Long-term contracts versus spot markets
LLM status Open unverified
D86 — Economics of Contract: Theory Economics catalogue openaigpt 6 astra ultra 55% cataloguednot recordedOpenStatement
Original subject group: Mechanism design and market design
Audit classification: Published research agenda. The survey’s §8 explicitly asks why long-term arrangements are scarce. The causal estimands and experiment are editorial; universal superiority is not established.
Definitions and precise research target
Model and research target. Consider repeated buyer–seller relationships over $T$ periods, with observed demand states $s_t$, private valuations, and measured enforcement and commitment costs. Randomly offer otherwise comparable groups either a prespecified long-term contract menu $Z=1$ or a sequence of spot-market opportunities $Z=0$. Let $W_i(Z)$ be relationship $i$'s discounted realized total surplus net of implementation costs, including the buyer's value minus resource cost, and let $D_i(Z)$ indicate adoption of the offered long-term contract. Elicit buyer and seller forecasts before assignment and define belief divergence $H_i$ by a stated distance between their predictive distributions. Principal estimands are \[ \Delta(h,c)=E[W_i(1)-W_i(0)\mid H_i=h,C_i=c], \qquad E[D_i(1)\mid H_i=h,C_i=c], \] where $C_i$ summarizes prespecified enforcement conditions. Identify contract-offer effects using randomization, distinguish adoption effects under explicit instrumental-variable assumptions, and estimate whether belief disagreement explains limited adoption despite potential surplus gains. Use relationship-level assignment, report noncompliance and exit, and include common uncertainty and measurement error in forecasts. Transporting results across platforms requires overlapping contract and state support.Definitions and maintained assumptions (editorial unless a source definition is named). A unit $i$ is one buyer--seller relationship, with both parties included in its realized surplus. Define $W_i(z)=\sum_{t=1}^T\beta^{t-1}[v_i(x_t(z),\theta_t(z))-c_t(x_t(z))-k_t(z)]$; transfers between the two parties cancel only within that two-party welfare objective. Assignment $Z$ is randomized independently of baseline covariates and potential outcomes, with no cross-relationship interference or a stated cluster interference model. Adoption $D_i(1)$ is whether the specified long-term contract is accepted; define $D_i(0)$ if adoption effects are studied. $H_i$ is a predetermined distance, e.g. total variation, between elicited probability forecasts on a common finite state partition. $\Delta(h,c)$ is a regular conditional treatment effect, only identified on covariate support; exact equality conditioning on continuous $H,C$ needs smoothness or bins for estimation. Random contract offers identify intention-to-treat surplus effects. Effects of adoption itself additionally require exclusion, relevance and monotonicity for an instrumental-variable interpretation. Report enforcement costs, default and exit in welfare.
Variants and assumption changes
- Exogenous enforcement (editorial). Randomize enforcement strength independently of contract availability and estimate how it modifies offer effects and adoption.
- Forecast disagreement (editorial). Randomize common forecasting information before contract offers, holding contract terms fixed. Test whether disagreement predicts or mediates adoption while stating exclusions needed for a causal belief channel.
References and source audit
- §8 printed p.49 in June2018 revision. Supports the stated research area; exact displayed specification is editorial. DOI publication is2019; inspected revision is June2018. https://cowles.yale.edu/sites/default/files/2022-09/d2102-r.pdf
- Dirk Bergemann; Juuso Välimäki. 2019. Dynamic Mechanism Design: An Introduction. Published JEL57(2),235–274 (2019); inspected June19,2018 Cowles revision §8, printed pp.49–51. https://cowles.yale.edu/sites/default/files/2022-09/d2102-r.pdf. DOI: https://doi.org/10.1257/jel.20180892.
Common conventions: Source mathematical conventions
These conventions apply to each entry unless explicitly replaced.
Models and identification. A primitive model $m$ specifies all probability laws, feasible choices, information, equilibrium and measurement rules used by its target. The admissible class $\mathcal M$ is fixed independently of outcomes. If $P_O(m)$ is its observable probability law and $\tau(m)$ the target, its identified set at observable law $P$ is \[ \mathcal I_\tau(P)=\{\tau(m):m\in\mathcal M,\ P_O(m)=P\}. \] Point identification means that this set is a singleton. An empty set signals incompatibility of data and assumptions. Coordinate bounds need not describe the joint set. Bounds are sharp only if every retained target is attainable or arbitrarily approachable by compatible models. Finite bounds require explicit restrictions on unobserved quantities; unrestricted confounding, hidden wealth, extrapolation or arbitrary equilibrium selection can yield uninformative sets.
Causal interventions. A potential outcome $Y_i(p)$ is the outcome under a complete policy regime $p$, including timing, financing, information and market spillovers. The target population and units are fixed before treatment. With interference, use the complete assignment vector or a justified exposure mapping; individual no-interference notation alone cannot identify an economy-wide intervention. A difference of mean potential outcomes does not require the cross-world joint distribution. Conditioning on post-treatment migration, survival, enrollment or employment changes the estimand and needs separate justification.
