Why trade matters after all

Transcript

1 277 – Journal of International Economics 97 (2015) 266 ScienceDirect Contents lists available at Journal of International Economics journal homepage: www.elsevier.com/locate/jie ☆ Why trade matters after all Ralph Ossa University of Chicago, Booth School of Business, 5807 South Woodlawn Avenue, Chicago, IL 60637, United States NBER, United States article info abstract Article history: es the estimated gains from I show that accounting for cross-industry variation in trade elasticities greatly magni fi Received 17 May 2013 trade. The main idea is as simple as it is general: while imports in the average industry do not matter too much, Received in revised form 22 June 2015 imports in some industries are critical to the functioning of the economy, so that a complete shutdown of inter- Accepted 13 July 2015 national trade is very costly overall. Available online 22 July 2015 © 2015 Elsevier B.V. All rights reserved. JEL classi fi cation: F10 Keywords: Gains from trade Multi-industry model Trade elasticity 1 1. Introduction multi-industry environment, where the aggregate is now a weighted ~ ε j 1 average of the industry-level . The point is that if is close to zero in ε s ε s Either the gains from trade are small for most countries or the work- 1 some industries, fi is close to in - fi nity in these industries which is suf ε horse models of trade fail to adequately capture those gains. This s 1 cient to push ε up a lot. Loosely speaking, ε is a weighted average of uncomfortable conclusion seems inevitable given recent results in quan- s ~ ε j Arkolakis et al. (2012) ,thegainsfrom titative trade theory. As shown by so that the exponent of the aggregate formula is the inverse of the average trade can be calculated in the most commonly used quantitative trade of the trade elasticities whereas the exponent of the industry-level for- λ models from the observed share of a country's trade with itself, ,and j average of the inverse of the trade elasticities. mula is the , ε the elasticity of aggregate trade fl ows with respect to trade costs, model Armington (1969) I make this point in the context of a simple 1 − 1 ε in which consumers have CES preferences within industries and goods Þ . λ ¼ð Using standard methods to obtain G using the formula j j are differentiated by country of origin. As is well-known, the trade elas- and ε , I show below that this implies that a move from λ estimates of j ticities then depend on the elasticities of substitution through the simple complete autarky to 2007 levels of trade would increase real income relationship ε 1. Estimating these elasticities at the 3-digit level = σ − by only 16.5% on average among the 50 largest economies in the world. s s ned by and re Feenstra (1994) using the standard method developed by fi In this paper, I argue that the workhorse models of trade actually Broda and Weinstein (2006) , I show that the industry-level formula pre- predict much larger gains once the industry dimension of trade fl ows dicts that a move from autarky to 2007 levels of trade increases real in- is taken into account. The main idea is as simple as it is general: while come by 48.6% on average which is around three times the number the imports in the average industry do not matter too much, imports in aggregate formula predicts. It increases even further once I allow for some industries are critical to the functioning of the economy, so that non-traded goods and intermediate goods which have opposing effects a complete shutdown of international trade is very costly overall. In par- 1 nd that the gains from on the gains from trade. All things considered, I fi − ~ ε j ticular, I show that the above formula can be written as G ¼ð Þ in a λ 2 j j trade average 55.9% among the 50 largest economies in the year 2007. 2 ☆ I would like to thank two referees as well as Ariel Burstein, Kerem Cosar, Chang-Tai While my general point also extends to imperfectly competitive gravity models such Hsieh, and Anson Soderbery for useful comments and discussions. This work is Melitz (2003) , the particular gains from trade predicted by my and Krugman (1980) as supported by the Business and Public Policy Faculty Research Fund at the University of Armington (1969) multi-sector model are only exactly the same in other perfectly com- Chicago Booth School of Business. The usual disclaimer applies. petitive gravity models such as . This is because the exact iso- Eaton and Kortum (2002) . [email protected] E-mail address: trade models does not apply in the case of multiple ” new “ ” morphism between “ old and 1 Costinot .However,recentcalculationsby Arkolakis et al. (2012) industries as shown by model, the Eaton and Armington (1969) Krugman (1980) model, the This includes the and Rodriguez-Clare (2014) suggest that even with multiple industries the gains from model, and the Kortum (2002) ε cor- model. The aggregate trade elasticity Melitz (2003) ” “ new ” trade models. and trade are quite similar in old “ responds to different structural parameters in different models. http://dx.doi.org/10.1016/j.jinteco.2015.07.002 0022-1996/© 2015 Elsevier B.V. All rights reserved.

