## 2.1 Setup (wpiea2020036-print-pdf) ## Source details **Canonical URL:** [2.1 Setup (wpiea2020036-print-pdf)](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020036-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020036-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020036-print-pdf.pdf.json) --- ### Model setup and price/transport-cost relationship - Trade costs τi j composed of ad valorem and per unit components: - τi j = πcif i j − πfob i j = (τi j − 1) πfob i j + Ti j - πcif i j and πfob i j are CIF and FOB prices; τi j is ad valorem component; Ti j is per unit component. - Trade costs are ad valorem only if Ti j = 0; per unit only if τi j = 1. If Ti j > 0, trade costs are less than proportional to the FOB price. - CIF–FOB price relationship: - πcif i j(τi j, Ti j, ci(θ)) = τi j πfob i j(τi j, Ti j, ci(θ)) + Ti j - ci(θ) is the marginal cost of firm i and increases with quality θ. ### Firm profit maximization, demand, and pricing - Exporter profit when exporting to country j: - πi j = (πfob i j − ci(θ)) qi j = [(πcif i j − Ti j)/τi j − ci(θ)] qi j - qi j is quantity sold by firm i in market j (depends on πcif i j). - Representative consumer utility (continuum of varieties ω): - U(qi j(ω)) = [∫− a(ω) (qi j(ω)) (σ−1)/σ dω] σ/(σ−1), with σ > 1 elasticity of substitution. - Inverse CES demand faced by firm i in country j: - πcif i j = μj qi j −1/σ i j θ(ω) (θ) (σ−1)/σ, where μj is a positive parameter (function of size and price index of destination); higher θ shifts demand up. - First-order condition and implied prices: - First-order condition yields: πcif i j = σ/(σ−1) (Ti j + τi j ci(θ)) - FOB price: πfob i j = 1/(σ−1) [Ti j/τi j + σ ci(θ)] - Markup over marginal cost: γfob i j = πfob i j − ci(θ) = 1/(σ−1) [Ti j/τi j + ci(θ)] ### Comparative statics and qualitative implications - If Ti j = 0 (ad valorem only), FOB price and markup do not depend on trade costs. - If Ti j > 0 (per unit costs present): - FOB price and markup depend positively on Ti j and negatively on τi j. - If τi j = 1 (per unit only), FOB price and markup increase with trade costs. - Interpretation for distance and tariffs: - Assume Ti j increases with distance; τi j independent of distance but varies with ad valorem tariffs. - Elasticities of FOB price and markup w.r.t. Ti j and τi j interpreted as elasticities w.r.t. distance and tariffs respectively. ### Elasticities (exact expressions) - With respect to per unit trade costs Ti j: - επfob T = 1/[1 + (σ ci(θ))/(Ti j/τi j)] > 0 - εγfob T = 1/[1 + (ci(θ))/(Ti j/τi j)] > 0 - Both increase with Ti j (and thus distance); magnitudes fall with quality since ci′(θ) > 0. - With respect to ad valorem trade costs τi j: - επfob τ = −1/[1 + (σ ci(θ))/(Ti j/τi j)] < 0 - εγfob τ = −1/[1 + (ci(θ))/(Ti j/τi j)] < 0 - Both decrease with τi j (and thus tariffs); magnitudes fall with quality. ### Theoretical predictions (stated) - Prediction 1: - Elasticities of FOB price and markup w.r.t. bilateral distance are positive; magnitude decreases with quality. - Empirical implication: positive coefficient on distance and negative coefficient on distance×quality when regressing FOB prices or markups. - Prediction 2: - Elasticities of FOB price and markup w.r.t. ad valorem tariffs are negative; magnitude decreases with quality. - Empirical implication: negative coefficient on tariffs and positive coefficient on tariffs×quality in the same regressions. ### Mechanisms behind predictions (demand elasticity) - Elasticity of demand to the FOB price εfob: - εfob = εcif [1 + (Ti j/τi j)/πfob i j] = −σ · {1 + 1/((σ/(σ−1))·(1/μj) + (τi j/Ti j)((σ/(σ−1))·(ci(θ))/Ti j))} (see equation (13) in source). - Key properties: - If Ti j = 0, elasticities to FOB and CIF prices coincide. - With per unit costs, εfob w.r.t. per unit trade costs (distance) is negative and increases with FOB price (hence with marginal cost ci(θ) and quality): - Interpretation: demand in more distant markets is less elastic to FOB price changes; exporters raise prices to offset higher transport costs but raise them less for higher-quality exports. - εfob w.r.t. ad valorem trade costs (tariffs) is positive and falls with FOB price and quality: - Interpretation: demand in higher-tariff countries is more elastic; exporters reduce prices to offset tariffs but reduce them less for higher-quality exports. ### Robustness to alternative demand systems - Demand systems examined: (a) CES, (b) quadratic non-separable (Ottaviano, Tabuchi, Thisse, 2002), (c) translog (Feenstra, 2003), (d) additively quasi-separable (Behrens and Murata, 2007). - Findings (preserving signs/qualitative patterns from source): - εfob w.r.t. per unit trade costs is negative with CES; can be negative with translog and additively quasi-separable depending on parameters; positive with quadratic preferences. For all demand systems, this elasticity increases with the FOB price. - εfob w.r.t. ad valorem trade costs is positive in all cases. It decreases with the FOB price for CES and translog, and for additively quasi-separable for some parameter values; it increases with the FOB price for quadratic preferences. - Conclusion: predictions are not specific to CES and hold for alternative utility functions except for quadratic preferences. - Empirical appendix (Appendix C) confirms: - εfob w.r.t. per unit trade costs (distance) is negative and rises with quality. - εfob w.r.t. ad valorem trade costs (tariffs) is positive and falls with quality. *Source: wpiea2020036-print-pdf — content as supplied.* --- ### Empirical framework, data, and descriptive statistics ### Empirical specification (unit values / markups) - Estimated reduced-form regression (equation (14)) for log FOB unit value ln uv_ijk,t: - ln uv_ijk,t = α1 ln dist_j + α2 ln dist_j × quality_k + α3 ln tar_j,t + α4 ln tar_j,t × quality_k + α5 ζ_j,t + Δ_k,t + ε_ijk,t - Quality measured using Wine Spectator ratings. - Distance dist_j and tariff tar_j,t (log of one plus tariff rate) interacted with quality. - Controls ζ_j,t include (log) GDP, GDP per capita, remoteness. - Product-time fixed effects Δ_k,t (and firm-destination-time fixed effects in more stringent specs). ### Data sources and sample construction (overview) - Argentinean customs transactions from Nosis for 2002–2009: exporter name, destination, shipment date, 12-digit HS code, FOB value (US dollars), volume (liters). - Wine product definition: name, type (red, white, rosé), grape, vintage year, container type. - Sample: 11,158 products corresponding to 8,361 different wines. - Unit values = value/volume (US dollars per liter); aggregated quarterly 2002Q1–2009Q4; shipments < 4.5 liters dropped; unit values > 100× median or < median/100 dropped. - Quality ratings: - Time-invariant Wine Spectator and Robert Parker ratings on (50,100) scale. - Wine Spectator match: 237 producers exporting 8,361 wines (11,158 products); scores range 55 to 97. - Parker match: 2,960 wines (4,128 products); scores vary 72 to 98. - Mean absolute difference WS vs Parker = 1.96; standard deviation = 3.19; Pearson correlation = 0.53; Kendall concordance = 0.36. - Wine Spectator used for main regressions; Parker for robustness. - Macroeconomic and trade-cost data: - Bilateral distances (km) from CEPII (population-weighted great circle distance between largest cities). - Bilateral ad valorem tariffs for HS 2204 from UNCTAD TRAINS (annual; effectively applied weighted average tariff rates in percentage terms). - Annual PPP GDPs and GDP per capita (constant 2011 US dollars) from World Bank WDI. - Quarterly nominal exchange rates from IMF IFS. - Country wine production and per-capita production/consumption from Anderson and Nelgen (2011). - Sample summary: - 237 exporters, 8,361 wines (11,158 products), 95 destination countries, 2002Q1–2009Q4, 91,810 observations; represents 69 percent of total value of red, white, rosé wine exported over period. - Growth 2002→2009: number of exporters, wines, products increased threefold (2002: 926 wines by 70 firms; 2009: 2,677 wines by 191 firms). ### Descriptive patterns (selected) - Correlation between unit values and quality = 36 percent. - Tariffs tend to rise with distance (correlation = 42 percent). - Unit values tend to fall with tariffs (correlation = −9.4 percent); unit values do not vary strongly with distance by region. - Regression of log unit values on product-time and destination fixed effects: country fixed effects vs log distance slope = 0.027 (significant at 1 percent). *Source: wpiea2020036-print-pdf — content as supplied.