Arkolakis–Costinot–Rodríguez-Clare Welfare Formula

The Arkolakis–Costinot–Rodríguez-Clare (ACR) Welfare Formula is a unifying sufficiency theorem in quantitative international trade and business and economics theory, established by economists Costas Arkolakis, Arnaud Costinot, and Andrés Rodríguez-Clare in their landmark 2012 American Economic Review article "New Trade Theories, Same Old Gains?" Over the three decades preceding 2012, international trade theory underwent two major microeconomic revolutions: first, Paul Krugman's (1980) new trade model of monopolistic competition and increasing returns to scale (introducing love-of-variety consumer gains), and second, the early-2000s heterogeneous-firm and Ricardian revolutions led by the Eaton–Kortum model (2002), the Bernard–Eaton–Jensen–Kortum model (2003), and Marc Melitz's (2003) heterogeneous-firm trade model (introducing self-selection of productive firms into exporting and within-industry productivity reallocation). Economists widely believed that these new microeconomic margins—imported product varieties (Broda–Weinstein variety gains) and Melitz firm selection—generated larger aggregate welfare gains from trade than classical Armington (1969) national product differentiation models. Arkolakis, Costinot, and Rodríguez-Clare proved a remarkable equivalence theorem: across any trade model satisfying three macro-level structural conditions—balanced trade, a constant ratio of aggregate profits to aggregate revenue, and a Constant Elasticity of Substitution (CES) structural gravity model of trade—the proportional change in real welfare Ŵ_j = W′_j / W_j resulting from any foreign trade shock is governed by the two-parameter formula:

where λ_jj is the share of country j's total expenditure devoted to domestically produced goods (λ̂_jj = λ′_jj / λ_jj) and ε < 0 is the partial trade elasticity of bilateral imports with respect to variable trade costs.

Four Generations of Quantitative Trade Models

Prior to Arkolakis, Costinot, and Rodríguez-Clare (2012), quantitative trade economists categorized general-equilibrium trade models according to the microeconomic channels through which lower iceberg trade costs τ_ij ≥ 1 raise real national income W_j = Y_j / P_j:

1. Armington (1969) National Product Differentiation (Perfect Competition, Intensive Margin): Each country i ∈ {1, ..., N} produces a single national good differentiated by country of origin, demanded by consumers in country j with constant elasticity of substitution σ > 1. Because the set of goods is fixed exogenously, trade liberalization operates along the intensive margin (lowering import prices τ_ijw_i), with partial trade elasticity ε = 1 − σ < 0.
2. Krugman (1980) New Trade Model (Monopolistic Competition, Homogeneous Firms, Extensive Variety Margin): Differentiated varieties are produced under increasing returns to scale (fixed overhead cost f plus constant marginal cost) by homogeneous monopolistically competitive firms charging a constant markup σ/(σ − 1). Opening to trade adds an extensive margin of imported product varieties M_i while keeping firm scale fixed, again with partial trade elasticity ε = 1 − σ < 0.
3. Eaton–Kortum (2002) Ricardian Trade Model (Perfect Competition, Fréchet Comparative Advantage): A continuum of goods ω ∈ [0, 1] can be produced in every country i with stochastic productivity z_i(ω) drawn independently from a Fréchet distribution F_i(z) = exp(−T_iz^−θ) with dispersion parameter θ > σ − 1. Trade operates along both the intensive price margin and the extensive margin of comparative-advantage specialization (changing the fraction of goods country j buys from each supplier i), with partial trade elasticity ε = −θ < 0.
4. Melitz (2003) Heterogeneous-Firm Model with Pareto Productivity (Chaney, 2008): Monopolistically competitive firms pay a sunk entry cost f_e to draw idiosyncratic productivity φ from a Pareto distribution G(φ) = 1 − (φ_min/φ)^θ with shape parameter θ > σ − 1, and must pay destination-specific fixed exporting costs f_ij > 0 to sell in market j. Trade liberalization triggers within-industry selection: low-productivity domestic firms exit (raising the domestic survival cutoff φ*_jj), high-productivity firms enter foreign export markets (lowering the export cutoff φ*_ij), and labor reallocates toward productive firms, yielding partial trade elasticity ε = −θ < 0.