2 277 – R. Ossa / Journal of International Economics 97 (2015) 266 267 σ s ! While my point may seem obvious once stated, I believe it has not 1 − σ N s 1 − σ X s σ s fl been made explicitly before. brie ydiscussa Arkolakis et al. (2012) C ð C Þ ¼ 3 js ijs ¼ 1 i multi-industry formula in an extension but never contrast it to their ag- gregate formula or use it to actually calculate the gains from trade. Caliendo and Parro (2015) Ossa (2014) , and , Hsieh and Ossa (2012) , others work with multi-industry versions of standard trade models Notice that I allow the Cobb Douglas shares of the aggregate inter- – but also do not point out that cross-industry heterogeneity in the mediate good to vary by country , upstream industry j s , and down- trade elasticities has the potential to greatly magnify the gains from stream industry , which allows me to match input – output tables from t Edmond et al. trade. Closest in spirit is perhaps the contribution by around the world. The aggregate nal good translates one-for-one into fi (2012) which measures the gains from trade originating from pro- utility U L . The aggregate intermediate good is combined with labor is j nding is fi competitive effects in an oligopolistic trade model. A key Douglas technology to produce the country-industry- – using a Cobb that such pro-competitive effects are large if there is a lot of cross- speci fi c traded varieties Q . In combi- with total factor productivities A is is industry variation in markups which is the case if there is a lot of nation, these assumptions imply: 3 cross-industry variation in the elasticities of substitution. perform Having said this, Costinot and Rodriguez-Clare (2014) F 4 C ¼ Þ ð U j j closely related calculations in recently published contemporaneous work. In particular, they also work outthe gainsfrom trade usingthe ag- ! β 1 −  is s I ; β gregate and industry-level formulas consideringcases with and without is C L is i A ¼ Þ 5 ð Q is is intermediate goods. While my analysis features more industries (252 β β 1 − is is instead of 31), features more countries (50 instead of 34), and uses dif- ferentdata (GTAPinsteadof WIOD),the main distinction liesin theelas- There is perfect competition and the shipment of an industry trad- s ticity estimates. Instead of relying on elasticity estimates from the i involves iceberg trade barriers to country ed variety from country j literature, I estimate them using the – Feenstra (1994) Broda and for one unit to ar- i units must leave country N τ 1 in the sense that τ fi Weinstein (2006) approach. This allows me to estimate con dence ijs ijs 4 N Q rive in country so that j C = ∑ . τ The model can be solved by dence intervals for fi intervals for the elasticities and, in turn, also con ijs ijs =1 j is invoking the standard requirements that consumers maximize utility, the gains from trade. Overall, the gains from trade appear to be quite ts, and all markets clear. rms maximize pro fi ts, fi rms make zero pro fi fi precisely estimated with the average 95% con fi dence interval ranging fi ne Since the model's solution should be intuitive to most readers, I con from 49.3% until 62.5%. myself to sketching some core aspects here. The remainder ofthis paper is divided into four sections. In , Section 2 to country s fl The value of industry i trade j , X owing from country , fi model of trade in I develop a multi-industry Armington (1969) nal and ijs − 1 σ σ − 1 s s intermediate goods and show what it implies for the measurement of X follows the gravity equation ¼ is the price ,where E p P P ijs js ijs js ijs , I describe the data and discuss all ap- Section 3 the gains from trade. In is the ideal price P s of the industry , variety from country i in country j js Section 4 plied aggregation, interpolation, and matching procedures. In , s varieties available in country j ,and E index of all industry is total ex- js I discuss the elasticity estimation and give an overview of the obtained penditure on all industry s varieties in country j originating from fi nal Section 5 results. In , I report the gains from trade for 50 countries in − 1 β is I β s ; 1 − is p and intermediate demand. Moreover, w ð τ ð Þ A ¼ Þ P , ijs i ijs is i the world and document that a small share of industries typically β − 1 is ; s β I is accounts for a large share of the gains from trade. P Þ and Þ ð w is a cost