* --- ### Empirical results — Prices/Markups, Heterogeneity, and Mechanisms ### 4.1 Homogeneous trade cost effects (baseline) - Key baseline findings (Table 5, product-time fixed effects in column (3)): - lndistance: 0.021 (0.005)*** (column (3)) - Interpretation: If distance doubles, prices and markups increase by 2.74 and 1.47 percent on average (reported in source). - lntarifs: −0.086 (0.022)*** (column (3)) - At sample mean tariff = 14.87 percent, if tariffs double, prices and markups fall by 1.37 and 1.04 percent on average (reported in source). - lnremoteness: 0.011 (0.008) (column (3)) - lnGDP: −0.013 (0.002)*** (column (3)) - lnGDP/cap: 0.011 (0.005)** (column (3)) - R-squared: 0.838 (column (3)); Observations: 71,952 (column (3)). - Non-parametric distance intervals (column (4), omitted interval < 2,900 km): - 2,900–7,700 km: markups increase by 0.016 log points on average. - 7,700–14,200 km: markups increase by 0.043 log points. - >14,200 km: markups increase by 0.061 log points. - Decomposition: - Variable markups account for around half of the impact of distance on within-firm price variation across markets; selection/composition effects account for the other half. - Markups explain three-quarters of the effect of tariffs on variation in firm-level prices across markets; remaining quarter due to selection/composition. ### 4.2 Heterogeneous trade cost effects by quality (Table 6) - Main interaction results (column (2) emphasized): - lndistance coefficient positive; lndistance × quality coefficient negative. - lntarifs coefficient negative; lntarifs × quality coefficient positive. - Conclusion: Elasticities of markups w.r.t. trade costs are smaller in magnitude for higher quality exports. - Magnitudes across quality distribution (based on column (2)): - Quality range in sample: between 55 and 97. - At mean quality = 85: - Distance elasticity = 0.022. - Tariff elasticity = −0.094. - At 5th percentile quality = 79: - Distance elasticity = 0.052. - Tariff elasticity = −0.227. - At 95th percentile quality = 91: - Distance elasticity = −0.007 (insignificant). - Tariff elasticity = 0.039 (insignificant). - For quality > 91, distance elasticity becomes negative and tariff elasticity positive (quality dampening outweighs trade-cost effects). - Selected coefficients reported: - lndistance: 0.446 (0.061)*** (column (2)) - lndistance × quality: −0.005 (0.001)*** (column (2)) - lntarifs: −0.087 (0.022)*** (column (2)) - lntarifs × quality: reported in table layouts (interaction patterns noted in source). - R-squared: 0.838 (columns 1–2); Observations: 71,952 (cols 1–2). ### Robustness and alternative specifications - Firm-destination-time fixed effects (column (3) of Table 6): distance × quality remains negative and tariff × quality positive. - Product definition robustness (column (4)): results robust when defining product by name, grape, type, vintage only (ignore packaging). ### Alternative mechanisms examined - Intensity of competition (σ estimated by quality): - Estimated σ by quality categories (2004–2009): - Low quality: 17.71. - Medium quality: 11.82. - High quality: 8.37. - Destination-quality varying σ: 75 elasticities range 1.41 to 41.73, average 9.92, standard deviation 9.27. - Regression: σ falls with quality and tariffs, increases with distance (all coefficients significant at 1%). - Controlling for σ does not alter main conclusions. - Country-level characteristics and firm-level interactions: - Interactions (exchange rate, foreign real GDP per capita, wine production/consumption, import share) with quality included; heterogeneous trade-cost effects by quality remain. - Selection correction (three-step estimator and Heckman two-step) implemented; results robust to selection bias. *Source: wpiea2020036-print-pdf — empirical sections 4 and robustness material as supplied.* --- ### Extensions, export volumes, and income heterogeneity ### 5.1 Income heterogeneity across destinations (selected) - Sample split by destination income (above/below sample median; World Bank GNI threshold 4,035 US dollars in 2011 yields similar results). - Interaction results (Table 9, columns (1)–(2)): - lndistance×quality = −0.003 (0.001) - lndistance×quality×rich = −0.001 (0.000) - lntariffs×quality = 0.004 (0.005) in column (1) and −0.007 (0.009) in column (2) - lntariffs×quality×rich = 0.036 (0.007) in column (1) and 0.038 (0.009) in column (2) - R-squared = 0.922; Observations = 66,941. - Interpretation: - Distance×quality interaction significant for both rich and poor groups and larger in magnitude for richer destinations. - Tariff×quality heterogeneity present for richer markets only. - Heterogeneous effects of trade costs on markups are stronger for exports to richer destinations. ### 5.2 Firm-level characteristics (Table 10 summary) - Heterogeneity by firm quality, firm size, and market share (interacting distance and tariff effects with dummies). - Selected interaction coefficients (standard errors in parentheses): - Firm quality (column (1)): - lndistance×quality×above median = −0.005 (0.001) - lntariffs×quality×above median = 0.033 (0.008) - Firm size (column (2)): - lndistance×quality×above median = −0.005 (0.001) - lntariffs×quality×above median = 0.017 (0.009) - Market shares (column (3)): - lntariffs×quality×above median = 0.026 (0.009) - R-squared = 0.922; Observations = 66,941 in all columns. - Interpretation: - Heterogeneous distance and tariff effects driven mainly by high-performance firms: larger firms, higher market-share firms, and firms exporting higher quality. - Reported correlations: larger firms have higher market shares correlation = 0.69; correlation between firm-level quality and market shares = 0.09; firm-level quality and firm size = 0.10. ### 5.3 Manufacturing industries (Table 11 summary) - Universe: Argentinean firm-level exports 2002–2009; manufacturing industries HS 30–38 and 42–97; 8-digit HS product definition; quality estimated via Khandelwal (2010) procedure. - Sample: 9,724 exporters, 3,155 products, 92 destination countries, 462,127 observations. - Selected estimates (column (4) with firm-destination-time fixed effects): - quality = 0.062 (0.010) - lndistance×quality = −0.005 (0.001) - lntariffs×quality = 0.127 (0.039) - lntariffs = −0.007 (0.093) - R-squared = 0.965; Observations = 154,760. - Interpretation: - Unit values rise with distance, income per capita, remoteness; fall with country size and tariffs. - Magnitude of distance and tariff effects on unit values is smaller for higher quality exports; result robust to firm-destination-time fixed effects. ### 5.4 Export volumes — model and empirical findings (Table 12) - Model predictions (equations (16) and (17)): - ∂ ln θ / ∂ ln T = −σ × (1 + (χ_i(α) T_ij / τ_ij)) < 0 - ∂ ln θ / ∂ ln τ = −σ × (1 + (T_ij / τ_ij) χ_i(α)) < 0 - Implications: quantities fall with T_ij (distance), especially for lower quality exports; quantities fall with τ_ij (tariffs) to a larger extent for higher quality exports. - Empirical specification uses PPML to include zero trade flows; balanced sample of all firm-wine-destination-time combinations; drop years prior to vintage. - Selected PPML estimates (columns (1)–(3)): - Column (1): - lndistance = −1.036 (0.069) - lntariffs = −2.346 (0.245) - Observations = 2,472,069. - Column (2): - lndistance = −2.399 (0.372) - lndistance×quality = −0.016 (0.004) - lntariffs = 8.098 (2.329) - lntariffs×quality = −0.124 (0.028) - Observations = 2,472,069. - Column (3) (adds firm-destination-time fixed effects): - lndistance = −0.352 (0.033) - lndistance×quality = 0.018 (0.006) - lntariffs = −0.394 (0.158) - lntariffs×quality = −0.219 (0.028) - Observations = 421,691. - OLS results for positive flows only (columns (4)–(6)) show qualitatively similar signs. - Point estimates at mean, 5th, and 95th percentiles of quality distribution (based on column (2)): - Distance elasticity = −1.040 (mean), −1.142 (5th percentile), −0.932 (95th percentile). - Tariff elasticity = −2.314 (mean), −1.540 (5th percentile), −3.147 (95th percentile). - All reported as significant at the one percent level. - Interpretation: - Export volumes fall with distance and tariffs. - Distance elasticity falls with quality; tariff elasticity increases with quality. - Heterogeneities persist with firm-destination-time fixed effects. *Source: wpiea2020036-print-pdf — Sections 4–5 empirical findings as supplied.* --- ### Robustness, selection, and measurement ### Robustness exercises (Appendix F, selected) - Alternative samples/specifications (Table F1): - Including wholesalers/retailers: results hold; share handled by intermediaries = 4.80 percent in 2002 and 5.33 percent in 2009. - Instrument tariffs×quality with (log(1+total wine production) × quality); Kleibergen-Paap F statistic = 83 rejects weak instrument null. - Excluding year 2002 (Argentina crisis), controlling for port of exit, export volumes and interactions with quality, annual unit values — conclusions unchanged. - Measurement of quality (Table F2): - Using Parker ratings; rescaling quality to 1–6 Wine Spectator bins; excluding “Great” wines; excluding US — results robust. - Instrumenting Wine Spectator with Parker scores (both interacted) — Kleibergen-Paap F statistic = 404 rejects weak instruments. - Estimating quality at 12-digit HS via Khandelwal (2010) — sample more than doubles; results hold. - Cross-sectional variation (Table F3): - For 2004–2009, markups rise with distance and fall with tariffs; effects stronger for lower quality exports. Results for 2002–2003 are insignificant (possible crisis/coverage issues). ### Selection bias and correction (Appendix D) - Three-step estimator (Harrigan et al., 2015) and Heckman two-step implemented. - Selection control coefficient ≈ 0.044 (0.007) positive and significant. - Controlling for selection yields economically and statistically similar benchmark findings. - Example: in column (1) of Table D1, a doubling of distance induces an exporter to increase its markup by 2.3 percent on average. - Conclusion: results robust to selection bias across firms. ### Appendix A — Evidence on per unit vs ad valorem trade costs (freight regression) - Regression for ln φ_{i j k, t} = log freight charges per liter on ln unit value, ln dist_j, ln q_{i j k, t}, firm-time and product characteristics fixed effects. - Sample: 107 firms, 1,440 wines (1,534 products), 82 export markets, 2005Q4–2009Q4, 6,802 observations (coverage incomplete). - Selected empirical results (Table A1): - ln unit value coefficients (IV specifications): 0.713 (0.043), 0.667 (0.048) (IV variants reported). - ln distance coefficients: 0.062 (0.031) (col 1); 0.069 (0.030) (col 2); 0.068 (0.030) (col 3). - ln export volume coefficients (negative): −0.121 (0.012); −0.133 (0.014); −0.187 (0.022) across columns. - ln unit value × ln distance interaction negative and significant in IV columns: e.g., −0.107 (0.035), −0.096 (0.035). - Unit value elasticity estimates (examples): mean 0.656 (0.048) in one IV; 5th percentile 0.744 (0.049); 95th percentile 0.569 (0.065). - Kleibergen-Paap F statistics reported (e.g., 346.35; 196.90; 147.73) indicate instrument relevance in IVs. - Interpretation: - Freight costs rise with distance and fall with export volumes. - Freight costs increase with unit values but with elasticity strictly less than unity (evidence of mixed ad valorem and per unit components). - Interaction implies freight costs become more per unit relative to ad valorem as distance increases. ### Appendix C — Empirical estimates of elasticity of demand to FOB price - Demand elasticity εfob estimated via specification (equations C1–C2) with instruments for unit values. - Selected estimates (Table C1, IV column (2)): - lnunit value: −11.024 (4.105)*** (IV) - lnunit value×lndistance (α2): 1.457 (0.494)*** (IV) - lnunit value×lntariffs (β2): −6.173 (3.683) (IV) - lnunit value×lndistance×quality (α4): −0.016 (0.006)*** (IV) - lnunit value×lntariffs×quality (β4): 0.068 (0.040) (IV) - Kleibergen-Paap F–178.76 for IV. - Elasticities derived (IV column (2)): - Elasticity of εfob w.r.t. distance = −0.137 (mean), −0.280 (5th), −0.033 (95th). - Elasticity of εfob w.r.t. tariffs = 0.445 (mean), 1.037 (5th), 0.012 (95th). - Conclusion: εfob w.r.t. distance negative and increases with quality; εfob w.r.t. tariffs positive and decreases with quality. *Source: wpiea2020036-print-pdf — Appendices A–D and C empirical content as supplied.