Why, if the Melitz (2003) model includes two microeconomic channels absent in Armington (1969)—endogenous selection of high-productivity firms and endogenous exit of low-productivity varieties—does it predict the same aggregate welfare gain Ŵ_j = λ̂_jj^1/ε once calibrated to the same observed domestic expenditure share λ_jj and trade elasticity ε?

General Equilibrium Notation and Expenditure Shares

Consider a global economy comprising N countries indexed by i, j ∈ {1, ..., N}, where each country j is endowed with an inelastic supply L_j of a primary factor of production (labor) earning nominal wage w_j. Let X_ij ≥ 0 denote country j's total bilateral expenditure on goods originating in country i, and let X_j = ∑_i=1^N X_ij denote country j's total aggregate expenditure. Define the bilateral import expenditure share λ_ij and the domestic expenditure share λ_jj as:

Following Dekle, Eaton, and Kortum (2008), for any variable x, denote its gross proportional change ("hat algebra") from an initial equilibrium x to a counterfactual equilibrium x′ after an arbitrary foreign shock (changes in iceberg variable trade costs τ_ij, fixed exporting costs f_ij, or foreign endowments L_i for i ≠ j, holding domestic technology and domestic trade costs τ_jj, f_jj fixed) as x̂ ≡ x′ / x.

Restriction R1 — Balanced Trade and Restriction R2 — Constant Profit Share

Arkolakis, Costinot, and Rodríguez-Clare (2012) showed that all four flagship trade models satisfy three macro-level equilibrium restrictions. First, under Restriction R1 (Balanced Trade), in every country j total national expenditure X_j equals total gross revenue (value of output) Y_j generated by firms located in country j:

Second, let Π_j denote aggregate net profits (after deducting any fixed overhead costs f_ij and sunk entry costs f_e) earned by producers in country j, so total national income is Y_j = w_jL_j + Π_j. Restriction R2 (Constant Profit Share) requires that the ratio of aggregate net profits to total gross revenue is a structural constant π_j ∈ [0, 1) invariant to foreign trade shocks:

• In competitive models (Armington 1969 and Eaton–Kortum 2002) and in monopolistically competitive models with unrestricted free entry (Krugman 1980 and free-entry Melitz 2003), aggregate net profits are zero (π_j = 0), so R2 holds (Y_j = w_jL_j).
• In a restricted-entry Melitz (2003) / Chaney (2008) model with a fixed mass of potential entrants and Pareto productivity G(φ) = 1 − (φ_min/φ)^θ, the constant tail index θ of the Pareto distribution ensures that aggregate net profits remain a constant fraction π_j = (σ − 1)/(σθ) of total revenue regardless of variable or fixed trade costs.

Restriction R3 — CES Import Demand System (Structural Gravity Equation)

Total bilateral import shares λ_ij satisfy a separable structural gravity model of trade with a constant partial trade elasticity ε < 0:

Here A_i reflects exporter i's wage w_i and technology/entry mass N_i, τ_ij ≥ 1 is the iceberg variable trade cost from i to j, γ_ij reflects fixed exporting cost cutoffs (with γ_ij = 1 in Armington, Krugman, and Eaton–Kortum), and ε ≡ ∂ ln(λ_ij / λ_jj) / ∂ ln τ_ij < 0 (holding importer and exporter fixed effects constant) is the partial trade elasticity.