term aggregating over the wage w ð where i i i the price index of the aggregate intermediate good demanded by indus- s γ S s ; I it ∏ . Combining these elements, the above gravity equa- P ¼ s P try , 1 ¼ t i it 2. Model tion becomes s There are or j and S industries indexed by i Ν countries indexed by  − 1 σ s S F s − 1 γ ðÞ β β 1 − is or t . In each country, consumers demand an aggregate fi C nal good 1 − and σ s it is j ð 6 Þ A ¼ P τ w ∏ P E X ijs ijs js is js i it t I , producers demand an aggregate intermediate good industry t C .These ¼ 1 t j – aggregate goods are Cobb Douglas combinations of industry-speci fi c I t F s , goods C ∑ , C , which are in turn CES aggregates of C = + C =1 js js js t js j / of country X ≡ s as the own trade share in industry , E fi De λ ning jjs js js s 1 fi C industry-speci c traded varieties differentiated by the location of Þ − β ð γ 1 ijs β js S − 1 jt js σ − 1 s ∏ A ¼ P w λ , which is a the above equation implies P js ¼ 1 t js jt j js their production. To be clear, C s denotes the quantity of the industry ijs . As is easy to verify, its solu- system of equations that is log-linear in P js j available in country and it is at that traded variety from country i s δ 1 jt level of disaggregation that trade physically takes place. In sum, S σ − 1 − 1 s t ð ∏ Þ ,where δ w A is element ( λ , t )ofmatrix ¼ s P tion is jt js j jt 1 ¼ t jt − 1 ) denoting the ma- I denoting the identity matrix and B with ( B − I j j ! s α js F , t )is trix whose element ( γ s ). Readers familiar with input (1 – − β S jt js C js F − 1 C ¼ ð 1 Þ ∏ B I output analysis will recognize ( − as the transpose of theLeontief ) j j α js 1 ¼ s s inverse which implies that δ is a measure of the importance of industry t jt s in the production process of industry s . In particular, a total of $ δ worth jt of industry s goods is required to meet $1 worth of industry fi nal de- t t ! γ I ; t js S C mand. This value combines industry goods used as inputs in industry t js t ; I ð 2 ∏ ¼ Þ C t j directly as well as industry t goods used as inputs in other industries s γ s ¼ 1 js 5 which then also produce inputs for industry s . 3 4 C = 1 throughout. Even though I refer to Related points have, of course, also been made in other areas of macroeconomics. For As usual, I set as traded varieties, the τ iis ijs example, Nakamura and Steinsson (2010) show how cross-industry heterogeneity in model can also accommodate non-traded ones by letting the corresponding τ . ∞ → ijs 5 Jones menu costs substantially increases the degree of monetary non-neutrality. Also, I thank a referee for suggesting this way of modeling input – output linkages which is argues that cross-industry complementarities through intermediate goods matter (2011) Costinot and more general than what I had originally done. It is based on section 3.4 of a great deal for understanding cross-country differences in incomes. and explained in more detail in their online appendix. Rodriguez-Clare (2014)

3 277 – R. Ossa / Journal of International Economics 97 (2015) 266 268 Table 1 Elasticity estimates. 95% CI SITC description SITC code Sigma 2.41 1.50 656 1.54 TULLES, LACE, EMBROIDERY, RIBBONS, TRIMMINGS AND OTHER SMALL WARES 1.37 2.00 NATURAL ABRASIVES, N.E.S. (INCLUDING INDUSTRIAL DIAMONDS) 277 1.56 2.63 1.50 1.57 WOOD, SIMPLY WORKED AND RAILWAY SLEEPERS OF WOOD 248 STONE, SAND AND GRAVEL 1.66 2.48 273 1.61 2.83 1.55 1.70 CRUDE ANIMAL MATERIALS, N.E.S. 291 ROAD MOTOR VEHICLES, N.E.S. 1.71 4.36 1.48 783 2.16 1.62 1.72 MINERAL MANUFACTURES, N.E.S. 663 1.73 1.55 2.27 657 SPECIAL YARNS, SPECIAL TEXTILE FABRICS AND RELATED PRODUCTS 2.15 1.61 MISCELLANEOUS CHEMICAL PRODUCTS, N.E.S. 598 1.75 1.76 1.66 2.53 DYEING AND TANNING EXTRACTS, AND SYNTHETIC TANNING MATERIALS 532 PARTS AND ACCESSORIES FOR TRACTORS, MOTOR CARS AND OTHER MOTOR VEHICLES, TR 1.80 1.64 2.01 784 VENEERS, PLYWOOD, PARTICLE BOARD, AND OTHER WOOD, WORKED, N.E.S. 634 1.81 1.44 3.58 1.76 2.50 MISCELLANEOUS NONFERROUS BASE METALS EMPLOYED IN METALLURGY AND CERMETS 689 1.84 723 CIVIL ENGINEERING AND CONTRACTORS' PLANT AND EQUIPMENT 1.87 1.43 2.73 NATURAL RUBBER, BALATA, GUTTA-PERCHA, GUAYULE, CHICLE AND SIMILAR NATURAL G 1.77 2.57 231 1.90 PAPER AND PAPERBOARD 1.90 1.76 2.46 641 4.55 1.90 WOVEN FABRICS OF TEXTILE MATERIALS, OTHER THAN COTTON OR MANMADE FIBERS AND 654 1.58 CLAY CONSTRUCTION MATERIALS AND REFRACTORY CONSTRUCTION MATERIALS 1.90 1.44 3.02 662 523 METALLIC SALTS AND PEROXYSALTS OF INORGANIC ACIDS 1.91 1.77 2.24 664 GLASS 1.91 1.70 2.46 325 1.92 1.82 3.05 COKE AND SEMICOKE (INCLUDING CHAR) OF COAL, OF LIGNITE OR OF PEAT, AGGLOMER 1.69 1.92 NAILS, SCREWS, NUTS, BOLTS, RIVETS AND SIMILAR ARTICLES, OF IRON, STEEL, CO 694 2.50 MANUFACTURES OF BASE METAL, N.E.S. 1.81 2.62 699 1.93 533 PIGMENTS, PAINTS, VARNISHES