* --- ### Conclusions and implications - Core empirical conclusions: - Firms raise markups in more distant markets (per unit trade costs effect). - Firms lower markups in high-tariff countries (ad valorem trade costs effect). - Heterogeneity: effects of trade costs on markups are smaller in magnitude for higher quality exports. - Heterogeneity stronger for exports to richer countries and driven mainly by higher-quality firms, larger firms, and firms with larger export market shares. - Broader implications: - Variation in firm-level export prices across markets is driven not only by quality differences but also by markup variation conditional on quality. - Firms price discriminate across destinations; extent depends on size and nature of trade costs (per unit versus ad valorem) and the quality exported. - Trade models assuming markups invariant to country-level characteristics miss a key channel; findings support models with variable markups dependent on trade costs and emphasize modeling per unit trade costs. - Suggested next step (from source): - Understanding the welfare implications of these results is an important next step. *Source: wpiea2020036-print-pdf — concluding sections and suggested next steps as supplied.* ### 2.1 Setup ### 2.1 Setup ### Model setup and price/transport-cost relationship - Trade costs τi j are modeled with ad valorem and per unit components: - τi j = πcif i j − πfob i j = (τi j − 1) πfob i j + Ti j - πcif i j and πfob i j are CIF and FOB prices; τi j is ad valorem component; Ti j is per unit component. Trade costs are ad valorem only if Ti j = 0; per unit only if τi j = 1. If Ti j > 0, trade costs are less than proportional to the FOB price. - CIF–FOB relationship: - πcif i j(τi j, Ti j, ci(θ)) = τi j πfob i j(τi j, Ti j, ci(θ)) + Ti j - ci(θ) is the marginal cost of firm i which increases with quality θ. ### Firm profit maximization and demand - Exporter profit when exporting to country j: - πi j = (πfob i j − ci(θ)) qi j = [πcif i j − Ti j)/τi j − ci(θ)] qi j - qi j is quantity sold by firm i in market j (depends on πcif i j). - Representative consumer utility over continuum of varieties ω: - U(qi j(ω)) = [∫− a(ω) (qi j(ω)) (σ−1)/σ dω] σ/(σ−1) - σ > 1 is the elasticity of substitution. - Inverse CES demand faced by firm i in country j: - πcif i j = μj qi j −1/σ i j θ(ω) (θ) (σ−1)/σ - μj is a positive parameter (function of size and price index of destination). Higher quality θ shifts demand up. ### First-order condition, FOB price, and markup - First-order condition from profit maximization combining demand: - πcif i j = σ/(σ−1) (Ti j + τi j ci(θ)) - FOB price: - πfob i j = 1/(σ−1) [Ti j/τi j + σ ci(θ)] - Corresponding markup over marginal cost: - γfob i j = πfob i j − ci(θ) = 1/(σ−1) [Ti j/τi j + ci(θ)] - Comparative statics observations: - If Ti j = 0 (ad valorem only), FOB price and markup do not depend on trade costs. - If per unit costs exist (Ti j > 0), FOB price and markup depend positively on Ti j and negatively on τi j. - If trade costs are per unit only (τi j = 1), FOB price and markup increase with trade costs. ### Predictions on trade costs, distance, tariffs, and quality - Assumptions: - Ti j increases with distance; τi j independent of distance but varies with ad valorem tariffs. - Elasticities of FOB price and markup w.r.t. Ti j and τi j interpreted as elasticities w.r.t. distance and tariffs respectively. - Elasticities with respect to per unit trade costs Ti j: - επfob T = 1/[1 + (σ ci(θ))/(Ti j/τi j)] > 0 - εγfob T = 1/[1 + (ci(θ))/(Ti j/τi j)] > 0 - Both price and markup increase with per unit trade costs (and thus distance); magnitudes fall with quality because ci′(θ) > 0. - Prediction 1: - The elasticities of the FOB price and markup with respect to bilateral distance are positive, and their magnitude decreases with quality. - Empirical implication: in regressions explaining FOB prices or markups, expect a positive coefficient on distance and a negative coefficient on the interaction between distance and quality. - Elasticities with respect to ad valorem trade costs τi j: - επfob τ = −1/[1 + (σ ci(θ))/(Ti j/τi j)] < 0 - εγfob τ = −1/[1 + (ci(θ))/(Ti j/τi j)] < 0 - Both price and markup decrease with ad valorem trade costs (and thus tariffs); magnitudes fall with quality. - Prediction 2: - The elasticities of the FOB price and markup with respect to ad valorem tariffs are negative, and their magnitude decreases with quality. - Empirical implication: in regressions explaining FOB prices or markups, expect a negative coefficient on tariffs and a positive coefficient on the interaction between tariffs and quality. ### Mechanisms behind predictions - Key mechanism: per unit trade costs generate an elasticity of demand to the FOB price that depends on trade costs and quality. - Elasticity of demand to the FOB price εfob: - εfob = εcif [1 + (Ti j/τi j)/πfob i j] = −σ · {1 + 1/((σ/(σ−1))·(1/μj) + (τi j/Ti j)((σ/(σ−1))·(ci(θ))/Ti j))} (see equation (13) for exact expression) - If Ti j = 0, elasticities to FOB and CIF prices coincide. - With per unit costs, εfob w.r.t. per unit trade costs (distance) is negative and increases with FOB price (hence with marginal cost ci(θ) and quality). - Interpretation: demand in more distant markets is less elastic to FOB price changes; exporters raise prices to offset higher transport costs but raise them less for higher-quality exports. - εfob w.r.t. ad valorem trade costs (tariffs) is positive and falls with FOB price and quality. - Interpretation: demand in higher-tariff countries is more elastic; exporters reduce prices to offset tariffs but reduce them less for higher-quality exports. ### Robustness to alternative demand systems - The two mechanisms (i) εfob w.r.t. per unit trade costs negative and increasing with FOB price, and (ii) εfob w.r.t. ad valorem trade costs positive and decreasing with FOB price, are examined for alternative preferences: - Demand systems considered: (a) CES, (b) quadratic non-separable (Ottaviano, Tabuchi, Thisse, 2002), (c) translog (Feenstra, 2003), (d) additively quasi-separable (Behrens and Murata, 2007). - Findings: - εfob w.r.t. per unit trade costs is negative with CES; can be negative with translog and additively quasi-separable depending on parameters; positive with quadratic preferences. For all demand systems, this elasticity increases with the FOB price. - εfob w.r.t. ad valorem trade costs is positive in all cases. It decreases with the FOB price for CES and translog, and for additively quasi-separable for some parameter values; it increases with the FOB price for quadratic preferences. - Conclusion: predictions are not specific to CES and hold for alternative utility functions except for quadratic preferences. - Empirical appendix results (Appendix C): - εfob w.r.t. per unit trade costs (distance) is negative and rises with quality. - εfob w.r.t. ad valorem trade costs (tariffs) is positive and falls with quality. ### Data and descriptive statistics (overview from sections 3.1–3.4) - Data sources combined: firm-level customs data, wine experts quality ratings, and macroeconomic indicators. - Customs data: - Argentinean customs transactions purchased from Nosis for 2002–2009. - For each transaction: exporter name, destination, shipment date, 12-digit HS code, FOB value (US dollars), and volume (liters). - Wine product definition: name, type (red, white, rosé), grape, vintage year, and container type. Sample: 11,158 products corresponding to 8,361 different wines. - Unit values computed as value/volume in US dollars per liter. - Data aggregated at quarterly frequency between 2002Q1 and 2009Q4; shipments below 4.5 liters dropped; observations with unit value > 100× median or < median/100 dropped. - Quality ratings: - Time-invariant ratings from Wine Spectator and Robert Parker on (50,100) scale. - Wine Spectator rating bins and Parker bins described (e.g., Great 95–100, Extraordinary 96–100, etc.). - Matching results: - Wine Spectator match: 237 producers exporting 8,361 wines (11,158 products), 1,066 different names, 3 types, 24 grapes, 22 vintage years (1977–2009). Scores range 55 to 97. - Parker match: 2,960 wines (4,128 products) by 151 firms; scores vary between 72 and 98. - Mean absolute difference between Wine Spectator and Parker = 1.96; standard deviation = 3.19; Pearson correlation = 0.53; Kendall concordance = 0.36. - Wine Spectator used for main regressions; Parker used for robustness. - Macroeconomic data: - Bilateral distances (km) from CEPII (population-weighted great circle distance between largest cities). - Bilateral ad valorem tariffs for HS 2204 from UNCTAD TRAINS (annual; effectively applied weighted average tariff rates in percentage terms). - Annual PPP GDPs and GDP per capita (constant 2011 US dollars) from World Bank WDI. - Quarterly nominal exchange rates from IMF IFS. - Country wine production and per-capita production/consumption from Anderson and Nelgen (2011). - Sample summary and descriptive patterns: - Sample: 237 exporters, 8,361 wines (11,158 products), 95 destination countries, 2002Q1–2009Q4, 91,810 observations; represents 69 percent of total value of red, white, rosé wine exported over period. - Growth between 2002 and 2009: number of exporters, wines, and products increased threefold (e.g., 2002: 926 wines by 70 firms; 2009: 2,677 wines by 191 firms). - Table 3 (Wine Spectator bins) highlights: - “Good” and “Very good” wines largest share of sample. - “Great” wines exported to fewer but richer destinations and to more distant locations; higher quality wines are on average more expensive. Correlation between unit values and quality = 36 percent. - Table 4 (by continent) highlights: - North America is main destination by number of exporters, wines, export share; closer to Argentina, lower tariffs. - South America: closest, lowest tariffs, but only imports 13 percent of exports due to low GDP and GDP per capita. - Africa: negligible export share, smallest and poorest region, high tariffs. - Tariffs tend to rise with distance (correlation = 42 percent). - Unit values tend to fall with tariffs (correlation = −9.4 percent); unit values do not vary strongly with distance by region. - Regression of log unit values on product-time and destination fixed effects: - Product-time fixed effects identify variation in markups; country fixed effects capture mean markup in each destination. - Figure 1 (country fixed effects vs log distance): slope = 0.027 (significant at 1 percent). If distance doubles, markups increase by (slope × ln 2) implied in figure (slope reported exactly as 0.027). *Source: wpiea2020036-print-pdf — content as supplied.* ### 1.9 percent on average(2 ### wpiea2020036-print-pdf - 1.9 percent on average(2 ### Empirical framework and identification - Estimated reduced-form regression for log FOB unit value ln uv_ijk,t: - ln uv_ijk,t = α1 ln dist_j + α2 ln dist_j × quality_k + α3 ln tar_j,t + α4 ln tar_j,t × quality_k + α5 ζ_j,t + Δ_k,t + ε_ijk,t (equation (14)). - Quality measured using Wine Spectator ratings. - Distance dist_j and (annual) tariff tar_j,t (log of one plus tariff rate) interacted with quality. - Controls ζ_j,t include (log) GDP, GDP per capita, remoteness. - Product-time fixed effects Δ_k,t (and firm-destination-time fixed effects in more stringent specs) used to isolate variation in unit values across destinations for a given exporter and product. - Predictions derived from model: - Prediction 1: α1 + (α2 × quality_k) > 0 with α2 < 0 (prices/markups rise with distance, effect smaller for higher quality). - Prediction 2: α3 + (α4 × quality_k) < 0 with α4 > 0 (prices/markups fall with tariffs, effect smaller for higher quality). ### 4.1 Homogeneous trade cost effects (key findings) - Baseline regressions (Table 5) omitting interactions with quality find: - Distance increases unit values; tariffs decrease unit values. - Prices higher in richer and remote destinations; lower in larger markets. - With product-time fixed effects (column (3)): - Distance coefficient falls from 0.039 (column (2)) to 0.021 (column (3)). - Interpretation: If distance doubles, prices and markups increase by 2.74 and 1.47 percent on average. - Variable markups account for around half of the impact of distance on within-firm price variation across markets; selection/composition effects account for the other half. - Tariff coefficient falls from −0.113 to −0.086. - If tariffs double (from their mean equal to 14.87 percent), prices and markups fall by 1.37 and 1.04 percent on average. - Markups explain three-quarters of the effect of tariffs on variation in firm-level prices across markets; remaining quarter due to selection/composition. - Non-parametric distance intervals (column (4), omitted first interval < 2,900 km): - 2,900–7,700 km: markups increase by 0.016 log points on average. - 7,700–14,200 km: markups increase by 0.043 log points. - >14,200 km: markups increase by 0.061 log points. - Table 5 reported coefficients and statistics: - lndistance: 0.042 (0.008)*** (col 1); 0.039 (0.008)*** (col 2); 0.021 (0.005)*** (col 3). - lntarifs: −0.115 (0.040)*** (col 1); −0.113 (0.040)*** (col 2); −0.086 (0.022)*** (col 3). - lnremoteness: 0.048 (0.013)*** (col 1); 0.047 (0.013)*** (col 2); 0.011 (0.008) (col 3). - lnGDP: −0.024 (0.002)*** (col 1–2); −0.013 (0.002)*** (col 3). - lnGDP/cap: 0.020 (0.007)*** (col 1); 0.021 (0.007)*** (col 2); 0.011 (0.005)** (col 3). - R-squared: 0.584 (col 1); 0.596 (col 2); 0.838 (col 3). - Observations: 91,307 (cols 1–2); 71,952 (cols 3–4). ### 4.2 Heterogeneous trade cost effects by quality - Interaction results (Table 6, column (2) emphasized): - Distance coefficient positive and distance × quality coefficient negative. - Tariff coefficient negative and tariff × quality coefficient positive. - Conclusion: Elasticities of markups with respect to trade costs are smaller in magnitude for higher quality exports. - Magnitudes across quality distribution (based on column (2)): - Quality range in sample: between 55 and 97. - At mean quality = 85: - Distance elasticity = 0.022. - Tariff elasticity = −0.094. - At 5th percentile quality = 79: - Distance elasticity = 0.052. - Tariff elasticity = −0.227. - At 95th percentile quality = 91: - Distance elasticity = −0.007 (insignificant). - Tariff elasticity = 0.039 (insignificant). - For quality levels above 91, distance elasticity becomes negative and tariff elasticity positive (quality dampening outweighs trade-cost effects). - Table 6 selected coefficients: - lndistance: 0.303 (0.056)*** (col 1); 0.446 (0.061)*** (col 2). - lndistance × quality: −0.003 (0.001)*** (col 1); −0.005 (0.001)*** (col 2). - lntarifs: −0.087 (0.022)*** (col 2). - lntarifs × quality: −0.022 (0.004)*** (column label shows sign; column (2) reports 0.027 (0.004)*** as table layout indicates interaction patterns—refer to table entries). - R-squared: 0.838 (cols 1–2); 0.922 (col 3); 0.909 (col 4). - Observations: 71,952 (cols 1–2); 66,941 (col 3); 69,219 (col 4). - Robustness: - Column (3) estimates equation (15) with firm-destination-time fixed effects; distance × quality remains negative and tariff × quality positive. - Column (4): results robust when product defined by wine’s name, grape, type, vintage only (ignore packaging). ### 4.3 Alternative mechanisms considered and conclusions - Intensity of competition (elasticity of substitution σ): - Estimated σ by quality categories (2004–2009): - Low quality: 17.71. - Medium quality: 11.82. - High quality: 8.37. - Destination-quality varying σ: 75 elasticities range from 1.41 to 41.73, average 9.92, standard deviation 9.27. - Regression shows σ falls with quality and tariffs, increases with distance (all coefficients significant at 1%). - Table 7 highlights interactions with ln σ: - lndistance × ln σ: −0.027 (0.006)*** (col 1). - lntarifs × ln σ: −0.197 (0.039)*** (col 1). - ln σ coefficient: −0.200 (0.049)*** (col 1). - Controlling for σ does not alter main conclusions: firms raise markups in more distant markets, lower them in high-tariff destinations; effects stronger for lower quality wines. - R-squared in specifications: 0.845 (col 1); 0.928 (col 2); 0.922 (col 3). - Country-level characteristics: - Interacted exchange rate (ln ER) with quality: lnER × quality coefficient positive (0.001 (0.000)**), inclusion does not change distance/tariff × quality effects. - Interaction of foreign real GDP per capita × quality: positive coefficient but distance/tariff heterogeneous effects persist. - Interaction with wine production per capita, wine consumption per capita, and wine import share × quality included in separate specs; heterogeneous trade-cost effects by quality remain. - Table 8 selected coefficients: - lndistance × quality: −0.004 (0.001)*** (col 1); −0.005 (0.001)*** (col 2); −0.009 (0.001)*** (col 5). - lntarifs × quality: 0.026 (0.004)*** (col 1); 0.029 (0.004)*** (col 2); 0.015 (0.004)*** (col 5). - R-squared: 0.922–0.925 across columns. - Selection correction: - Implemented three-step estimator of Harrigan et al. (2015) and Heckman two-step; results robust to removing selection bias across firms. ### Extensions (brief) - Extensions discussed include: - Heterogeneity of results by destination country income level. - Differences across types of exporters. - Extension to manufacturing industries other than wine. - Derivation and testing of model predictions for export volumes. *Source: wpiea2020036-print-pdf* ### 5.1 Income Heterogeneity across Destinations ### 5.1 Income Heterogeneity across Destinations ### Main findings - Consumers in richer countries are generally assumed to have a stronger preference for higher quality goods (citations in source). - Sample split: countries classified as rich or poor based on whether income per capita is above or below the sample median; alternative split using World Bank GNI threshold of 4,035 US dollars in 2011 yields similar results. - Interaction results (Table 9, column (1) and (2)): - lndistance×quality = -0.003 (0.001) - lndistance×quality×rich = -0.001 (0.000) - lntariffs×quality = 0.004 (0.005) in column (1) and -0.007 (0.009) in column (2) - lntariffs×quality×rich = 0.036 (0.007) in column (1) and 0.038 (0.009) in column (2) - R-squared = 0.922; Observations = 66,941. - Interpretation: - The coefficient on the interaction of distance with quality is significant for both rich and poor destination groups and larger in magnitude for richer destinations. - The effect of tariffs on markups is heterogeneous for richer markets only. - Heterogeneous effects of trade costs on markups are stronger for exports to richer destinations. --- ### 5.2 Firm-level Characteristics ### Research