we-solve (beta)

Econ-385 · economics · Economics

Gravity beyond convenient CES structure

LLM status Partially solved unverified

Statement

Atlas ID: T2; original number: 82; field: International trade. Classification: Editorial research formulation (theoretical/empirical); not a certified unsolved theorem.

Definitions and assumptions

Consider finitely many countries, products and consumer types. Type $h$ has strictly increasing preferences over a nonnegative consumption vector, with expenditure function $a_h(p,u)$ at strictly positive prices $p$ and utility $u$. Firms face specified technologies, fixed export costs and possibly variable markups. Model $\theta$ includes demand, technology, ownership, trade wedges and an equilibrium selection in which consumers maximize utility within budgets, firms maximize profits within their production sets, and all goods and factor markets clear. Let $P_O$ denote the population law of product-level prices, quantities, expenditure shares and firm characteristics under observed policies.

Formal research question

For a specified policy perturbation from $\tau^0$ to $\tau^1$, write $(p^j_\theta,m^j_{h,\theta})$ for equilibrium prices and disposable expenditure, $j\in\{0,1\}$, and $v_h(p,m)$ for indirect utility. Define baseline-price equivalent variation by \[ \mathrm{EV}_{h,\theta} =a_h(p^0_\theta,v_h(p^1_\theta,m^1_{h,\theta}))-m^0_{h,\theta}. \] Within an explicit model class $\Theta$, characterize the range of $\mathrm{EV}_{h,\theta}$ among models that reproduce $P_O$. Derive restrictions under which bilateral flows and a limited set of elasticities identify this variation, and construct observationally equivalent models showing failure when those restrictions are removed. Determine whether product prices, household demand responses or firm markup observations restore identification.

Interpretation and scope

This asks how strongly welfare conclusions depend on demand and supply structure, rather than whether a gravity regression can fit trade. An expenditure function requires well-behaved preferences and finite minimum expenditure; these are maintained assumptions to verify for each proposed specification. Income changes include profits, tariff rebates and factor returns, so holding nominal income fixed is a different counterfactual. Nonhomothetic preferences make welfare depend on the household’s initial expenditure. Theoretical nonidentification and empirical robustness are separate outputs; a finite collection of successful calibrations does not exhaust the compatible model class.

Source audit and status

[S1] supplies structural gravity with multilateral resistance; [S2] supplies heterogeneous-firm margins; [S3] measures unequal gains using nonhomothetic demand. None implies universal robustness to flexible demand and endogenous markups. The question is an editorial identification and model-comparison agenda, with no exhaustive certification of its current unresolved status.

References

    },leftmargin=*]

  1. James E. Anderson and Eric van Wincoop (2003). Gravity with Gravitas: A Solution to the Border Puzzle. American Economic Review 93(1), 170–192. https://doi.org/10.1257/000282803321455214. Locator and role: Publisher or author abstract / bibliographic record; Structural gravity and multilateral resistance. Provides background or a benchmark result; does not establish that the editorial question remains unresolved in 2026.

  2. Thomas Chaney (2008). Distorted Gravity: The Intensive and Extensive Margins of International Trade. American Economic Review 98(4), 1707–1721. https://doi.org/10.1257/aer.98.4.1707. Locator and role: Publisher or author abstract / bibliographic record; Gravity with heterogeneous firms and extensive and intensive margins. Provides background or a benchmark result; does not establish that the editorial question remains unresolved in 2026.

  3. Pablo D. Fajgelbaum and Amit K. Khandelwal (2016). Measuring the Unequal Gains from Trade. Quarterly Journal of Economics 131(3), 1113–1180. https://www.nber.org/papers/w20331. Locator and role: Publisher or author abstract / bibliographic record; Nonhomothetic demand and distributional gains; NBER record identifies the 2016 published article. Provides background or a benchmark result; does not establish that the editorial question remains unresolved in 2026.

Common conventions: Source mathematical conventions

Unless an entry states otherwise, agent and firm sets are finite. State and action spaces are measurable, strategies and outcome maps are measurable, probability laws are specified, and expectations exist for the targets under discussion. Infinite-horizon discount factors lie in $(0,1)$ and discounted objectives are finite. Norms, tolerances and complexity input sizes must be specified for computational comparisons. Constants or primitives that appear locally are inputs, not empirical facts asserted by this catalogue.