AND RELATED MATERIALS 1.94 1.77 2.25 SYNTHETIC RUBBER; RECLAIMED RUBBER; WASTE, PAIRINGS AND SCRAP OF UNHARDENED 1.81 2.56 232 1.95 ELECTRICAL APPARATUS FOR SWITCHING OR PROTECTING ELECTRICAL CIRCUITS OR FOR 1.96 1.78 2.73 772 2.51 1.98 SULFUR AND UNROASTED IRON PYRITES 274 1.79 WIRE PRODUCTS (EXCLUDING INSULATED ELECTRICAL WIRING) AND FENCING GRILLS 1.98 1.73 2.70 693 211 HIDES AND SKINS (EXCEPT FURSKINS), RAW 1.99 1.84 2.98 1.99 281 IRON ORE AND CONCENTRATES 1.88 3.07 678 1.99 1.75 3.13 IRON AND STEEL WIRE 2.81 263 COTTON TEXTILE FIBERS 2.01 1.91 592 2.01 1.95 2.62 STARCHES, INULIN AND WHEAT GLUTEN; ALBUMINOIDAL SUBSTANCES; GLUES 882 PHOTOGRAPHIC AND CINEMATOGRAPHIC SUPPLIES 2.01 1.76 2.54 2.93 1.83 MOTORCYCLES (INCLUDING MOPEDS) AND CYCLES, MOTORIZED AND NOT MOTORIZED; INV 785 2.03 562 FERTILIZERS (EXPORTS INCLUDE GROUP 272; IMPORTS EXCLUDE GROUP 272) 2.04 1.93 2.64 2.42 695 TOOLS FOR USE IN THE HAND OR IN MACHINES 2.04 1.88 2.97 HEATING AND COOLING EQUIPMENT AND PARTS THEREOF, N.E.S. 741 2.04 1.79 3.05 1.67 2.05 FISH, FRESH (LIVE OR DEAD), CHILLED OR FROZEN 34 3.19 775 1.74 2.05 HOUSEHOLD TYPE ELECTRICAL AND NONELECTRICAL EQUIPMENT, N.E.S. 1.91 3.33 FURSKINS, RAW (INCLUDING FURSKIN HEADS, TAILS AND OTHER PIECES OR CUTTINGS, 212 2.06 2.75 1.98 2.08 ALLOY STEEL FLAT-ROLLED PRODUCTS 675 2.92 HOUSEHOLD EQUIPMENT OF BASE METAL, N.E.S. 1.82 697 2.08 1.85 2.09 ELECTRICAL MACHINERY AND APPARATUS, N.E.S. 778 2.71 BALL OR ROLLER BEARINGS 2.11 2.05 3.12 746 1.91 629 ARTICLES OF RUBBER, N.E.S. 2.12 2.86 1.82 2.71 2.13 WOOD MANUFACTURES, N.E.S. 635 2.65 2.14 CRUDE MINERALS, N.E.S. 278 1.88 VEGETABLE TEXTILE FIBERS (OTHER THAN COTTON AND JUTE), RAW OR PROCESSED BUT 265 1.87 3.55 2.16 2.75 673 IRON OR NONALLOY STEEL FLAT-ROLLED PRODUCTS, NOT CLAD, PLATED OR COATED 2.16 2.03 2.60 1.92 2.18 ORGANO-INORGANIC COMPOUNDS, HETEROCYCLIC COMPOUNDS, NUCLEIC ACIDS AND THEIR 515 1.73 4.42 2.19 NONALCOHOLIC BEVERAGES, N.E.S. 111 3.16 2.19 HYDROCARBONS, N.E.S. AND THEIR HALOGENATED, SULFONATED, NITRATED OR NITROSA 511 1.92 661 LIME, CEMENT, AND FABRICATED CONSTRUCTION MATERIALS, EXCEPT GLASS AND CLAY 2.19 1.94 2.50 2.85 1.98 2.22 IRON AND STEEL TUBES, PIPES AND HOLLOW PROFILES, FITTINGS FOR TUBES AND PIP 679 522 INORGANIC CHEMICAL ELEMENTS, OXIDES AND HALOGEN SALTS 2.23 1.84 2.60 685 LEAD 2.24 1.96 3.08 1.98 743 PUMPS (NOT FOR LIQUIDS), AIR OR GAS COMPRESSORS AND FANS; VENTILATING HOODS 2.25 3.19 2.87 2.26 TEXTILE YARN 651 2.08 2.96 724 1.90 2.27 TEXTILE AND LEATHER MACHINERY, AND PARTS THEREOF, N.E.S. 1.93 3.02 MATERIALS OF RUBBER, INCLUDING PASTES, PLATES, SHEETS, RODS, THREAD, TUBES, 621 2.29 2.59 1.94 2.31 GLASSWARE 665 872 2.31 1.88 11.85 INSTRUMENTS AND APPLIANCES, N.E.S., FOR MEDICAL, SURGICAL, DENTAL OR VETERI ORGANIC CHEMICALS, N.E.S. 516 1.99 2.82 2.34 7.63 245 FUEL WOOD (EXCLUDING WOOD WASTE) AND WOOD CHARCOAL 2.35 2.02 574 2.11 2.35 POLYACETALS, OTHER POLYETHERS AND EPOXIDE RESINS, IN PRIMARY FORMS; POLYCAR 3.79 571 POLYMERS OF ETHYLENE, IN PRIMARY FORMS 2.36 2.08 2.83 2.00 735 PARTS AND ACCESSORIES SUITABLE FOR USE SOLELY OR PRINCIPALLY WITH METAL WOR 2.36 3.05 3.07 2.36 NONELECTRIC PARTS AND ACCESSORIES OF MACHINERY, N.E.S. 749 2.01 METAL CONTAINERS FOR STORAGE OR TRANSPORT 4.05 692 2.02 2.39 342 LIQUEFIED PROPANE AND BUTANE 2.05 11.05 2.41 524 INORGANIC CHEMICALS, N.E.S.; ORGANIC AND INORGANIC COMPOUNDS OF PRECIOUS ME 2.41 2.02 3.70 1.93 551 ESSENTIAL OILS, PERFUME AND FLAVOR MATERIALS 2.41 3.21 24.55 2.42 ZINC 686 2.05

4 277 – R. Ossa / Journal of International Economics 97 (2015) 266 269 Table 1 continued ) ( Sigma SITC description SITC code 95% CI TRUNKS, SUITCASES, VANITY CASES, BINOCULAR AND CAMERA CASES, HANDBAGS, WALL 831 2.16 3.21 2.42 2.10 2.43 MEDICINAL AND PHARMACEUTICAL PRODUCTS, OTHER THAN MEDICAMENTS (OF GROUP 542 541 3.05 2.01 3.07 WASTE, PARINGS AND SCRAP, OF PLASTICS 579 2.43 3.03 1.93 2.43 ELECTRIC POWER MACHINERY (OTHER THAN ROTATING ELECTRIC PLANT OF POWER GENER 771 PULP AND WASTE PAPER 2.45 3.46 251 2.18 2.98 2.09 2.45 PLATES, SHEETS, FILM, FOIL AND STRIP OF PLASTICS 582 VEGETABLES, FRESH, CHILLED, FROZEN OR SIMPLY PRESERVED; ROOTS, TUBERS AND O 2.46 3.66 2.24 54 3.71 2.07 2.46 FERTILIZER, CRUDE, EXCEPT THOSE OF DIVISION 56, (IMPORTS ONLY) 272 2.46 1.93 2.93 512 ALCOHOLS, PHENOLS, PHENOL-ALCOHOLS AND THEIR HALOGENATED, SULFONATED, NITRA 3.08 2.02 MONOFILAMENT WITH A CROSS-SECTIONAL DIMENSION EXCEEDING 1 MM, RODS, STICKS 583 2.46 2.47 2.24 3.77 ALUMINUM 684 CRUDE VEGETABLE MATERIALS, N.E.S. 2.48 2.05 3.28 292 INSECTICIDES, FUNGICIDES, HERBICIDES, PLANT GROWTH REGULATORS, ETC., DISINF 591 2.48 2.18 3.52 2.14 3.06 NITROGEN-FUNCTION COMPOUNDS 514 2.49 72 COCOA 2.50 2.13 3.78 FERROUS WASTE AND SCRAP; REMELTING INGOTS OF IRON OR STEEL 2.12 4.10 282 2.50 CEREAL PREPARATIONS AND PREPARATIONS