design - Heterogeneity examined by interacting distance and tariff effects with dummies for: - higher-quality firms (above median average firm-level quality), - larger firms (total yearly exports above median), - firms with larger export market shares (above median of firm’s share of destination-year total exports). - Firm size used as proxy for productivity when direct productivity measures are unavailable. ### Key empirical results (Table 10) - Column (1) — Firm quality: - lndistance×quality = -0.001 (0.001) - lndistance×quality×above median = -0.005 (0.001) - lntariffs×quality = 0.007 (0.005) - lntariffs×quality×above median = 0.033 (0.008) - Column (2) — Firm size: - lndistance×quality = 0.001 (0.001) - lndistance×quality×above median = -0.005 (0.001) - lntariffs×quality = 0.014 (0.007) - lntariffs×quality×above median = 0.017 (0.009) - Column (3) — Market shares: - lndistance×quality = -0.003 (0.001) - lndistance×quality×above median = 0.001 (0.000) - lntariffs×quality = 0.008 (0.007) - lntariffs×quality×above median = 0.026 (0.009) - R-squared = 0.922 in all columns; Observations = 66,941 in all columns. ### Interpretation - Interactions of distance and tariffs with quality are only significant for higher-quality firms (column (1)). - Distance effects are heterogeneous for large firms only (column (2)); tariff heterogeneity is present for both size groups but larger for bigger firms. - Distance effects are equally heterogeneous for high and low market-share firms, while tariff effects are heterogeneous for high market-share firms only. - Summary statement from source: heterogeneous effects of distance and tariffs on markups are mainly driven by high performance firms: larger firms, firms with higher market shares, and firms exporting higher quality. - Correlations reported: larger firms have higher market shares correlation = 0.69; correlations between firm-level quality and market shares = 0.09; firm-level quality and firm size = 0.10. --- ### 5.3 Manufacturing Industries ### Data and methods - Universe: Argentinean firm-level exports from Nosis, 2002–2009. - Observables: exporter name, destination, transaction date, 12-digit HS, FOB value (US dollars), mass (kg). - Aggregation: quarterly frequency; unit values in US dollars per kilogram. - Focus: manufacturing industries (HS 30–38 and 42–97). Product defined at 8-digit HS level. - Quality estimation: Khandelwal (2010) procedure adapted (Bernini and Tomasi, 2015; Chen and Juvenal, 2018); see Appendix E in source. - Tariffs: weighted average tariff rates (percentage terms) from UNCTAD TRAINS at 6-digit HS and annual frequency. - Sample: 9,724 exporters, 3,155 products, 92 destination countries, 462,127 observations. - Controls: firm-product-time fixed effects; final column adds firm-destination-time fixed effects. ### Key empirical results (Table 11) - Column (1): - lndistance = 0.035 (0.003) - lnremoteness = 0.031 (0.010) - lnGDP = -0.028 (0.002) - lnGDP/cap = 0.035 (0.006) - lntariffs = -0.101 (0.046) - R-squared = 0.930; Observations = 215,476. - Column (2): - quality = -0.037 (0.002) - lndistance = 0.035 (0.003) - lnremoteness = 0.040 (0.011) - lnGDP = -0.028 (0.002) - lnGDP/cap = 0.035 (0.006) - R-squared = 0.930; Observations = 215,476. - Column (3): - lndistance×quality = -0.003 (0.001) - lntariffs×quality = -0.102 (0.035) - quality = 0.056 (0.008) - other coefficients similar to above - R-squared = 0.930; Observations = 215,476. - Column (4) (adds firm-destination-time fixed effects): - quality = 0.062 (0.010) - lndistance×quality = -0.005 (0.001) - lntariffs×quality = 0.127 (0.039) - lntariffs = -0.007 (0.093) - R-squared = 0.965; Observations = 154,760. ### Interpretation - Unit values rise with distance, income per capita, and remoteness, and fall with a country’s size and tariffs (column 1). - Higher quality goods are exported at a higher price (column 2). - The magnitude of the effects of distance and tariffs on unit values is smaller for higher quality exports (column 3); result robust to firm-destination-time fixed effects (column 4). --- ### 5.4 Export Volumes ### Model predictions (equations (16) and (17)) - Elasticity of export quantities θ_ij with respect to T_ij: - ∂ ln θ / ∂ ln T = -σ × (1 + (χ_i(α) T_ij / τ_ij)) < 0 (equation (16)) - Elasticity of export quantities θ_ij with respect to τ_ij: - ∂ ln θ / ∂ ln τ = -σ × (1 + (T_ij / τ_ij) χ_i(α)) < 0 (equation (17)) - Implications: - Quantities fall with T_ij (and therefore with distance), especially for lower quality exports. - Quantities fall with τ_ij (and thus with tariffs), but to a larger extent for higher quality exports. ### Empirical specification - Reduced-form regression (equation (18)) includes ln distance, ln distance×quality, ln tariffs, ln tariffs×quality, controls z_jt, product-time fixed effects Δ_k,t. Estimation uses Poisson Pseudo-Maximum Likelihood (PPML) to include zero trade flows; balanced sample of all firm-wine-destination-time combinations; drop years prior to wine vintage. ### Key empirical results (Table 12) - Columns (1)–(3): PPML estimates; dependent variable FOB export volume (liters). - Column (1): - lndistance = -1.036 (0.069) - lntariffs = -2.346 (0.245) - lnremoteness = -0.713 (0.109) - lnGDP = 0.737 (0.019) - lnGDP/cap = 0.942 (0.062) - Observations = 2,472,069; Zero observations included: Yes. - Column (2): - lndistance = -2.399 (0.372) - lndistance×quality = -0.016 (0.004) - lntariffs = 8.098 (2.329) - lntariffs×quality = -0.124 (0.028) - lnGDP and lnGDP/cap same as column (1) - Observations = 2,472,069; Zero observations included: Yes. - Column (3) (adds firm-destination-time fixed effects): - lndistance = -0.352 (0.033) - lndistance×quality = 0.018 (0.006) - lntariffs = -0.394 (0.158) - lntariffs×quality = -0.219 (0.028) - Observations = 421,691; Zero observations included: Yes. - Columns (4)–(6): OLS estimates; dependent variable log FOB export volume for positive flows only. - Column (4): - lndistance = -0.858 (0.267) - lntariffs = 8.632 (1.631) - lnGDP = -0.281 (0.009) - lnGDP/cap = -0.413 (0.034) - Observations = 71,952; Zero observations included: No. - Column (5): - lndistance×quality = -0.006 (0.003) - lntariffs×quality = -0.106 (0.019) - Observations = 71,952; Zero observations included: No. - Column (6) (adds firm-destination-time fixed effects): - lndistance×quality = -0.004 (0.003) - lntariffs×quality = -0.079 (0.017) - Observations = 66,941; Zero observations included: No. - R-squared: columns (4)–(6) report 0.555, 0.555, 0.778 respectively for OLS specifications; PPML columns report “–” for R-squared where not applicable. - Notes: Robust standard errors clustered by destination-time. ### Interpretation and elasticity magnitudes - PPML estimates (columns (1)–(3)) show export volumes fall with distance and tariffs. - Distance elasticity falls with quality; tariff elasticity increases with quality. - These heterogeneities remain when including firm-destination-time fixed effects. - OLS estimates smaller in magnitude but signs and significance similar except for remoteness and interaction distance×quality in column (6). - Point estimates at mean, 5th, and 95th percentiles of quality distribution (based on column (2)): - Distance elasticity = -1.040 (mean), -1.142 (5th percentile), -0.932 (95th percentile). - Tariff elasticity = -2.314 (mean), -1.540 (5th percentile), -3.147 (95th percentile). - All reported as significant at the one percent level. --- ### 6 Robustness ### Robustness exercises summarized (Appendix F in source) - Alternative samples and specifications (Table F1): - Including wholesalers and retailers: results continue to hold; share of wine exports handled by intermediaries = 4.80 percent in 2002 and 5.33 percent in 2009. - Control for firm-product-time fixed effects when each wine can be exported by more than one firm. - Instrumenting tariffs×quality interaction with (log(1+total wine production) × quality) to address tariff endogeneity; Kleibergen-Paap F statistic = 83 rejects weak instrument null. - Excluding year 2002 to account for Argentina’s December 2001 banking freeze and crisis. - Controlling for port of exit (wine product redefined including port) — conclusions unchanged. - Controlling for export volumes and their interaction with quality. - Computing unit values at annual frequency to address quarterly price vs annual tariff timing. - Including small shipments < 4.5 liters. - Measurement of quality (Table F2): - Using Parker ratings; rescaling quality to 1–6 Wine Spectator bins; excluding “Great” wines; excluding US from sample. - Instrumenting Wine Spectator with Parker scores (both interacted with distance and tariffs); Kleibergen-Paap F statistic = 404 rejects weak instrument null. - Estimating quality at 12-digit HS level via Khandelwal (2010) methodology — sample size more than doubles; results hold. - Interacting distance and tariffs with both estimated quality and Wine Spectator ratings — coefficients on both sets remain significant with expected signs; those on estimated quality less significant. - Classifying quality into high quality (“Very good,” “Outstanding,” “Great”) and low quality (“Not recommended,” “Mediocre,” “Good”) and estimating separate coefficients. - Cross-sectional variation (Table F3): - Interacting distance and tariff variables with year dummies (equation (14)): For 2004–2009, markups rise with distance, fall with tariffs, effects stronger for lower quality exports; insignificant results for 2002–2003 possibly due to crisis or poor data coverage. --- ### 7 Concluding Remarks ### Core conclusions - Empirical evidence shows exporters adjust markups across destinations depending on per unit trade costs (distance) and ad valorem trade costs (tariffs). - Firms raise markups in more distant markets. - Firms lower markups in high-tariff countries. - Heterogeneity: - Effects of trade costs on markups are heterogeneous and smaller in magnitude for higher quality exports. - Heterogeneity stronger for exports to richer countries and driven predominantly by higher quality firms, larger firms, and firms with larger export market shares. - High performance firms tend to charge higher markups and better adjust markups across countries and quality levels in response to changes in trade costs. - Implications: - Variation in firm-level export prices across markets is driven not only by quality differences but also by markup variation conditional on quality. - Firms price discriminate across destinations; extent depends on size and nature of trade costs (per unit versus ad valorem) and the quality exported. - Trade costs are key to explaining deviations from the Law of One Price and international market segmentation. - Aggregate export prices and markups are likely affected because high performance firms contribute the bulk of aggregate exports. - Trade models assuming markups invariant to country-level characteristics miss a key channel; findings support models with variable markups dependent on trade costs and emphasize modeling per unit trade costs. ### Suggested next steps (from source) - Understanding the welfare implications of these results is identified as an important next step. *Source: 5.1–7 excerpts from the IMF working paper content unit "wpiea2020036-print-pdf - 5.1 Income Heterogeneity across Destinations."