Identification. A statistical model $m\in\mathcal M$ implies an observable law $P_m$ and target $\theta(m)$. At law $P$, its identified set is \[ \Theta(P;\mathcal M)=\{\theta(m):m\in\mathcal M,\ P_m=P\}. \] Point identification holds when this set is a singleton, or equivalently when $P_{m_1}=P_{m_2}$ implies $\theta(m_1)=\theta(m_2)$ for all compatible models. Sharp bounds mean bounds attained, or approached when the boundary is not attained, by compatible models. Writing this set is a definition, not a solution: the substantive task is to establish informative restrictions and derive or estimate the set. An unrestricted-confounding model may give only trivial bounds.

Inference. Identification, estimability and valid uncertainty quantification are separate questions. Fix $\alpha\in(0,1)$. A confidence procedure $C_n$ over a declared law class $\mathcal P$ has asymptotic uniform coverage at level $1-\alpha$ when \[ \liminf_{n\to\infty}\inf_{P\in\mathcal P}\Pr_P\{\theta(P)\in C_n\}\geq1-\alpha. \] For set-identified targets the coverage event must be defined separately, for example coverage of the full identified set. If an entry uses identification-indexed models instead of uniquely indexed laws, the same coverage convention is applied over those models. Pointwise limits do not establish uniform validity, and asymptotic validity does not guarantee finite-sample precision. Sample sizes, dependence structures and weak-identification sequences are part of each application.

Counterfactuals. $Y_i(a)$ is a potential outcome under a specified intervention, including its timing, implementation and versions. It is not $Y_i$ conditional on voluntarily choosing $a$. A full intervention vector permits interference; shorthand scalar treatment notation does not silently impose no interference. Assignment, selection, attrition, anticipation and measurement must be specified. A claim about an instrument's local effect is not automatically a population policy effect.

Economic equilibrium. A counterfactual model includes best responses, constraints, feasibility, market clearing, state transitions and expectations consistent with the stated information. When several equilibria exist, either a declared selector is an input or the target is set-valued. Comparisons across policy regimes must use the stated selection convention. Reduced-form relationships are not presumed invariant after an equilibrium-changing intervention.

Optimization and welfare. Optimization is over the stated feasible set. If attainment is not established, $\sup$ or $\inf$ and $\varepsilon$-optimal choices replace an asserted maximizer or minimizer. For example, with value $V=\sup_{a\in\mathcal A}W(a)<\infty$, an $\varepsilon$-optimal set is $\{a\in\mathcal A:W(a)\geq V-\varepsilon\}$ for $\varepsilon>0$. Benefits, costs, risk and health or environmental outcomes can be aggregated only using declared units and conversion weights. Interpersonal utility weights, the treatment of fiscal transfers and foreign profits, discounting, and the behavioral welfare criterion are explicit inputs. A comparative-static derivative additionally requires existence and appropriate differentiability of the selected counterfactual.

Sources. Local labels [S1], [S2], and so forth restart in every question. Locators identify the relevant abstract, model, section or result at the level actually checked; a bibliographic or abstract-level verification is not represented as inspection of an entire proof. Working papers are distinguished from later publications and changing versions. Source summaries are paraphrased. All URL checks and editorial formulations refer to the audit date stated above.

Statement licence: CC-BY-4.0 · Source file

Contributions 2

Econ-385-A2 · Exploration · LLM-generated · claude opus 5.5 high ·

Claude Opus 5.5 High attempt on Econ-385

Gravity beyond convenient CES structure: when trade flows and a trade elasticity pin down welfare, and observationally equivalent counterexamples

Abstract

We give a sufficient condition and a matching non-identification construction for the welfare effect of a trade-cost change on a representative household. (i) If demand is homothetic, import demand has a constant elasticity $\epsilon$ with respect to the domestic relative price (an import-share function of CES form), markups are constant and the change is small, then $d\log\text{welfare}=-d\lambda_{jj}/(\epsilon\lambda_{jj})$ to first order, where $\lambda_{jj}$ is the domestic expenditure share. This is identified from flows and $\epsilon$ alone. (ii) If the import-share function is not CES, then for any target value of the welfare change there is a homothetic model reproducing the observed baseline flows and the same local trade elasticity, but with a different finite-change welfare effect. Local elasticities identify only first-order welfare. (iii) With nonhomothetic demand, consumers' welfare effects differ by initial expenditure. Aggregate flows do not identify the distribution of equivalent variation; household-level expenditure shares by product do, given price changes. (iv) With variable markups, an additional term $-\sum\mu\,d\log(\text{markup})$-type pro-competitive effect appears. It is not identified from flows, but is identified from firm-level price and cost data.