OF FLOUR OR STARCH OF FRUITS OR VEGETA 2.52 2.32 3.52 48 3.98 2.52 EXPLOSIVES AND PYROTECHNIC PRODUCTS 593 2.03 PRINTING AND BOOKBINDING MACHINERY, AND PARTS THEREOF 2.52 2.03 3.60 726 744 MECHANICAL HANDLING EQUIPMENT, AND PARTS THEREOF, N.E.S. 2.52 1.89 3.83 334 PETROLEUM OILS AND OILS FROM BITUMINOUS MINERALS (OTHER THAN CRUDE), AND PR 2.55 2.05 3.55 672 2.55 2.05 3.55 IRON OR STEEL INGOTS AND OTHER PRIMARY FORMS, AND SEMIFINISHED PRODUCTS OF 2.27 2.56 WOOL AND OTHER ANIMAL HAIR (INCLUDING WOOL TOPS) 268 3.56 PAPER MILL AND PULP MILL MACHINERY, PAPER CUTTING MACHINES AND MACHINERY FO 1.88 3.98 725 2.56 786 TRAILERS AND SEMI-TRAILERS; OTHER VEHICLES, NOT MECHANICALLY PROPELLED; SPE 2.57 1.70 5.40 RESIDUAL PETROLEUM PRODUCTS, N.E.S. AND RELATED MATERIALS 2.05 4.05 335 2.58 MANMADE FIBERS, N.E.S. SUITABLE FOR SPINNING AND WASTE OF MANMADE FIBERS 2.60 2.20 4.09 267 3.70 2.60 TRANSMISSION SHAFTS AND CRANKS; BEARING HOUSINGS AND PLAIN SHAFT BEARINGS; 748 2.14 SOAP, CLEANSING AND POLISHING PREPARATIONS 2.62 2.22 3.25 554 884 OPTICAL GOODS, N.E.S. 2.62 2.28 3.31 2.63 581 TUBES, PIPES AND HOSES OF PLASTICS 2.19 3.34 776 2.63 2.00 3.86 THERMIONIC, COLD CATHODE OR PHOTOCATHODE VALVES AND TUBES; DIODES, TRANSIST 2.98 773 EQUIPMENT FOR DISTRIBUTING ELECTRICITY, N.E.S. 2.67 2.26 553 2.68 2.32 3.55 PERFUMERY, COSMETICS, OR TOILET PREPARATIONS, EXCLUDING SOAPS 791 RAILWAY VEHICLES (INCLUDING HOVERTRAINS) AND ASSOCIATED EQUIPMENT 2.68 2.01 6.49 3.58 2.13 SUGARS, MOLASSES, AND HONEY 61 2.70 733 MACHINE TOOLS FOR WORKING METAL, SINTERED METAL CARBIDES OR CERMETS, WITHOU 2.70 1.93 16.07 11.28 289 ORES AND CONCENTRATES OF PRECIOUS METALS; WASTE, SCRAP AND SWEEPINGS OF PRE 2.71 1.89 3.87 2.71 881 PHOTOGRAPHIC APPARATUS AND EQUIPMENT, N.E.S. 2.12 3.26 2.28 2.71 MISCELLANEOUS MANUFACTURED ARTICLES, N.E.S. 899 3.72 2.35 2.74 266 SYNTHETIC FIBERS SUITABLE FOR SPINNING 2.41 13.88 WOMEN'S OR GIRLS' COATS, CAPES, JACKETS, SUITS, TROUSERS, DRESSES, UNDERWEA 844 2.75 4.80 2.40 2.77 SHIPS, BOATS (INCLUDING HOVERCRAFT) AND FLOATING STRUCTURES 793 25.05 TRACTORS (OTHER THAN MECHANICAL HANDLING EQUIPMENT) 1.55 722 2.82 2.36 2.83 ORES AND CONCENTRATES OF BASE METALS, N.E.S. 287 3.52 FLOOR COVERINGS, ETC. 2.83 1.86 25.05 659 2.36 676 IRON AND STEEL BARS, RODS, ANGLES, SHAPES AND SECTIONS, INCLUDING SHEET PIL 2.84 3.52 1.81 7.34 2.87 NICKEL ORES AND CONCENTRATES; NICKEL MATTES, NICKEL OXIDE SINTERS AND OTHER 284 3.80 2.90 PREPARED ADDITIVES FOR MINERAL OILS ETC.; LIQUIDS FOR HYDRAULIC TRANSMISSIO 597 2.37 TIN 687 2.34 13.50 2.90 3.72 642 PAPER AND PAPERBOARD, CUT TO SIZE OR SHAPE, AND ARTICLES OF PAPER OR PAPERB 2.91 2.39 3.42 2.37 2.93 SYNTHETIC ORGANIC COLORING MATTER AND COLOR LAKES AND PREPARATIONS BASED TH 531 2.14 5.24 2.94 POLYMERS OF STYRENE, IN PRIMARY FORMS 572 6.28 2.95 MEAT AND EDIBLE MEAT OFFAL, PREPARED OR PRESERVED N.E.S. 17 2.30 74 TEA AND MATE 2.96 2.43 3.78 6.59 2.27 2.98 CORK MANUFACTURES 633 885 WATCHES AND CLOCKS 2.98 2.09 13.69 4.01 658 MADE-UP ARTICLES, WHOLLY OR CHIEFLY OF TEXTILE MATERIALS, N.E.S. 2.99 2.31 3.79 2.99 893 ARTICLES, N.E.S. OF PLASTICS 2.29 4.01 2.47 3.00 PIG IRON AND SPIEGELEISEN, SPONGE IRON, IRON OR STEEL GRANULES AND POWDERS 671 4.29 2.36 3.04 56 VEGETABLES, ROOTS AND TUBERS, PREPARED OR PRESERVED, N.E.S. 2.43 4.75 WORN CLOTHING AND OTHER WORN TEXTILE ARTICLES; RAGS 269 3.04 4.34 2.21 3.08 SPICES 75 891 3.08 2.27 7.39 ARMS AND AMMUNITION POLYMERS OF VINYL CHLORIDE OR OTHER HALOGENATED OLEFINS, IN PRIMARY FORMS 573 2.31 3.75 3.09 4.02 716 ROTATING ELECTRIC PLANT AND PARTS THEREOF, N.E.S. 3.11 2.36 36 2.37 3.13 CRUSTACEANS MOLLUSCS,AQUTC INVRTBRTS FRSH (LVE/DEAD) CH SLTD ETC.; CRUSTACE 4.04 4.70 222 OIL SEEDS AND OLEAGINOUS FRUITS USED FOR THE EXTRACTION OF SOFT FIXED VEGET 3.13 2.53 25.05 3.16 897 JEWELRY, GOLDSMITHS' AND SILVERSMITHS' WARES, AND OTHER ARTICLES OF PRECIOU 2.40 5.53 2.53 3.17 CEREAL MEALS AND FLOURS, N.E.S. 47 OIL SEEDS AND OLEAGINOUS FRUITS, WHOLE OR BROKEN, OF A KIND USED FOR EXTRAC 3.93 1.99 3.19 223 742 PUMPS FOR LIQUIDS, WHETHER OR NOT FITTED WITH A MEASURING DEVICE; LIQUID EL 1.97 4.71 3.19 10.62 261 SILK TEXTILE FIBERS 3.20 2.46 3.91 3.24 98 EDIBLE PRODUCTS AND PREPARATIONS, N.E.S. 2.55 4.22 737 METALWORKING MACHINERY (OTHER THAN MACHINE TOOLS) AND PARTS THEREOF, N.E.S. 2.26 3.25 764 TELECOMMUNICATIONS EQUIPMENT, N.E.S.; AND PARTS, N.E.S., AND ACCESSORIES OF 3.25 2.51 4.19 (continued on next page)