* ### References ### wpiea2020036-print-pdf - References ### Key bibliographic entries - Alchian, A.A., Allen, W.R., 1964. University Economics. Wadsworth Publishing, Belmont, California. - Amiti, M., Itskhoki, O., Konings, J., 2014. Importers, exporters, and exchange rate disconnect. American Economic Review 104 (7), 1942–1978. - Amiti, M., Itskhoki, O., Konings, J., 2019. International shocks, variable markups and domestic prices. Review of Economic Studies, forthcoming. - Anderson, K., Nelgen, S., 2011. Global Wine Markets, 1961–2009: A Statistical Compendium. University of Adelaide Press, Adelaide. - Arkolakis, C., Costinot, A., Donaldson, D., Rodríguez-Clare, A., 2019. The elusive pro-competitive effects of trade. Review of Economic Studies 86 (1), 46–80. - Atkeson, A., Burstein, A., 2008. Trade costs, pricing-to-market, and international relative prices. American Economic Review 98 (5), 1998–2031. - Atkin, D., Donaldson, D., 2015. Who’s getting globalized? The size and implications of intra-national trade costs, mimeo. - Atkin, D., Khandelwal, A.K., Osman, A., 2017. Exporting and firm performance: evidence from a randomized experiment. Quarterly Journal of Economics 132 (2), 551–615. - Baldwin, R.E., Harrigan, J., 2011. Zeros, quality, and space: trade theory and trade evidence. American Economic Journal: Microeconomics 3 (2), 60–88. - Bastos, P., Silva, J., 2010. The quality of a firm’s exports: where you export to matters. Journal of International Economics 82 (2), 99–111. - Behrens, K., Murata, Y., 2007. General equilibrium models of monopolistic competition: a new approach. Journal of Economic Theory 136 (1), 776–787. - Bellone, F., Musso, P., Nesta, L., Warzynski, F., 2014. International trade and firm-level markups when location and quality matter. Journal of Economic Geography 16 (1), 67–91. - Berman, N., Martin, P., Mayer, T., 2012. How do different exporters react to exchange rate changes? Quarterly Journal of Economics 127 (1), 437–492. - Bernini, M., Tomasi, C., 2015. Exchange rate pass-through and product heterogeneity: does quality matter on the import side? European Economic Review 77, 117–138. - Bosker, M., Buringh, E., 2019. Iceberg transport costs. Evidence from the 19th century ice trade. Economic Journal, forthcoming. - Broda, C., Weinstein, D.E., 2006. Globalization and the gains from variety. Quarterly Journal of Economics 121 (2), 541–585. - Cavallo, A., Gopinath, G., Neiman, B., Tang, J., 2019. Tariff passthrough at the border and at the store: evidence from US trade policy, mimeo. - Chen, N., Imbs, J., Scott, A., 2009. The dynamics of trade and competition. Journal of International Economics 77 (1), 50–62. - Chen, N., Juvenal, L., 2016. Quality, trade, and exchange rate pass-through. Journal of International Economics 100, 61–80. - Chen, N., Juvenal, L., 2018. Quality and the Great Trade Collapse. Journal of Development Economics 135, 59–76. - Crinò, R., Epifani, P., 2012. Productivity, quality, and export behaviour. Economic Journal 122 (565), 1206–1243. - Crozet, M., Head, K., Mayer, T., 2012. Quality sorting and trade: firm-level evidence for French wine. Review of Economic Studies 79 (2), 609–644. - Daudin, G., Héricourt, J., Patureau, L., 2018. International transport costs: new findings from modelling additive costs. CEPII Working Paper 2018–10. - De Loecker, J., Goldberg, P.K., Khandelwal, A.K., Pavcnik, N., 2016. Prices, markups, and trade reform. Econometrica 84 (2), 445–510. - De Loecker, J., Warzynski, F., 2012. Markups and firm-level export status. American Economic Review 102 (6), 2437–2471. - Eaton, J., Kortum, S., 2002. Technology, geography and trade. Econometrica 70 (5), 1741–1779. - Emlinger, C., Lamani, V., 2017. International trade, quality sorting and trade costs: the case of Cognac. CEPII Working Paper 2017–18. - Fajgelbaum, P., Grossman, G.M., Helpman, E., 2011. Income distribution, product quality, and international trade. Journal of Political Economy 119 (4), 721–765. - Fan, H., Li, Y.A., Xu, S., Yeaple, S.R., 2017. Quality, variable markups, and welfare: a quantitative general equilibrium analysis of export prices, mimeo. - Feenstra, R.C., 1994. New product varieties and the measurement of international prices. American Economic Review 84 (1), 157–177. - Feenstra, R.C., 2003. A homothetic utility function for monopolistic competition models, without constant price elasticity. Economics Letters 78 (1), 79–86. - Feenstra, R.C., Romalis, J., 2014. International prices and endogenous quality. Quarterly Journal of Economics 129 (2), 477–527. - Fitzgerald, D., Haller, S., 2014. Pricing-to-market: evidence from plant-level prices. Review of Economic Studies 81 (2), 761–786. - Gaulier, G., Zignago, S., 2010. BACI: International trade database at the product-level. The 1994–2007 version. CEPII Working Paper 2010–23. - Görg, J., Halpern, L., Muraközy, B., 2017. Why do within-firm-product export prices differ across markets? Evidence from Hungary. The World Economy 40 (6), 1233–1246. - Greenhut, M.L., Ohta, H., Sailors, J., 1985. Reverse dumping: a form of spatial price discrimination. Journal of Industrial Economics 34 (2), 167–181. - Hallak, J.C., 2006. Product quality and the direction of trade. Journal of International Economics 68 (1), 238–265. - Harrigan, J., Ma, X., Shlychkov, V., 2015. Export prices of US firms. Journal of International Economics 97 (1), 100–111. - Head, K., Mayer, T., 2014. Gravity equations: workhorse, toolkit, and cookbook, in: Gopinath, G., Helpman, E., Rogoff, K. (Eds), Handbook of International Economics, Vol. 4, Elsevier, Amsterdam, pp. 131–195. - Hoover, E.M., 1937. Spatial price discrimination. Review of Economic Studies 4 (3), 182–191. - Hummels, D., Klenow, P.J., 2005. The variety and quality of a nation’s exports. American Economic Review 95 (3), 704–723. - Hummels, D., Skiba, A., 2004. Shipping the good apples out? An empirical confirmation of the Alchian-Allen conjecture. Journal of Political Economy 112 (6), 1384–1402. - Imbs, J., Méjean, I., 2015. Elasticity optimism. American Economic Journal: Macroeconomics 7 (3), 43–83. - Irarrazabal, A., Moxnes, A., Opromolla, L.D., 2015. The tip of the iceberg: a quantitative framework for estimating trade costs. Review of Economics and Statistics 97 (4), 777–792. - Johnson, R.C., 2012. Trade and prices with heterogeneous firms. Journal of International Economics 86 (1), 43–56. - Khandelwal, A., 2010. The long and short (of) quality ladders. Review of Economic Studies 77 (4), 1450–1476. - Krugman, P., 1980. Scale economies, product differentiation, and the pattern of trade. American Economic Review 70 (5), 950–959. - Kugler, M., Verhoogen, E., 2012. Prices, plant size, and product quality. Review of Economic Studies 79 (1), 307–339. - Lashkaripour, A., 2017. The international gains from declining transport costs, mimeo. - Lugovskyy, V., Skiba, A., 2015. How geography affects quality. Journal of Development Economics 115, 156–180. - Lugovskyy, V., Skiba, A., 2016. Positive and negative effects of distance on export prices. Journal of Economic Behavior and Organization 127, 155–181. - Manova, K., Zhang, Z., 2012. Export prices across firms and destinations. Quarterly Journal of Economics 127 (1), 379–436. - Martin, J., 2010. Markups, quality, and transport costs. CREST Working Paper 2010–17. - Martin, J., 2012. Markups, quality, and transport costs. European Economic Review 56 (4), 777–791. - Medina, P., 2018. Import competition, quality upgrading and exporting: evidence from the Peruvian apparel industry, mimeo. - Melitz, M.J., 2003. The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica 71 (6), 1695–1725. - Melitz, M.J., Ottaviano, G.I.P., 2008. Market size, trade, and productivity. Review of Economic Studies 75 (1), 295–316. - Ottaviano, G., Tabuchi, T., Thisse, J.