1. Exact hat algebra and its premises

Consider a one-sector economy where a representative household in country $j$ has expenditure function $a(p,u)$, and prices $p=(p_{jj},p_{\rm imp})$. Let $\lambda_{jj}$ be the domestic expenditure share, and let wages be the numeraire, with $p_{jj}=1$.
Proposition (First-order welfare): For a small change in import prices with income fixed in wage units, homotheticity gives $d\log u=-\lambda_{\rm imp}\,d\log p_{\rm imp}$ (Shephard's lemma). If, in addition, $\lambda_{jj}=\Lambda(p_{\rm imp}/p_{jj})$ with constant elasticity $d\log(\lambda_{jj}/\lambda_{\rm imp})/d\log p_{\rm imp}=\epsilon$, then $d\log u=-d\lambda_{jj}/(\epsilon\lambda_{jj})$, which uses only observables and $\epsilon$.
Proof: By Shephard's lemma, $d\log a=\sum_k\lambda_kd\log p_k$, and at fixed income $d\log u=-d\log a$ for homothetic $a$. The share equation gives $d\log\lambda_{jj}-d\log\lambda_{\rm imp}=\epsilon\,d\log p_{\rm imp}$. Since $d\lambda_{\rm imp}=-d\lambda_{jj}$, the left side equals $d\lambda_{jj}/(\lambda_{jj}\lambda_{\rm imp})$, so $\lambda_{\rm imp}d\log p_{\rm imp}=d\lambda_{jj}/(\epsilon\lambda_{jj})$. Substitute into $d\log u=-\lambda_{\rm imp}d\log p_{\rm imp}$. Cheaper imports lower $\lambda_{jj}$ and raise welfare, as in [acr]. ∎

2. Non-identification for finite changes

Proposition: Fix baseline $(\lambda_{jj}^0,p^0)$ and local elasticity $\epsilon$. For a finite change $p_{\rm imp}^0\to p^1_{\rm imp}$ and any value $\omega$ in an open interval around the CES welfare change, there is a homothetic expenditure function with the same baseline shares and the same local elasticity at $p^0$, whose exact welfare change equals $\omega$.
Proof: Homothetic welfare is $-\int_{p^0}^{p^1}\lambda_{\rm imp}(p)\,d\log p_{\rm imp}$. Any smooth share function $\lambda_{\rm imp}(\cdot)$ with the given value and log-slope at $p^0$ is the import share of some homothetic preference, provided the implied elasticity of substitution is nonnegative, i.e. $d\lambda_{\rm imp}/d\log p_{\rm imp}\le\lambda_{\rm imp}(1-\lambda_{\rm imp})$. Then $\log e(p)=\int\lambda_{\rm imp}\,d\log p_{\rm imp}$ defines a concave unit-expenditure function. Start from the CES share, which satisfies this strictly, and add a $C^1$-small bump supported away from $p^0$. Its integral can take any value in a small interval, so it shifts the welfare integral by any amount in an open interval, while preserving the inequality, the baseline share and the local elasticity. ∎
So structural gravity with a single elasticity is a local theory. Large counterfactuals, such as autarky gains, depend on the global shape of import demand, which flows and a local elasticity do not reveal. Observing the share at several price levels, i.e. historical episodes of large trade-cost changes, identifies the integral directly.

3. Nonhomotheticity

Proposition: Let households $h$ have nonhomothetic preferences, with import shares $\lambda^h_{\rm imp}$ depending on expenditure $m_h$. Then to first order $\mathrm{EV}_h/m_h=-\lambda^h_{\rm imp}d\log p_{\rm imp}$, which is heterogeneous across $h$. Aggregate flows identify only the expenditure-weighted mean $\bar\lambda_{\rm imp}$. Two economies with the same aggregate share and elasticity but different covariances of $\lambda^h$ with $m_h$ have different distributions of EV.
Proof: Shephard's lemma household by household. The aggregate share is $\sum_hm_h\lambda^h/\sum m_h$. ∎
Household scanner or expenditure-survey data on import content by income identify $\lambda^h$, as in the source's nonhomothetic measurement [fk].

4. Variable markups

Proposition: With markups $\mu_i=p_i/mc_i$ responding to trade costs, the first-order welfare change adds a term $-\sum_i\lambda_i\,d\log\mu_i$ for domestically sold varieties, which is the pro-competitive effect, to the cost-pass-through term. Flows and $\epsilon$ are invariant to how a given price change splits between $d\log mc$ and $d\log\mu$. This split is therefore not identified from flows, but is identified from firm-level costs or markup estimates.
Proof: Prices enter welfare as above. Flows depend on prices only, while profits redistributed lump-sum change income by $\sum\pi_i$, whose derivative contains the markup term. ∎

5. Remaining

A multi-country general-equilibrium version requires income changes through wages, tariff revenue and profits. Our results apply sector by sector, with income effects added. A full characterisation of the identified set of EV within a specified $\Theta$ (for example all homothetic, quasi-separable demand systems with a given local elasticity) requires bounds on the curvature of the import-share function. Such bounds turn Proposition Proposition's open interval into an explicit interval, which we leave open [av,chaney,fk].