Step 1: Log-Linearization of the Consumer Price Index and Relative Trade Shares

Under CES representative consumer preferences with elasticity of substitution σ > 1 (paired with R1, R2, and R3 while keeping domestic technology L_j, τ_jj, and f_jj constant), choose domestic labor in country j as the local numeraire (w_j = 1, so d ln w_j = 0). By R1 and R2, nominal per-capita income Y_j/L_j = w_j/(1 − π_j) is constant (d ln(Y_j/L_j) = 0), so the infinitesimal logarithmic change in real welfare W_j = (Y_j/L_j) / P_j equals the negative logarithmic change in the ideal consumer price index P_j:

From the structural gravity equation R3, the ratio of bilateral import share λ_ij to the domestic expenditure share λ_jj is λ_ij / λ_jj = (c_i/c_j)^ε (τ_ij/τ_jj)^ε (γ_ij/γ_jj). Taking the logarithmic differential while noting that d ln c_j = d ln τ_jj = d ln γ_jj = 0 yields:

Step 2: Summing Across All Trading Partners and Integrating to Hat Form

Substitute this identity into the price index differential d ln P_j:

Because ∑_i=1^N λ_ij = 1 and ∑_i=1^N λ_ij d ln λ_ij = ∑_i=1^N dλ_ij = 0, the summation over foreign suppliers vanishes, leaving:

Integrating this differential between any initial trading equilibrium (with domestic share λ_jj and welfare W_j) and any counterfactual trading equilibrium (with domestic share λ′_jj and welfare W′_j) yields the ACR Welfare Formula:

In the special counterfactual of moving from the observed trading equilibrium to autarky (where τ_ij → ∞ for all i ≠ j, so λ′_jj = 1 and λ̂_jj = 1 / λ_jj), the total welfare gain from trade relative to autarky (G_j ≡ 1 − W_j^autarky / W_j) is:

Why Microeconomic Margins Cancel Out Conditional on ε

Arkolakis, Costinot, and Rodríguez-Clare identified two offsetting mechanisms that explain why Pareto Melitz yields the same welfare formula as Armington conditional on (λ_jj, ε):

1. Re-Interpretation of the Gravity Trade Elasticity ε: In an Armington or Krugman model without firm heterogeneity, the gravity trade elasticity equals ε = 1 − σ, whereas in a Pareto Melitz model the gravity trade elasticity equals ε = −θ (where θ > σ − 1 is the Pareto tail index). Because part of the observed trade response ε = −5 comes from the extensive margin of marginal exporting firms, calibrating the Melitz model to the same macro trade elasticity ε = −θ = −5 requires setting θ = 5, and the marginal exporting firms that enter at the cutoff φ*_ij are the least productive exporters.
2. Displacement of Domestic Varieties ("Anti-Variety" Effect): In the free-entry Melitz model, when foreign exporters penetrate country j (lowering λ_jj), heightened competition for domestic labor bids up the real wage w_j/P_j and raises the domestic survival cutoff φ*_jj, forcing the least productive domestic firms to exit. Under a Pareto distribution, the consumer welfare loss from losing domestic varieties offsets the consumer welfare gain from acquiring new imported varieties. For the United States with λ_US,US = 0.93 and ε = −5, the single-sector ACR formula yields G_US = 1 − (0.93)^0.20 = 1.44% of GDP.

Extensions that Magnify Welfare Gains Beyond ACR

The ACR theorem clarified the structural assumptions (R1, R2, R3, single sector, no intermediate inputs) that must be relaxed for microeconomic structure to generate larger gains from trade:

1. Multiple Sectors with Heterogeneous Trade Elasticities (Ossa, 2015): When an economy consists of S > 1 sectors with Cobb–Douglas expenditure weights β_j,s and sector-specific trade elasticities ε_s < 0, the multi-sector ACR formula becomes Ŵ_j = ∏_s=1^S (λ̂_jj,s)^{β_j,s / ε_s}. Because λ_jj,s^1/ε_s is convex in 1/ε_s, low-elasticity differentiated inputs (|ε_s| ≈ 1.5–2.0) more than triple the estimated welfare gains from trade (to ~4%–5%).
2. Intermediate Inputs and Input-Output Linkages (Caliendo and Parro, 2015): When imported goods are used as intermediate inputs into domestic production (with labor value-added share γ_j ∈ (0, 1), typically γ_j ≈ 0.50), cheaper imported intermediates lower domestic production costs across consecutive stages of production, multiplying the ACR exponent by 1 / γ_j:

1. Non-Pareto Productivity Distributions (Melitz and Redding, 2015) and Variable Markups (Edmond, Midrigan, and Xu, 2015): Under lognormal or bounded-Pareto firm productivity distributions, the partial trade elasticity ε_ij varies endogenously with trade openness, and under oligopolistic competition trade liberalization compresses domestic markups, creating pro-competitive welfare gains outside the CES framework.