5 277 – R. Ossa / Journal of International Economics 97 (2015) 266 270 Table 1 continued ) ( Sigma SITC description SITC code 95% CI 874 2.55 MEASURING, CHECKING, ANALYSING AND CONTROLLING INSTRUMENTS AND APPARATUS, N 4.31 3.25 2.40 3.26 CARBOXYLIC ACIDS AND ANHYDRIDES, HALIDES, PEROXIDES AND PEROXYACIDS; THEIR 513 4.22 2.08 4.56 MUSICAL INSTRUMENTS, PARTS AND ACCESSORIES THEREOF; RECORDS, TAPES AND OTHE 898 3.26 23.30 2.91 3.29 MEAT OF BOVINE ANIMALS, FRESH, CHILLED OR FROZEN 11 OPTICAL INSTRUMENTS AND APPARATUS, N.E.S. 3.29 5.18 871 2.49 5.61 2.77 3.29 WORKS OF ART, COLLECTORS' PIECES AND ANTIQUES 896 FIXED VEGETABLE FATS AND OILS (OTHER THAN SOFT), CRUDE, REFINED OR FRACTION 3.30 6.39 2.47 422 3.74 2.43 3.30 PLASTICS, N.E.S., IN PRIMARY FORMS 575 3.31 2.31 5.20 1 LIVE ANIMALS OTHER THAN ANIMALS OF DIVISION 03 5.08 2.73 POWER GENERATING MACHINERY AND PARTS THEREOF, N.E.S. 718 3.34 3.39 2.42 7.50 CLOTHING ACCESSORIES, OF TEXTILE FABRICS, WHETHER OR NOT KNITTED OR CROCHET 846 FURNITURE AND PARTS THEREOF; BEDDING, MATTRESSES, MATTRESS SUPPORTS, CUSHIO 3.41 2.56 4.73 821 SPECIAL TRANSACTIONS AND COMMODITIES NOT CLASSIFIED ACCORDING TO KIND 931 3.42 2.60 15.02 2.55 5.99 RICE 42 3.43 285 ALUMINUM ORES AND CONCENTRATES (INCLUDING ALUMINA) 3.43 2.23 5.43 IRON AND NONALLOY STEEL FLAT-ROLLED PRODUCTS, CLAD, PLATED OR COATED 2.16 4.38 674 3.43 COAL, PULVERIZED OR NOT, BUT NOT AGGLOMERATED 3.44 2.50 5.60 321 5.56 3.44 AGRICULTURAL MACHINERY (EXCLUDING TRACTORS) AND PARTS THEREOF 721 1.98 CINEMATOGRAPHIC FILM, EXPOSED AND DEVELOPED, WHETHER OR NOT INCORPORATING S 3.48 2.15 27.26 883 44 MAIZE (NOT INCLUDING SWEET CORN) UNMILLED 3.51 2.62 5.99 57 FRUIT AND NUTS (NOT INCLUDING OIL NUTS), FRESH OR DRIED 3.55 2.70 4.14 122 3.56 2.24 7.80 TOBACCO, MANUFACTURED (WHETHER OR NOT CONTAINING TOBACCO SUBSTITUTES) 2.27 3.56 KNITTED OR CROCHETED FABRICS (INCLUDING TUBULAR KNIT FABRICS, N.E.S., PILE 655 6.34 STEAM OR OTHER VAPOR GENERATING BOILERS, SUPER-HEATED WATER BOILERS AND AUX 2.34 10.50 711 3.56 666 POTTERY 3.57 2.92 4.11 NONFERROUS BASE METAL WASTE AND SCRAP, N.E.S. 2.87 6.86 288 3.58 CEREALS, UNMILLED (OTHER THAN WHEAT, RICE, BARLEY AND MAIZE) 3.60 2.60 7.30 45 6.01 3.61 MILK AND CREAM AND MILK PRODUCTS OTHER THAN BUTTER OR CHEESE 22 3.08 FEEDING STUFF FOR ANIMALS (NOT INCLUDING UNMILLED CEREALS) 3.61 2.34 5.01 81 714 ENGINES AND MOTORS, NONELECTRIC (OTHER THAN STEAM TURBINES, INTERNAL COMBUS 3.63 1.93 25.05 3.65 421 FIXED VEGETABLE FATS AND OILS, SOFT, CRUDE, REFINED OR FRACTIONATED 2.81 5.33 683 3.65 2.75 7.01 NICKEL 5.67 728 MACHINERY AND EQUIPMENT SPECIALIZED FOR PARTICULAR INDUSTRIES, AND PARTS TH 3.78 2.09 745 3.78 2.38 5.85 NONELECTRICAL MACHINERY, TOOLS AND MECHANICAL APPARATUS, AND PARTS THEREOF, 112 ALCOHOLIC BEVERAGES 3.79 1.92 9.28 6.79 2.27 FOOD-PROCESSING MACHINES (EXCLUDING DOMESTIC) 727 3.84 894 BABY CARRIAGES, TOYS, GAMES AND SPORTING GOODS 3.88 2.73 5.78 333 PETROLEUM OILS AND OILS FROM BITUMINOUS MINERALS, CRUDE 3.96 2.55 6.31 2.55 411 ANIMAL OILS AND FATS 3.96 15.89 22.60 3.97 MANUFACTURES OF LEATHER OR COMPOSITION LEATHER, N.E.S.; SADDLERY AND HARNES 612 2.51 7.01 244 1.86 3.98 CORK, NATURAL, RAW AND WASTE (INCLUDING NATURAL CORK IN BLOCKS OR SHEETS) 2.05 10.56 WHEAT (INCLUDING SPELT) AND MESLIN, UNMILLED 41 4.00 6.07 2.82 4.04 MEAT, OTHER THAN OF BOVINE ANIMALS, AND EDIBLE OFFAL, FRESH, CHILLED OR FRO 12 4.60 COPPER 2.57 682 4.04 1.55 4.05 JUTE AND OTHER TEXTILE BAST FIBERS, N.E.S., RAW OR PROCESSED BUT NOT SPUN; 264 6.55 MEDICAMENTS (INCLUDING VETERINARY MEDICAMENTS) 4.05 2.75 6.36 542 2.84 58 FRUIT PRESERVED, AND FRUIT PREPARATIONS (EXCLUDING FRUIT JUICES) 4.07 6.20 3.05 21.80 4.08 BARLEY, UNMILLED 43 5.22 4.10 CHOCOLATE AND OTHER FOOD PREPARATIONS CONTAINING COCOA, N.E.S. 73 3.36 MEN'S OR BOYS' COATS, JACKETS, SUITS, TROUSERS, SHIRTS, UNDERWEAR ETC. OF W 841 3.19 5.73 4.11 6.00 71 COFFEE AND COFFEE SUBSTITUTES 4.19 2.79 5.63 2.80 4.19 ELECTRO-DIAGNOSTIC APPARATUS FOR MEDICAL, SURGICAL, DENTAL OR VETERINARY SC 774 2.11 14.34 4.20 MACHINE TOOLS WORKING BY REMOVING METAL OR OTHER MATERIAL 731 6.71 4.23 WOMEN'S OR GIRLS' COATS, CAPES, JACKETS, SUITS, TROUSERS, DRESSES, SKIRTS, 842 2.71 59 FRUIT JUICES (INCL. GRAPE MUST) AND VEGETABLE JUICES, UNFERMENTED AND NOT C 4.27 2.63 7.21 4.90 2.92 4.28 METAL STRUCTURES AND PARTS, N.E.S., OF IRON, STEEL OR ALUMINUM 691 713 INTERNAL COMBUSTION PISTON ENGINES AND PARTS THEREOF, N.E.S. 4.30 2.58 5.95 712 STEAM TURBINES AND OTHER VAPOR TURBINES, AND PARTS THEREOF, N.E.S. 4.35 2.57 20.75 2.67 322 BRIQUETTES, LIGNITE AND PEAT 4.45 12.27 7.30 4.45 FOOTWEAR 851 2.99 12.59 121 3.64 4.51 TOBACCO, UNMANUFACTURED; TOBACCO REFUSE 3.10 6.27 SANITARY, PLUMBING AND HEATING FIXTURES AND FITTINGS, N.E.S. 812 4.56 21.35 2.21 4.60 WOVEN FABRICS OF MANMADE TEXTILE MATERIALS (NOT INCLUDING NARROW OR SPECIAL 653 747 4.62 2.86 5.14 TAPS, COCKS, VALVES AND SIMILAR APPLIANCES FOR PIPES, BOILER SHELLS, TANKS, METERS AND COUNTERS, N.E.S. 873 2.77 7.72 4.63 34.45 25 BIRDS' EGGS AND EGG YOLKS, FRESH, DRIED OR OTHERWISE PRESERVED, SWEETENED O 4.73 2.13 246 2.68 4.77 WOOD IN CHIPS OR PARTICLES AND WOOD WASTE 6.49 813 LIGHTING FIXTURES AND FITTINGS, N.E.S. 4.88 2.85 6.05 3.04 35 FISH, DRIED, SLTD R IN BRINE; SMKD FISH (WHETHR R NT COOKD BEFORE OR DURNG 4.92 12.74 7.50 4.97 ARTICLES OF APPAREL AND CLOTHING ACCESSORIES OF OTHER THAN TEXTILE FABRICS; 848 2.64 PEARLS, PRECIOUS AND SEMIPRECIOUS STONES, UNWORKED OR WORKED 25.05 667 1.87 5.11 24 CHEESE AND CURD 3.66 7.31 5.13 46 MEAL AND FLOUR OF WHEAT AND FLOUR OF MESLIN 5.19 3.54 9.95 3.36 23 BUTTER AND OTHER FATS AND OILS DERIVED FROM MILK 5.26 8.82 7.02 5.26 SOUND RECORDERS OR REPRODUCERS; TELEVISION IMAGE AND SOUND RECORDERS OR REP 763 3.50