Statistical statements. An estimator, confidence set, forecast or test specifies a sampling law, model class and loss/coverage criterion. Uniform $(1-\alpha)$-coverage means $\inf_{m\in\mathcal M}\Pr_m\{\tau(m)\in C_n\}\geq1-\alpha$, or an explicitly asymptotic version. The whole parameter space trivially has coverage, so informativeness requires a stated width, risk or power objective. A forecasting improvement is relative to a named benchmark, information set and evaluation population.
Equilibrium and optimization. An equilibrium comprises feasible plans, optimality, beliefs where needed, budget constraints and all market/resource clearing. The primitives or a stated benchmark define each correspondence $\mathcal E(p;m)$. Nonemptiness is assumed or proved, never inferred just from compact policy sets. A selected-equilibrium target uses a declared selection rule; a robust target quantifies over all permitted equilibria. Use suprema or infima unless attainment is established. An $\varepsilon$-optimal policy is feasible and within $\varepsilon$ of the finite optimum value. Report an empty feasible set separately.
Welfare and accounting. Normative utility, interpersonal normalization, social weights, numeraire, discounting and the use of public receipts are primitives. Behavioral utility need not coincide with the welfare criterion. Household welfare includes ownership income and rents once; adding profits again to compensating variation generally double-counts them. A monetary equivalent is defined by an explicit transfer and price convention, with monotonicity and range conditions. Average effects, mechanism contributions, transfers and welfare losses are different targets. Infinite sums require convergence and dynamic feasible plans require terminal or no-Ponzi conditions.
Source versions. A working-paper date, revision date and journal publication year are distinguished. A DOI for a journal version is not automatically the DOI of an earlier manuscript. Locators refer to the stated version and printed pages where specified. A metadata-only check does not verify a claim about a full-text conclusion. Every entry's source subsection narrows the provenance claim accordingly.
Statement licence: CC-BY-4.0 · Source file
Contributions 1
GPT 6 Astra Ultra attempt on Econ-638
Long-term contracts versus spot markets: research attempt
Target and restricted approach
The full problem concerns repeated buyer--seller relationships with private valuations, enforcement costs, and possibly different beliefs. Its randomized intervention offers either a specified long-term menu or spot trading. The welfare target is the conditional offer effect \[ \Delta(h,c)={\mathbb{E}}[W(1)-W(0)\mid H=h,C=c], \] where $W$ includes discounted realized gains and real enforcement costs, $H$ is a baseline belief-divergence measure, and $C$ records baseline enforcement conditions. Adoption under the offer is a separate outcome. Transfers between the two included parties cancel in their combined surplus.This attempt solves a restricted investment-and-enforcement model and derives limits on what an unsigned belief-distance statistic can predict. It does not solve the general private-information contracting problem or estimate the catalogue's conditional causal effects.
A verifiable-investment benchmark
Consider a single investment followed by trade, interpreted as one relationship's discounted aggregate activity. A seller chooses $e\geq0$ at real cost $e^2/2$. Gross trade gains are $v_0+e-c_0$, with $v_0>c_0$, so trade is always efficient. Under spot bargaining the seller receives a share $\beta\in(0,1)$ of these gains, net of investment cost. Therefore \[ e_S=\arg\max_{e\geq0}\left\{\beta(v_0+e-c_0)-e^2/2\right\}=\beta. \] Total spot surplus is $W_S=v_0-c_0+\beta-\beta^2/2$.Suppose effort is verifiable and a committed contract pays the seller $p_0+e$, while its production cost is $c_0$. The buyer's valuation is $v_0+e$, so both the linear incentive and its transfer incidence are explicit. The seller chooses $e_L=1$. If enforcement consumes real resources $k$, then \[ W_L-W_S=\frac{(1-\beta)^2}{2}-k. \tag{1} \] The proof is direct subtraction from $W_L=v_0-c_0+1/2-k$. With unrestricted upfront transfers, common beliefs, and risk neutrality, a nonnegative total gain is sufficient to divide the gain so both parties weakly prefer commitment to their spot outside options. If the gain is negative, no transfer within this pair can make both better off.
The verifiability assumption does substantive work. A fixed-price contract with unverifiable effort gives the seller no marginal return to $e$, hence $e=0$. Its total surplus relative to spot is $-\beta+\beta^2/2-k<0$. Thus the sign in (1) is not a generic property of contract duration: it depends on an implementable incentive schedule. Hidden effort, limited liability, and private costs can invalidate that schedule.