What are the two sufficient statistics in the ACR welfare formula?

The Arkolakis–Costinot–Rodríguez-Clare (ACR) welfare formula states that the proportional change in real income (Ŵ_j = W′_j/W_j) from any foreign trade shock equals Ŵ_j = λ̂_jj^1/ε. The only two parameters required to compute this welfare change are:

1. The proportional change in the domestic expenditure share (λ̂_jj = λ′_jj/λ_jj), where λ_jj is the fraction of country j's total spending allocated to domestically produced goods (1 minus the import penetration ratio); and
2. The structural gravity trade elasticity (ε < 0), which measures the percentage change in bilateral imports relative to domestic purchases when variable iceberg trade costs τ_ij change by 1%.

Why do the Krugman (1980) and Melitz (2003) models yield the same ACR welfare formula as the Armington (1969) model?

Conditional on observing the same change in aggregate domestic expenditure share λ̂_jj and calibrating each model to the same empirical gravity trade elasticity ε < 0, the additional microeconomic margins in Krugman (new imported varieties) and Pareto Melitz (firm selection and exporter entry) are already captured inside the observed macro trade elasticity ε, while the exit of displaced domestic varieties offsets the entry of marginal foreign varieties.

Does the ACR theorem imply that firm heterogeneity (Melitz, 2003) is irrelevant for trade policy?

No. As demonstrated by Melitz and Redding (2015), even when the ACR formula holds ex post, λ̂_jj is an endogenous outcome, so solving for counterfactual tariff changes ex ante or under non-Pareto firm productivity distributions requires modeling heterogeneous firms. Furthermore, once intermediate input linkages, multi-sector trade-elasticity heterogeneity (Ossa, 2015), or variable oligopolistic markups are included, firm- and sector-level margins amplify aggregate welfare gains beyond the one-sector ACR baseline.

How do intermediate inputs modify the ACR welfare formula?

When domestic production uses tradable intermediate inputs alongside domestic labor (with labor's share in gross output equal to γ_j ∈ (0, 1)), lower prices on imported intermediate goods reduce domestic marginal production costs in a Leontief input-output multiplier loop. This modifies the ACR formula to Ŵ_j = λ̂_jj^{1 / (γ_j ε)}, multiplying the logarithmic welfare gain from trade by the inverse of the value-added share 1/γ_j ≈ 2.

1. Arkolakis, C., Costinot, A., & Rodríguez-Clare, A. (2012). "New Trade Theories, Same Old Gains?" American Economic Review, 102(1), 94–130. https://doi.org/10.1257/aer.102.1.94
2. Melitz, M. J. (2003). "The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity." Econometrica, 71(6), 1695–1725. https://doi.org/10.1111/1468-0262.00467
3. Eaton, J., & Kortum, S. (2002). "Technology, Geography, and Trade." Econometrica, 70(5), 1741–1779. https://doi.org/10.1111/1468-0262.00352
4. Melitz, M. J., & Redding, S. J. (2015). "New Trade Models, New Welfare Implications." American Economic Review, 105(3), 1105–1146. https://doi.org/10.1257/aer.20130351
5. Costinot, A., & Rodríguez-Clare, A. (2014). "Trade Theory with Numbers: Quantifying the Consequences of Globalization." In G. Gopinath, E. Helpman, & K. Rogoff (Eds.), Handbook of International Economics (Vol. 4, pp. 197–261). Elsevier. https://doi.org/10.1016/B978-0-444-54314-1.00004-5