6 277 – 271 R. Ossa / Journal of International Economics 97 (2015) 266 ( continued Table 1 ) Sigma SITC code 95% CI SITC description 1.82 20.38 611 5.30 LEATHER 677 5.63 2.35 14.13 IRON AND STEEL RAILS AND RAILWAY TRACK CONSTRUCTION MATERIAL 247 2.38 25.82 WOOD IN THE ROUGH OR ROUGHLY SQUARED 5.64 OFFICE AND STATIONERY SUPPLIES, N.E.S. 2.50 7.34 895 5.79 5.84 2.73 11.16 625 RUBBER TIRES, INTERCHANGEABLE TIRE TREADS, TIRE FLAPS AND INNER TUBES FOR W ARTICLES OF APPAREL, OF TEXTILE FABRICS, WHETHER OR NOT KNITTED OR CROCHETE 845 3.45 11.71 6.10 16 6.35 2.92 11.58 MEAT AND EDIBLE MEAT OFFAL, SALTED, IN BRINE, DRIED OR SMOKED; EDIBLE FLOUR AUTOMATIC DATA PROCESSING MACHINES AND UNITS THEREOF; MAGNETIC OR OPTICAL R 3.46 7.98 752 6.40 525 RADIOACTIVE AND ASSOCIATED MATERIALS 6.51 2.35 40.13 SUGAR CONFECTIONERY 6.85 62 18.20 3.36 971 6.88 2.48 80.04 GOLD, NONMONETARY (EXCLUDING GOLD ORES AND CONCENTRATES) PRINTED MATTER 7.13 3.49 11.16 892 751 15.23 3.38 7.83 OFFICE MACHINES 761 TV RECEIVERS (INCLUDING VIDEO MONITORS & PROJECTORS) WHETH R NT INCORP RADI 4.40 20.13 7.88 MEN'S OR BOYS' COATS, CAPES, JACKETS, SUITS, BLAZERS, TROUSERS, SHIRTS, ETC 3.56 16.24 843 7.97 681 SILVER, PLATINUM AND OTHER PLATINUM GROUP METALS 8.25 3.18 70.33 COPPER ORES AND CONCENTRATES; COPPER MATTES; CEMENT COPPER 8.52 3.45 25.05 283 37 FISH, CRUSTACEANS, MOLLUSCS AND OTHER AQUATIC INVERTEBRATES, PREPARED OR PR 3.49 14.61 8.73 CUTLERY 10.70 696 21.20 4.38 652 COTTON FABRICS, WOVEN (NOT INCLUDING NARROW OR SPECIAL FABRICS) 10.95 7.39 30.97 762 RADIO-BROADCAST RECEIVERS, WHETHER OR NOT INCORPORATING SOUND RECORDING OR 12.13 5.27 19.74 613 FURSKINS, TANNED OR DRESSED (INCLUDING PIECES OR CUTTINGS), ASSEMBLED OR UN 2.05 40.62 12.59 AIRCRAFT AND ASSOCIATED EQUIPMENT; SPACECRAFT (INCLUDING SATELLITES) AND SP 16.55 792 39.29 6.55 91 MARGARINE AND SHORTENING 18.05 3.05 44.81 781 MOTOR CARS AND OTHER MOTOR VEHICLES PRINCIPALLY DESIGNED FOR THE TRANSPORT 21.55 1.95 25.05 782 MOTOR VEHICLES FOR THE TRANSPORT OF GOODS AND SPECIAL PURPOSE MOTOR VEHICLE 25.05 2.05 47.20 MEAN 3.63 2.32 8.16 imports in some industries are critical to the functioning of the – Since theidealpriceindex fortheaggregate fi nal good is justa Cobb 8 economy. fi Douglas aggregate of the ideal price indices of the industry-speci c α S js Notice that this point is overlooked if the aggregate formula is used. goods, P P , the above solution for P implies an expression ∏ ¼ js j ¼ s 1 js 1 1 − β σ 1 − b w j j for real income which is just in terms of technology parameters and λ ¼ 1is ,where σ − In thespecialcase es to =1,Eq. (7) simpli fi S P j j s 1 α − δ js w jt S S − σ 1 j t now the aggregate trade elasticity. If the multi-industry model is cor- λ ¼ A ,whereIhave ∏ ∏ trade shares. In particular, j t 1 ¼ s ¼ 1 P jt j s − 1 is some weighted average of rect, the aggregate trade elasticity σ δ α js S S jt to simplify the notation. Since λ A ≡∏ =1forall ∏ A de fi ned js j ¼ t ¼ 1 1 s jt the industry-level trade elasticities σ − 1 because the latter ultimately s s under autarky, the proportional gains of moving from autarky to current fl ows respond to trade costs. Loosely speaking, the govern how trade s 1 − α δ b js w jt S S − σ 1 exponent of the aggregate formula is therefore the inverse of the average j t ¼ ∏ ∏ .To λ levels of trade are captured by the formula ¼ ¼ 1 1 t s p jt j of the trade elasticities whereas the exponent of the industry-level for- be able to clearly contrast this to the aggregate formula, I implicitly de ne fi average of the inverse of the trade elasticities which is differ- mula is the s 1 α − δ js 9 jt S S σ − 1 x 6 t ent as long as the elasticities vary across industries. λ x and solve for ∏ , which then implies ≡∏ λ 1 1 t ¼ s ¼ j jt In the empirical application, I report results using the industry-level and aggregate formulas discussed above. In addition, I also consider the X X ln λ 1 S S jt s simpler formulas which arise in the special case without non-traded α δ − js jt 1 ¼ t s 1 ¼ c w σ 1 ln λ − j t j and intermediate goods. While non-traded goods tend to dampen the ¼ ð 7 Þ λ j P j gains from trade, intermediate goods tend to amplify them so that abstracting from both turns out to be a reasonable fi rst pass. I remove non-traded