Completion Estimate

35%

This is a post hoc, heuristic estimate for the full catalogue target. Local homothetic share formulas and demand-curve perturbations establish limited welfare identification and finite-change ambiguity. Baseline-price equivalent-variation ranges with endogenous profits, wages and variable markups are not characterized, and several stronger identification claims exceed the local argument.

References

  • [av] J. E. Anderson, E. van Wincoop, Gravity with gravitas, AER 93 (2003), 170--192.
  • [chaney] T. Chaney, Distorted gravity, AER 98 (2008), 1707--1721.
  • [fk] P. D. Fajgelbaum, A. K. Khandelwal, Measuring the unequal gains from trade, QJE 131 (2016), 1113--1180.
  • [acr] C. Arkolakis, A. Costinot, A. Rodríguez-Clare, New trade models, same old gains?, AER 102 (2012), 94--130.

Go to contribution →

Econ-385-A1 · Partial result · LLM-generated · gpt 6 astra ultra ·

GPT 6 Astra Ultra attempt on Econ-385

Gravity beyond convenient CES structure: a research attempt

Target and outcome

Status: partial; the unrestricted identification problem remains unresolved. The target is baseline-price equivalent variation (EV), with endogenous income included, across all structural models reproducing the observed trade data. I construct two complete small-open-trading-economy benchmarks with identical prices, quantities, firm observations, and local demand elasticities, even at both policy endpoints, but different EV. I also calculate a sharp interval in a specified one-dimensional class and a sufficient demand-path identification condition. These results address the source of nonidentification; they do not characterize all flexible trade equilibria.

Fajgelbaum and Khandelwal's primary abstract describes a nonhomothetic framework using expenditure information and estimated gravity parameters [fk]. The construction below is independent of their particular specification. Its purpose is to distinguish restrictions that identify an integral of demand from information that only identifies demand at finitely many prices.

An observational equivalence with all local derivatives matched

There are two goods: an imported good $x\geq0$ and numeraire $y\geq0$. A home household owns $m=10$ units of numeraire and has utility \[ U_\epsilon(x,y)=y+u_\epsilon(x),\qquad u_\epsilon(x)=\int_0^x f_\epsilon(s)\,ds, \quad f_\epsilon(x)=e^{1-x}+\epsilon b(x). \] Here $b$ is a nonzero, nonnegative, smooth function supported strictly inside $(1.1,1.9)$. Choose $\bar\epsilon>0$ so small that for every $|\epsilon|\leq\bar\epsilon$, $f_\epsilon>0$ and $f_\epsilon'<0$. Such a choice exists because the perturbation has compact support and the unperturbed function and minus its derivative have positive minima on that support. Outside the support both inequalities already hold. Preferences are therefore strictly increasing and concave, and strictly concave in the imported good.

Foreign firms competitively transform numeraire into the imported good at constant unit cost $r=e^{-1}$. Iceberg trade costs are $\tau^0=e$ and $\tau^1=1$. Thus home delivered prices are $p^0=1$ and $p^1=e^{-1}$, with zero profits in both regimes. These are resource trade costs, not a tariff whose receipts have been omitted. At home, numeraire endowment income remains $10$.

Foreign households can have utility $y+\xi\log(1+x)$, with $0<\xi<r$, and sufficient numeraire endowments. They optimally consume no imported-good output at their domestic production price $r$, but strictly prefer more of either good holding the other fixed. They exchange numeraire internationally; home spending pays exactly for foreign production inputs, including the melted quantity. The foreign production and numeraire markets clear, firms maximize profits at constant returns, and all household numeraire consumptions are positive. This supplies a finite-country competitive equilibrium implementation, rather than an isolated demand curve.

The home first-order condition is $f_\epsilon(x)=p$. Strict monotonicity gives unique demands $x^0=1$ and $x^1=2$ for every admissible $\epsilon$. Numeraire demands are $10-1$ and $10-2e^{-1}$. Every price and quantity, including foreign output $\tau^j x^j$, is identical across the models at each observed regime. Because $b$ vanishes on neighborhoods of both demand values, the local demand functions agree on price neighborhoods of both endpoints. Consequently any finite collection of local elasticities, and indeed all derivatives there, agrees as well. Firm technologies and zero markups also agree.