-F., 2002. Agglomeration and trade revisited. International Economic Review 43 (2), 409–435. - Piveteau, P., Smagghue, G., 2019. Estimating firm product quality using trade data. Journal of International Economics 118, 217–232. - Santos Silva, J.M.C., Tenreyro, S., 2006. The log of gravity. Review of Economics and Statistics 88 (4), 641–658. - Schott, P.K., 2004. Across-product versus within-product specialization in international trade. Quarterly Journal of Economics 119 (2), 647–678. - Simonovska, I., 2015. Income differences and prices of tradables: insights from an online retailer. Review of Economic Studies 82 (4), 1612–1656. - Stock, J.H., Yogo, M., 2005. Testing for weak instruments in linear IV regression, in: Andrews, D.W.K., Stock, J.H. (Eds), Identification and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg. Cambridge University Press, Cambridge, pp. 80–108. - Takechi, K., 2015. The quality of distance: quality sorting, Alchian-Allen effect, and geography. RIETI Discussion Paper 15–E–018. - Verhoogen, E., 2008. Trade, quality upgrading and wage inequality in the Mexican manufacturing sector. Quarterly Journal of Economics 123 (2), 489–530. ### Appendix A — Per Unit versus Ad Valorem Trade Costs: empirical evidence and findings - Objective: provide evidence that both per unit and ad valorem trade costs are pervasive in the data and that per unit trade costs increase with distance, making trade costs more per unit than ad valorem as distance increases. - Estimated reduced-form regression (equation A1) for each wine product k exported by firm i to country j in quarter t: - Dependent variable: ln φ_{i j k, t} = log freight charges divided by volume exported (US dollars per liter). - Regressors include: ln uv_{i j k, t} (export unit value, US dollars per liter), ln dist_j (bilateral distance between Argentina and country j), ln q_{i j k, t} (shipment size in liters), firm-time Δ_{i,t} fixed effects, product characteristics Δ^p fixed effects (grape, type, vintage year, HS-level, packaging, and province of origin). Robust standard errors clustered by destination-time. - Endogeneity and instruments: - Unit values instrumented with mean unit values on exports to other destinations. - Shipment size instrumented with destination’s GDP and GDP per capita. - Sample used for equation (A1): - 107 firms, 1,440 wines (1,534 products), 82 export markets, between 2005Q4 and 2009Q4, for 6,802 observations (coverage of freight charges incomplete so estimation sample sizes differ). - Main empirical results (summarized from Table A1 and discussion): - Column (1) OLS: ln unit value coefficient = 0758 (0032) ¤¤¤. - Column (2) IV (unit values instrumented): ln unit value coefficient = 0713 (0043) ¤¤¤. - Column (3) IV (unit values and shipment size instrumented): ln unit value coefficient = 0667 (0048) ¤¤¤. - ln distance coefficients: - Column (1): 0062 (0031) ¤¤ - Column (2): 0069 (0030) ¤¤ - Column (3): 0068 (0030) ¤¤ - ln export volume coefficients (negative, scale economies): - Column (1): ¡0121 (0012) ¤¤¤ - Column (2): ¡0133 (0014) ¤¤¤ - Column (3): ¡0187 (0022) ¤¤¤ - Interaction ln unit value × ln distance introduced in Columns (4)–(6): - Interaction coefficients (Column (4)–(6) reported as): ln unit value × ln distance = ¡0059 (0034) ¤ ; ¡0107 (0035) ¤¤¤ ; ¡0096 (0035) ¤¤¤ respectively. - Unit value elasticities (mean and percentiles from columns with interaction): - Mean: 0756 (0033) ¤¤¤ ; 0702 (0043) ¤¤¤ ; 0656 (0048) ¤¤¤ - 5th percentile: 0810 (0039) ¤¤¤ ; 0800 (0045) ¤¤¤ ; 0744 (0049) ¤¤¤ - 95th percentile: 0702 (0051) ¤¤¤ ; 0604 (0061) ¤¤¤ ; 0569 (0065) ¤¤¤ - R-squared values across columns: 0.660, 0.299, 0.287, 0.660, 0.301, 0.289. - Observations reported: 6,706; 5,116; 5,116; 6,706; 5,116; 5,116 (as in Table A1 layout). - Kleibergen-Paap F statistics reported: 346.35; 196.90; –; 143.25; 147.73 (as in Table A1 layout). - Interpretation of empirical findings: - Freight costs rise with distance and fall with export volumes. - Freight costs increase with unit values but with elasticity strictly less than unity (e.g., 0667–0758), indicating a mix of ad valorem and per unit components. - Instrumenting reduces the estimated unit value elasticity (OLS upward biased). - The negative and significant interaction between unit value and distance implies freight costs become more per unit (relative to ad valorem) as distance increases. - Reported example elasticities at mean and percentiles of distance (from Column (6)): - At mean distance: unit value elasticity = 0656 (significant and < 1). - At 5th percentile of distance: elasticity = 0744. - At 95th percentile of distance: elasticity = 0569. - Based on Kleibergen-Paap F statistics, instrument relevance is acceptable in reported IV specifications (rejecting weak instruments in columns (5) and (6)). ### Appendix B — Alternative demand systems: theoretical implications for per unit and ad valorem trade costs - Framework: - General demand system for differentiated goods following Arkolakis, Costinot, Donaldson, and Rodríguez-Clare (2019). - Marshallian demand specification used (equation B1): ln θ(π_cif, π*, y) = ¡β ln π_cif + γ ln y + φ(ln π_cif ¡ ln π*), where π_cif = τ π_fob + T, with τ (ad valorem) and T (per unit) trade cost components. - Four demand systems encompassed by the framework: - CES utility: γ = β = 1 and φ(σ) = (1 ¡ ρ) σ. - Quadratic, non-separable utility (Ottaviano et al., 2002): γ = 0, β = ¡1, φ(σ) = ¡ln μ^2 + ln(ε ¡ σ ¡1). - Translog preferences (Feenstra, 2003): γ = β = 1, φ(σ) = ln α + ln(¡σ). - Additively quasi-separable utility (Behrens and Murata, 2007): γ = β = 0, φ(σ) = ln ξ + ln(¡σ). - Per unit trade costs (T > 0 vs T = 0): - For all four demand systems, the elasticity of demand with respect to the FOB price, ε_{π_fob}, is negative (rows (1) and (2) of Table B1). - The elasticity of ε_{π_fob} with respect to per unit trade costs (∂ ε_{π_fob} / ∂ T) is: - Negative for CES, translog, and additively quasi-separable utility (for some σ values). - Positive for quadratic preferences. - The elasticity ∂ ε_{π_fob} / ∂ T is increasing with the FOB price for all demand systems (row (4) of Table B1). - Implication noted from Irarrazabal et al. (2015): higher per unit trade costs reduce the magnitude of the elasticity of demand to the FOB price, especially among low-price firms; results vary across demand systems. - Ad valorem trade costs (τ > 1 vs τ = 1, with T > 0): - For both τ = 1 and τ > 1 (with T > 0), ε_{π_fob} remains negative across all demand systems (rows (5) and (6) of Table B1). - The hypothesis that export prices fall with ad valorem trade costs—especially for low-price/low-quality exports—requires ∂ ε_{π_fob} / ∂ τ to be positive and decreasing with the FOB price. - Analytical results for ∂ ε_{π_fob} / ∂ τ by demand system: - CES preferences: - ∂ ε_{π_fob} / ∂ τ × (τ / ε_{π_fob}) = T / (τ π_fob + T) > 0. (equation B2) - This elasticity decreases with π_fob: derivative is ¡T τ / (τ π_fob + T)^2 < 0. (equation B3) - Quadratic, non-separable utility: - ∂ ε_{π_fob} / ∂ τ × (τ / ε_{π_fob}) = (T ¡ π*) / (π_cif ¡ π*) > 0. (equation B4) - This elasticity increases with π_fob: derivative = ¡(T ¡ π*) τ / (τ π_fob ¡ π* + T)^2 > 0. (equation B5) - Translog preferences: - ∂ ε_{π_fob} / ∂ τ × (τ / ε_{π_fob}) = σ^2 (T ¡ σ T + τ π_fob ¡ σ π_cif) / (1 ¡ σ) > 0. (equation B6) - The elasticity decreases with π_fob (argument via chain rule in equation B7 and convexity properties). - Additively quasi-separable utility: - ∂ ε_{π_fob} / ∂ τ × (τ / ε_{π_fob}) = τ π_fob ¡ σ T ¡ σ π_cif > 0. (equation B8) - The change of this elasticity with π_fob is given by equation B9; sign depends on σ relative to roots of a quadratic in parameters and can be positive or negative depending on σ and parameter values. - Summary implications (end of presented material): - Across demand systems, ∂ ε_{π_fob} / ∂ τ is generally positive (meaning ad valorem trade costs tend to reduce the demand elasticity to FOB price), but the way this elasticity changes with FOB price depends on the specific demand system: - CES and translog: elasticity with respect to τ decreases with FOB price. - Quadratic: elasticity increases with FOB price. - Additively quasi-separable: sign of change with FOB price depends on parameter ranges (σ and others). - These theoretical results inform expectations about how export prices and pass-through respond to per unit versus ad valorem trade costs and how effects vary across firms with different FOB prices. *Italic: Content unit: wpiea2020036-print-pdf - References* ### 1. The prediction that prices and markups increase with distance requires ### 1. The prediction that prices and markups increase with distance requires ### Key theoretical predictions and conditions - Prediction that prices and markups increase with distance requires ¡    ¢ ¡   ¢ 0. - "This condition is satis...ed with CES. It also holds with translog and additively quasi-separable utility (for some values of). It is not satis...ed with quadratic preferences." - Prediction that the effect of distance on prices and markups is smaller in magnitude for higher price (and higher quality) exports requires ¡   ¢ ¡¡    ¢ ¡   ¢¢ 0. - "This condition holds in all cases." - Prediction that prices and markups fall with tari¤s requires ¡    ¢ ¡   ¢ 0. - "This condition holds for all demand systems." - Prediction that the effect of tari¤s on prices and markups is smaller in magnitude for higher price (and higher quality) exports requires ¡   ¢ ¡¡    ¢ ¡   ¢¢ 0. - "This condition is satis...ed with CES, translog, and additively quasi-separable utility (for some values of). It is not satis...ed with quadratic preferences." - Summary conclusion: "Our predictions that (1) export prices and markups increase with distance, (2) fall with tari¤s, (3) and to a larger extent for lower quality exports hold with CES, and with translog and additively quasi-separable utility (for some parameter values only). Models with quadratic, non-separable utility can yield some, but not all, of our predictions." ### Table B1: Alternative Demand Systems (high-level signs/conditions preserved) - Per unit trade cost T: - (1)   = ln ln  j =0 : signs reported for CES, Quadratic, Translog, Quasi-separable in table. - (2)   = ln ln  j >0 : signs reported (all 0). - (3)    (/ ) : mixed signs depending on  (see table). - (4)    (   (/ ) ) 0 for all demand systems in table. - Ad valorem trade cost : - (5)   = ln ln  j =1 : signs reported (all 0). - (6)   = ln ln  j >1 : signs reported (all 0). - (7)    (/ ) 0 for all demand systems in table. - (8)    (   (/ ) ) : signs vary by demand system and parameter ranges (table). --- ### C Elasticity of Demand to the FOB Price ### Model and empirical specification - Estimate (Irarrazabal et al., 2015) specification: - ln_{i j k,t} = ª [ln_{i j k,t} £ ln_{j} £ _{k}] + ̈ [ln_{i j k,t} £ ln_{j,t} £ _{k}] + _{i j,t} + _{i j k,t} (equation C1) - Demand elasticity   = ln_{i j k,t} / ln_{i j k,t} given by: -   = ª1 + ª2 ln_{j} + ª3 _{k} + ª4 ln_{j} £ _{k} + ̈2 ln_{j,t} + ̈4 ln_{j,t} £ _{k} (equation C2) - Expectations on signs: - ª2 0 (distance increases the negative   i.e.,   approaches zero). - ª4 0 (effect of distance on   smaller for higher quality). - ̈2 0 (tari¤s decrease the negative   i.e.,   becomes more negative). - ̈4 0 (effect of tari¤s on   smaller for higher quality). ### Endogeneity and instruments - Instrument unit values with mean unit value of each wine product on exports to other destinations (and instrument interactions analogously). - Firm-destination-time fixed effects _{i j,t} included; robust standard errors clustered at destination-time. ### Empirical estimates (Table C1) - Column (1): OLS; Column (2): IV (unit values and interactions instrumented). - Selected coefficient estimates (standard errors in parentheses): - lnunit value: (1) ¡15209 (4328) ¤¤¤ ; (2) ¡11024 (4105) ¤¤¤ - quality: (1) ¡0124 (0061) ¤¤ ; (2) ¡0065 (0073) - lnunit value£quality: (1) 0153 (0046) ¤¤¤ ; (2) 0106 (0046) ¤¤ - lnunit value£lndistance (ª2): (1) 2115 (0498) ¤¤¤ ; (2) 1457 (0494) ¤¤¤ - lnunit value£lntari¤s ( ̈2): (1) ¡9186 (3637) ¤¤ ; (2) ¡6173 (3683) ¤ - lnunit value£lndistance£quality (ª4): (1) ¡0023 (0005) ¤¤¤ ; (2) ¡0016 (0006) ¤¤¤ - lnunit value£lntari¤s£quality ( ̈4): (1) 0106 (0040) ¤¤¤ ; (2) 0068 (0040) ¤ - Goodness of fit and samples: - R-squared: (1) 0.566 ; (2) 0.076 - Observations: (1) 87,078 ; (2) 58,646 - Estimation: (1) OLS ; (2) IV - Kleibergen-Paap F–178.76 reported for IV. ### Elasticities calculated from estimates - Elasticity expressions: -   / dist_{j} * dist_{j} /   = 1/  (ª2 + ª4 £ quality_{k}) (equation C3) -   / tar_{j,t} * tar_{j,t} /   = 1/  ( ̈2 + ̈4 £ quality_{k}) (equation C4) - Using OLS (column (1)), elasticities at mean, 5th, and 95th percentiles of quality: - Elasticity of   w.r.t. distance = ¡0454, ¡1416, and ¡0098 (mean, 5th, 95th percentiles). - Elasticity of   w.r.t. tari¤s = 0346 at mean; falls from 3349 at 5th percentile to ¡0761 at 95th percentile (turns negative). - Using IV (column (2)), elasticities at mean, 5th, and 95th percentiles: - Elasticity of   w.r.t. distance = ¡0137, ¡0280, and ¡0033 (mean, 5th, 95th percentiles). - Elasticity of   w.r.t. tari¤s = 0445, 1037, and 0012 (mean, 5th, 95th percentiles). - Conclusion: "The elasticity of   with respect to distance is therefore negative and increases with quality. Instead, the elasticity of   with respect to tari¤s is positive and decreases with quality." --- ### D Selection Bias across Firms ### Three-step estimator implementation (Harrigan et al., 2015) - Construct balanced sample of all firm-wine-destination-time combinations including zero trade flows; drop years prior to vintage year for each wine. - Step 1 (probit for entry): - pr(_{i j k,t} > 0) = © (1 lnY_{j,t} + _{k,t}) (equation D1) - Y_{j,t} includes distance, tari¤s, GDP, GDP per capita, and remoteness (distance and tari¤s can be interacted with quality); _{k,t} product-time fixed effects. - Obtain estimated inverse Mills ratio b_{i j k,t}. - Step 2 (OLS for positive export values): - ln_{i j k,t} = 1 lnY_{j,t} + 2 b_{i j k,t} + _{k,t} + _{i j k,t} (equation D2) - Compute quasi-residuals b_{i j k,t} = b2 b_{i j k,t} + b_{i j k,t} = ln_{i j k,t} ¡ b1 lnY_{j,t} ¡ b_{k,t}. - Step 3 (unit value regression with selection control): - ln_{i j k,t} = 1 lnY_{j,t} + 2 b_{i j k,t} + _{k,t} + _{i j k,t} (equation D3) - Equations (D1) and (D2) estimated separately for each wine; equation (D3) pooled across wines. - Standard errors: third-stage standard errors clustered by destination-time. ### Results (Table D1) — selected estimates and interpretation - Dependent variable: log FOB unit value (US$ per liter). Product-time fixed effects included. - Selected coefficients (standard errors in parentheses): - lndistance: 0033 (0006) ¤¤¤ in column (1) - lndistance£quality: –0006 (0001) ¤¤¤ in column (2) - lntari¤s: ¡0135 (0035) ¤¤¤ in column (1); ¡0130 (0031) ¤¤¤ in column (2) - lntari¤s£quality: 0016 (0005) ¤¤¤ in column (2) - lnremoteness: 0046 (0011) ¤¤¤ in column (1) - lnGDP: ¡0017 (0002) ¤¤¤ in column (1) - lnGDP/cap: 0019 (0006) ¤¤¤ in column (1) - selection control: 0044 (0007) ¤¤¤ across columns - Observations: column (1) 66,785 ; column (2) 66,578 ; column (3) 66,785 - R-squared: 0.785, 0.784, 0.785 respectively. - Interpretation: - Positive coefficient on selection control implies correlation between errors of export price and export value regressions around four percent. - Controlling for selection yields results economically and statistically similar to benchmark findings. - When measuring distance continuously or with interval dummies, markups increase with bilateral distance, remoteness, and destination income per capita, and decrease with tari¤s and country size. - "In column (1), a doubling of distance induces an exporter to increase its markup by 2.3 percent on average." - Column (3) results at mean, 5th, and 95th percentiles of quality: - Distance elasticity = 0.035, 0.071, and 0.001 (95th percentile insigni...cant). - Tari¤ elasticity = ¡0140, ¡0238, and ¡0047 (95th percentile not signi...cant). - Conclusion: "Our results are therefore robust to controlling for selection bias across firms." --- ### E Estimation of Quality ### Quality definition and estimation approach - Follow Bernini and Tomasi (2015) adapting Khandelwal (2010): quality is the part of market share not explained by price. - Estimation equation (E1): - lnS_{i j k,t} ¡ lnS_{j K,t} = 1 u v_{i j k,t} + 2 ln nS_{i j k,t} + _{j,t} + _{i k} + _{i j k,t} - S_{i j k,t}: market share of product k exported by firm i to country j in period t. - S_{j K,t}: market share of "outside variety" K. - nS_{i j k,t}: "nest share." - u v_{i j k,t}: export unit value. - _{j,t} and _{i k}: destination-time and firm-product fixed effects. - Construction of variables: - Use Nosis quarterly data 2002Q1–2009Q4 for export value _{i j k,t} (US$) and quantity _{i j k,t} (kg) to compute unit values u v_{i j k,t} (US$ per kg). - Use annual BACI data (2002–2009) to compute outside variety share S_{j K,t} as share of non-Argentinean import quantities in total import quantities of country j in a 6-digit HS product category K. - Match annual S_{j K,t} with quarterly Nosis by year. - Market share and nest share formulas: - S_{i j k,t} = q_{i j k,t} / (Σ_i q_{i j K,t} / (1 ¡ S_{j K,t})) (equation E2) - nS_{i j k,t} = q_{i j k,t} / (Σ_i q_{i j k,t} / (1 ¡ S_{j K,t})) (equation E3) - Instruments to address endogeneity: - Instrument unit values with mean unit value of each 8-digit HS product by destination-time. - Instrument nest shares with number of 8-digit HS products by firm-destination-time. - Estimation by 2-digit HS category separately. - Quality measure (E4): - quality_{i j k,t} = b_{j,t} + b_{i k} + b_{i j k,t} = [lnS_{i j k,t} ¡ lnS_{j K,t}] ¡ [b1 u v_{i j k,t} + b2 ln nS_{i j k,t}] - Output: quality estimated for each 8-digit HS product by firm-destination-quarter between 2002Q1 and 2009Q4; same approach used for wine exports in Section 6. --- ### F Robustness ### Robustness on samples and specifications (Table F1) — selected coefficients - Dependent variable: log FOB unit value (US$ per liter). - Key interaction: lndistance£quality coefficients: - Column (1): ¡0004 (0001) ¤¤¤ - Column (2): ¡0004 (0001) ¤¤¤ - Column (3): ¡0003 (0001) ¤¤¤ - Column (4): ¡0003 (0001) ¤¤¤ - Column (5): ¡0003 (0001) ¤¤¤ - Column (6): ¡0005 (0001) ¤¤¤ - Column (7): ¡0004 (0001) ¤¤¤ - lntari¤s£quality coefficients: - Column (1): 0028 (0004) ¤¤¤ - Column (2) (IV): 0033 (0013) ¤¤ - Column (3): 0027 (0004) ¤¤¤ - Column (4): 0025 (0004) ¤¤¤ - Column (5): 0024 (0004) ¤¤¤ - Column (6): 0023 (0006) ¤¤¤ - Column (7): 0026 (0004) ¤¤¤ - Observations vary by sample: 67,105; 56,605; 64,884; 63,743; 66,941; 53,197; 71,466. - Samples include: Retailers, Full, >2002, Ports, FullAnnual, Less 4.5l. - Estimation methods: OLS and IV reported. ### Robustness on quality measures (Table F2) — selected coefficients - lndistance£quality: - (1) ¡0007 (0002) ¤¤¤ - (2) ¡0019 (0004) ¤¤¤ - (3) ¡0004 (0001) ¤¤¤ - (4) ¡0003 (0001) ¤¤¤ - (5) ¡0009 (0003) ¤¤¤ - (6) –¡0003 (0001) ¤¤¤ - lntari¤s£quality: - (1) 0046 (0011) ¤¤¤ - (2) 0113 (0018) ¤¤¤ - (3) 0027 (0004) ¤¤¤ - (4) 0020 (0004) ¤¤¤ - (5) 0061 (0015) ¤¤¤ - (6) –0028 (0004) ¤¤¤ - Quality measures/samples: - WS refers to Wine Spectator ratings; Parker ratings used in some IVs. - quality^ε estimated via Khandelwal (2010) in some columns. - Columns cover Full, No “Great”, No US, Est./WS, WS variants. - Sub-sample and instrument robustness indicate consistency of sign and statistical significance across alternative quality measures. ### Cross-sectional estimates (Table F3) - Yearly cross-sectional coefficients (selected): - Year 2002: lndistance 0159 (0269) ; lndistance£quality ¡0001 (0003) ; lntari¤s 1815 (1132) ; lntari¤s£quality ¡0021 (0014) - Year 2004: lndistance 0951 (0195) ¤¤¤ ; lndistance£quality ¡0011 (0002) ¤¤¤ ; lntari¤s ¡3812 (0985) ¤¤¤ ; lntari¤s£quality 0045 (0012) ¤¤¤ - Year 2009: lndistance 0438 (0167) ¤¤¤ ; lndistance£quality ¡0005 (0002) ¤¤ ; lntari¤s ¡2378 (0943) ¤¤ ; lntari¤s£quality 0027 (0011) ¤¤ - Observations: 71,952. Product-time fixed effects included; GDP, GDP per capita, remoteness included but not reported. --- *Italic: Content derived from "wpiea2020036-print-pdf - 1. The prediction that prices and markups increase with distance requires" (IMF PDF chapter/section).* --- _Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020036-print-pdf.pdf_