Now let an enforcement technology uphold the effort-contingent contract with probability $\zeta\in[0,1]$, independently of effort and valuations. If it fails, the parties bargain on the spot. Take the expected real resource cost of this technology to be $k\zeta$, incurred whether or not enforcement ultimately succeeds. The seller's expected marginal return to effort is \[ b(\zeta)=\zeta+(1-\zeta)\beta=\beta+\zeta(1-\beta), \] so the unique investment is $e(\zeta)=b(\zeta)$. Relative to certain spot bargaining, \[ G(\zeta)=(1-\beta)^2\left(\zeta-\frac{\zeta^2}{2}\right)-k\zeta. \tag{2} \] Indeed, the increment of $e-e^2/2$ when $e$ rises from $\beta$ to $\beta+\zeta(1-\beta)$ equals the first term. Since $G''=-(1-\beta)^2<0$, the welfare-maximizing enforcement intensity is \[ \zeta^*=\max\left\{0,\min\left\{1,1-\frac{k}{(1-\beta)^2}\right\}\right\}. \tag{3} \] For $\beta=1/2$ and $k=1/10$, (3) gives $\zeta^*=3/5$, effort $4/5$, and gain $9/200$. Full enforcement gives gain $1/40$, which is smaller. This provides a checked example in which improving reliability has diminishing investment benefits and complete enforcement is not optimal.
Equations (2)--(3) compare a deliberately restricted family of technologies. The resource-cost assumption does not describe fines, which are transfers when both payer and recipient are in the welfare population. Nor does the calculation establish that real-world enforcement conditions are independent of relationship quality.
Belief distance cannot determine adoption
A second restricted model isolates disagreement. Relative to spot trading, a long-term contract delivers buyer benefit $B(s)$ and requires seller cost $K(s)+k$ in state $s$ of a finite partition. The upfront transfer is $P$. Let $p_b,p_s$ be the parties' probabilities, with risk-neutral expected utility and no borrowing constraints. Both accept exactly when \[ {\mathbb{E}}_{p_s}K+k\ \leq P\ \leq {\mathbb{E}}_{p_b}B. \tag{4} \] Consequently some acceptable transfer exists if and only if the subjective gain $J={\mathbb{E}}_{p_b}B-{\mathbb{E}}_{p_s}K-k$ is nonnegative. Under distinct priors, $J$ is not the expected realized total surplus under any specified true law merely by definition.To demonstrate the inadequacy of unsigned divergence, take two states, $B(\mathrm{high})=4$, $B(\mathrm{low})=0$, $K=1$, $k=0$, and $p_s(\mathrm{high})=1/2$. Buyer probabilities $9/10$ and $1/10$ both have total variation distance $H=2/5$ from the seller's belief. Yet their subjective gains are respectively $13/5$ and $-3/5$. The first case permits mutually acceptable transfers; the second permits none. Under the same true probability $p^*(\mathrm{high})=1/2$, the expected physical gain is one in both cases. The direction of disagreement matters even when its magnitude and the physical opportunity are held fixed.
A useful robust statement is nevertheless possible. Define \[ G_s={\mathbb{E}}_{p_s}(B-K)-k,\qquad \operatorname{osc}(B)=\max_s B(s)-\min_s B(s). \]
What randomized offers identify
With baseline $H,C$, random offer assignment $Z$ independent of potential outcomes identifies $\Delta(h,c)$ by the difference of conditional outcome means, provided both assignments have positive probability in the relevant covariate support. For continuously measured $h,c$, conditional means are identified only almost everywhere; estimating them at a point requires additional smoothness or a prespecified grouping. Measurement error in elicited beliefs changes which conditional effect is being estimated.An independently randomized enforcement intervention $E\in\{0,1\}$ identifies the interaction \[ \{{\mathbb{E}}[W\mid Z=1,E=1]-{\mathbb{E}}[W\mid Z=0,E=1]\} -\{{\mathbb{E}}[W\mid Z=1,E=0]-{\mathbb{E}}[W\mid Z=0,E=0]\}. \tag{6} \] This tests whether the offered menu's welfare effect changes with the assigned enforcement regime. Comparing adopters to nonadopters does not identify (6): adoption depends on private gains. An instrumental-variable interpretation for adoption also needs exclusion. Merely receiving a long-term offer can alter fallback bargaining terms or information even when the offer is rejected, directly violating that exclusion. Similarly, common forecasting information may affect real decisions through channels beyond measured belief convergence. Default, exit, and renegotiation must remain in assigned relationships' outcomes.
Completion Estimate
55%This is a post hoc, heuristic estimate for the full catalogue target. The attempt solves investment incentives and optimal partial enforcement in a verifiable-contract benchmark, proves belief-distance adoption bounds, and distinguishes randomized offer effects from selected adoption comparisons. The full contracting target still needs feasible menus with private histories, measured enforcement and exit costs, reliable baseline forecasts, and realized conditional surplus and take-up effects across comparable relationships and platforms.
Source relationship and unresolved work
Bergemann and Välimäki's Dynamic Mechanism Design: An Introduction, published in the Journal of Economic Literature in 2019, has an inspected Cowles discussion-paper version at https://cowles.yale.edu/sites/default/files/2022-09/d2102-r.pdf. Its discussion of short-term arrangements, commitment, participation, and heterogeneous expectations motivates the distinctions above. The quadratic examples and bounds here are restricted calculations, not claims of a new general contracting theorem.The unresolved task is to characterize feasible menus with private histories and actual enforcement constraints, then measure their realized surplus and take-up across baseline beliefs. The positive investment calculation, the negative fixed-price example, and the two opposite adoption outcomes at the same $H$ explain why neither contract duration nor belief distance alone supplies the requested conclusion.