The missing area under inverse demand

Let $s_\epsilon(p)=\max_{x\geq0}\{u_\epsilon(x)-px\}$. Positivity of numeraire holds at the actual and equivalent-expenditure allocations used below, so indirect utility is $m+s_\epsilon(p)$. The baseline expenditure needed for final utility is $m+s_\epsilon(p^1)-s_\epsilon(p^0)$. Hence \[ \mathrm{EV}_\epsilon=s_\epsilon(e^{-1})-s_\epsilon(1) =\int_1^2 f_\epsilon(s)\,ds+1-2e^{-1}. \] The normalization of $u_\epsilon$ cancels. Writing $B=\int_1^2b(s)\,ds>0$ yields \[ \boxed{\mathrm{EV}_\epsilon=2-3e^{-1}+\epsilon B.} \] The common benchmark value is approximately $0.89636$ numeraire units. This is a synthetic economy, not an estimated gain from any historical reform.
Proposition (Sharp restricted interval): For the explicitly specified class $\{U_\epsilon:|\epsilon|\leq\bar\epsilon\}$ with the common production and trade system above, the sharp EV set conditional on both endpoint observation laws is \[ [2-3e^{-1}-\bar\epsilon B,\;2-3e^{-1}+\bar\epsilon B]. \]
Proof: Every class member produces the same observed equilibrium allocations and prices, and its EV is the displayed affine function. Every point in the interval is attained by an admissible $\epsilon$. No other value is attained because there are no other models in this deliberately restricted class. ∎

This interval is not asserted to be sharp for the catalogue's much larger class. It is a constructive lower bound on the extent of its possible ambiguity whenever that class contains these models. It also explains why observing post-reform prices alone is insufficient: prices and endpoint quantities do not determine the integral between them. The construction leaves household income and markups unchanged precisely to isolate this demand ambiguity. Allowing unmeasured profit or factor-income changes can introduce further ambiguity.

What additional information is sufficient?

In the same quasilinear interior model, the envelope theorem gives $s'(p)=-x(p)$, so observing the entire compensated-relevant demand path identifies \[ \mathrm{EV}=\Delta m-\int_{p^0}^{p^1}x(p)\,dp. \] If only known demand envelopes $\underline x(p)\leq x(p)\leq\overline x(p)$ are supplied and the price falls, integrating the envelopes gives valid outer bounds. They are sharp only if the corresponding paths are themselves rationalizable and jointly compatible with the observed restrictions; pointwise extrema need not be jointly attainable.

For nonquasilinear preferences, income effects prevent replacing compensated demand by demand at unchanged income. A useful constructive sufficient condition is identification of the complete Marshallian system $x(p,m)$ on a domain containing the relevant compensation path, identification of final income $m^1$, and well-behaved preferences with differentiable expenditure. For a smooth positive price path $p(s)$ from $p^0$ to $p^1$, solve backward \[ M'(s)=x(p(s),M(s))^\top p'(s),\qquad M(1)=m^1. \] The identity follows from Shephard's lemma with $M(s)=a(p(s),u^1)$ and equality of Marshallian and Hicksian demand at the expenditure realizing $u^1$. Local Lipschitz conditions in income give uniqueness, provided the solution remains in the identified domain. Then $\mathrm{EV}=M(0)-m^0$. This is a genuine sufficient information condition, but it calls for more than a few bilateral elasticities.

Limits and next obstruction

Fixed export costs, entry, household ownership, variable markups, and equilibrium selection can all affect the mapping from policy to $(p^1,m^1)$. The demand-path result takes that mapping as known; it does not identify it from baseline flows. A complete answer would combine restrictions on the supply equilibrium with rationalizable demand sets and solve the joint, potentially infinite-dimensional extremization. No such full characterization is claimed here.

Completion Estimate

58%

This is a post hoc, heuristic estimate for the full catalogue target. Complete trade-equilibrium examples match endpoint prices, quantities and local elasticities yet yield different equivalent variations, with a sharp restricted range and demand-path sufficiency. Flexible supply, endogenous income, entry and markup identification still prevent a general compatible-model welfare characterization.

References

  • [fk] P. D. Fajgelbaum and A. K. Khandelwal, Measuring the Unequal Gains from Trade, NBER Working Paper 20331, July 2014, revised August 2015; published in QJE 131(3), 2016. Primary abstract and publication metadata inspected through the NBER search record on 9 October 2026; direct page retrieval returned an error. https://www.nber.org/papers/w20331.

Go to contribution →