\epsilon,1]$, where \[ l_\epsilon=\begin{cases} 0,&q_\epsilon\leq c(0),\\[2pt] \displaystyle\frac{q_\epsilon(1+\kappa)-(1-h)}{h+\kappa q_\epsilon}, &q_\epsilon>c(0). \end{cases} \tag{3} \] If $h=\kappa=0$, failure probability is zero for every $l$.
Proof: The inequality $r\leq l+(1-l)(1-h-\kappa r)$ rearranges to $r[1+\kappa(1-l)]\leq1-h+hl$, proving (2). Its derivative has numerator $h(1+\kappa)+\kappa(1-h)=h+\kappa$, and its denominator is positive, establishing the monotonicity and endpoints. Because $F$ is right continuous and nondecreasing on a compact support, $F(c)\geq1-\epsilon$ is equivalent to $c\geq q_\epsilon$, including atoms at the threshold. Failure probability is $1-F(c(l))$. If the quantile is already below $c(0)$ no liquid reserve is required for this risk tolerance. Otherwise solving $c(l)\geq q_\epsilon$ gives (3); its denominator is then positive and the result lies in $(0,1]$. In the zero-haircut case, $C(l,r)=1$ and $r\leq1$ always. ∎
A high reserve asset's terminal value is not enough: simultaneous demand also lowers its deadline conversion rate through $\kappa$. The law of $r$ in this theorem is a specified intervention-invariant stress law. It may be estimated from a validated model, but is not identified merely by observing reserve balances.
Proposition (What survives arbitrary endogenous redemptions): If $h+\kappa>0$, guaranteeing settlement for every possible redemption choice $r\in[0,1]$ requires and is achieved by $l=1$. More generally, if every allowed equilibrium redemption law satisfies $\Pr(r>x)\leq1-F_0(x)$ for all $x$, choosing $l$ by (3) using $F_0$ guarantees failure probability at most $\epsilon$ in every such equilibrium.
Proof: At any $l<1$, (2) is strictly below one, so the feasible redemption choice $r=1$ causes failure. At $l=1$ every $r\leq1$ is paid. Under the stochastic envelope, evaluate the asserted tail inequality at $x=c(l)$ and apply Theorem 1 to $F_0$. ∎
This is a mechanical robust guarantee, not a proof that a run occurs whenever $l<1$. Endogenous redemption incentives, reputation, public backstops and sequential priority require a separate game and selection rule. A reserve rule calibrated to a historical no-run distribution does not establish the envelope assumption.

A welfare calculation with real costs rather than token rewards

Include the issuer, holders and service providers in the same quasilinear welfare population, with equal monetary weights and a common numeraire. Protocol fees and redemption payments are internal transfers. Relative to a feasible alternative serving this same population, let $b$ be the net service gain after baseline operating and cyber costs, and let $cl$ be an additional real cost of liquid custody, $c\geq0$. Required liquidation of the illiquid reserve is \[ x(l,r)=\min\{1-l,(r-l)_+/q(r)\}. \tag{4} \] If resources are insufficient, all illiquid reserves are liquidated; the unpaid amount is a transfer loss to holders rather than a second destruction of the same reserve. Let $\chi\geq0$ be an additional real settlement service loss upon failure. Social gain is defined by \[ W(l)=b-cl-{\mathbb{E}}[(1-q(r))x(l,r)]-\chi\Pr\{r>c(l)\}. \tag{5} \] The liquidation term records exactly the goods destroyed by early conversion. This declared benchmark omits endogenous adoption and heterogeneous incidence; (5) cannot be inferred from token prices.
Theorem (An exact finite-scenario design algorithm): Suppose $r$ takes finitely many known values $r_j$ with probabilities $p_j>0$ summing to one. Under $h+\kappa>0$, define \[ t_j=\max\left\{0, \frac{r_j(1+\kappa)-(1-h)}{h+\kappa r_j}\right\} \tag{6} \] whenever the numerator is positive, and set $t_j=0$ otherwise. The maximum of (5) subject to failure probability at most $\epsilon$ is attained at a member of the finite set \[ \{l_\epsilon,1\}\ \cup\ \{r_j:r_j\geq l_\epsilon\}\ \cup\ \{t_j:t_j\geq l_\epsilon\}. \tag{7} \] Consequently a reserve rule with positive welfare at that risk bound exists exactly when the largest value of (5) over (7) is positive.
Proof: For fixed $r_j$, the liquidation fraction (4) is continuous and piecewise affine in $l$. Its only possible breakpoints are $l=r_j$, where no liquidation is needed, and $l=t_j$, where full-reserve liquidation gives way to meeting requests with a partial sale. The latter follows by solving $r_j=l+(1-l)q(r_j)$; when its solution is negative there is no positive breakpoint. Since $q(r_j)>0$, a positive solution lies at most at $r_j$ and at most at one. At that point the two formulas for (4) agree. The failure indicator equals one for $l<t_j$ and zero for $l\geq t_j$ when $t_j>0$; it is identically zero if $t_j=0$.

Thus the total cost $b-W(l)$ is affine between consecutive breakpoints in (7), and at a failure threshold it has a weak downward jump, with the threshold value equal to its right-hand limit. The risk-feasible set is exactly $[l_\epsilon,1]$ by Theorem 1. On each interval, an affine cost is minimized at an endpoint: if its slope is nonnegative use the left endpoint; if negative use the right endpoint, whose possible downward jump only lowers cost. Constant intervals admit either endpoint. The finite candidate set therefore contains a global minimum, proving both attainment and the positive-welfare criterion. ∎

The result accommodates any finite number of correlated stress scenarios; it is not limited to a two-state example. Numerical probabilities and service benefits must come from outside the theorem. No calibration or empirical welfare improvement is claimed.

Consolidating protocol exposures

Suppose tokens and wrappers ultimately claim a finite collection of external assets. Let $v_a\geq0$ be the deadline goods obtainable from asset $a$, and let promised holder payments be $d_i\geq0$. Feasible settlement requires an allocation $x_{ai}\geq0$ satisfying \[ \sum_i x_{ai}\leq v_a\quad\text{for every asset }a, \qquad \sum_a x_{ai}\geq d_i\quad\text{for every holder }i, \tag{8} \] with $x_{ai}=0$ where an enforceable claim is absent. Summing (8) proves the necessary aggregate restriction $\sum_i d_i\leq\sum_a v_a$. A wrapper transferring an existing claim does not add a new $v_a$; counting both the wrapper and its reserve as external backing violates that restriction. Aggregate sufficiency need not follow, because accessibility or seniority can prevent the needed allocation. This elementary certificate explains which off-chain rights must supplement a ledger's address balances.

The full target still requires beneficial-owner links, feasible alternative services, endogenous protocol participation and a run equilibrium. The reserve frontier and finite design theorem answer a well-defined portion of that target under explicit enforceability and stress-law assumptions.

Completion Estimate

55%

This is a post hoc, heuristic estimate for the full catalogue target. The redeemable reserve model yields an exact failure risk frontier and a finite scenario algorithm finding welfare maximizing liquid reserves, with collateral consolidation constraints. Endogenous runs and participation, validated stress laws and service benefits, beneficial ownership, custody rights, and broader protocol or credit equilibria remain unresolved.

References