goods by simply narrowing down the set of included indus- For the purposes of calculating the gains from trade, the correct tries, as I discuss below. I remove intermediate goods by considering approach is therefore to take a weighted average of the inverse of the the special case with β and s . which yields the modi fi ed = 1 for all i is 1 . The weights capture how depen- industry-level trade elasticities − 1 σ ln λ 1 t S js − α ∑ js ln λ 1 ¼ s jt 1 σ − λ ln 1 b b , t is on trade in industry j dent country j , how dependent country − w w s j j j − 1 σ ln λ j formulas λ ¼ . and λ ¼ P P j j j j is on upstream industry for producing fi nal output in downstream in- t s , and how important industry fi nal consumers in coun- is to s dustry s , δ jt 8 Imbs In the context of their discussion of aggregation biases in elasticity estimations, b w 7 j seem to conjecture that the gains from trade estimated using the ag- and Mejean (2015) ∞ as σ → 1 in some industries as long as . → As a consequence, try α j , js t P j gregate formula would be the same as the gains from trade estimated using the ln λ s jt industry-level formula if the aggregate trade elasticity is estimated using a method which is strictly positive there. While Eq. is admittedly based on (7) δ α js jt λ ln j does not suffer from aggregation bias. A simple thought experiment reveals that this can- w very special assumptions, it nevertheless captures what has to be a gen- j σ not bethe case. Inparticular, supposethat 1 inone industrysothat → ∞ as discussed → t P j eral point: even if imports in the average industry do not matter too 'stradeelasticity iszero, it inthemaintext. Whilethis situation would imply that industry t would certainly not imply that any reasonably measured aggregate trade elasticity is zero, much, a complete shutdown of international trade is still very costly, if which would be required, however, for the aggregate formula to correctly predict in fi nite gains from trade. 9 Notice that this can also be understood in terms of the familiar Jensen's inequality. To be able to use the aggregate formula, one essentially has to compute the aggregate trade S X ∑ X jjs 6 jjs 1 − 1 s ¼ 1 ≡ . λ To be clear, λ ≡ and = ε elasticity as f f E ( ε . )], where f ð ε Þ¼ [ ε is a convex and decreasing function of js j s s s N S N ε s X ∑ X ∑ ∑ ijs ijs 1 ¼ i 1 ¼ s ¼ i 1 ≤ As a result, ε E [ ε E [ ] represents the weighted arithmetic ε ] by Jensen's inequality, where s s λ ln λ − 1 7 js js average that is implicitly estimated when estimating aggregate trade elasticities. I would are the shares of industry-level and λ − and that 1 ≈ Notice that λ − and 1 j js 1 λ − λ ln j j like to thank a referee for suggesting to point this out. aggregate imports in country j 's total expenditure.

7 277 R. Ossa / Journal of International Economics 97 (2015) 266 – 272 Table 2 expenditure shares α , the shares of value added in gross production js t Gains from trade. β – γ output matrices , the elements of the input , and the elasticities js js of substitution σ . My main data source is the eighth version of the Adjusted Unadjusted s Global Trade Analysis Project database (GTAP 8) which I supplement True gain Ratio Naive gain True gain Ratio Naive gain UN trade data from the time periods – with the widely used NBER (%) (%) (%) (%) 1994 – 2008 when I need time variation or a fi ner disaggregation 4.2 United Arab 39.8 133.2 3.3 35.9 148.8 of industries. The GTAP 8 database is a carefully cleaned, fully Emirates documented, publicly available, and globally consistent database 3.1 9.6 31.5 3.3 Argentina 28.3 9.2 13.1 Australia 3.0 35.9 2.7 9.7 28.7 3.5 103.4 32.1 Austria 3.2 27.1 95.5 Belgium 8.5 505.2 4.9 59.5 259.9 53.3 2.1 9.5 Brazil 4.7 9.8 1.9 4.9 19.0 Canada 44.0 3.0 2.8 14.4 53.6 Table 3 Switzerland 39.0 134.6 3.5 24.1 4.6 111.0 Decomposition of the gains from trade. 3.9 16.0 67.0 17.0 Chile 6.8 109.0 Adjusted Unadjusted 2.2 5.7 12.9 2.2 13.8 30.8 China 30.8 9.5 Colombia 3.2 7.6 29.2 3.8 True gain Lambda Exponent True gain Lambda Exponent 3.6 Czech Republic 22.6 71.4 3.2 38.0 137.4 (%) (%) (%) (%) 2.5 17.7 40.2 2.3 18.5 Germany 45.7 79.2 Denmark 3.0 25.4 3.0 75.4 26.5 0.86 148.82 64.46 − 133.19 37.29 United Arab 2.08 − Spain 3.5 53.4 3.2 15.4 52.0 16.4 Emirates 3.1 17.2 52.6 3.1 22.0 68.0 Finland 31.46 88.46 2.23 Argentina 28.28 77.18 − 0.96 − 39.2 15.0 2.6 13.1 2.7 France 35.3 0.85 − 2.31 − Australia 28.68 89.65 35.93 69.59 44.7 United Kingdom 18.3 2.4 12.6 2.5 31.8 103.43 44.05 − 0.87 95.50 74.06 − 2.23 Austria 3.5 19.4 Greece 20.8 72.6 121.9 6.3 4.04 1.02 505.22 64.02 − − 259.88 28.45 Belgium Hungary 26.0 86.5 3.3 45.4 166.1 3.7 Brazil − − 1.31 9.76 87.39 9.48 93.34 0.69 3.2 35.6 3.0 11.3 25.2 8.3 Indonesia − Canada 53.57 59.94 0.84 43.97 83.75 − 2.06 India 20.9 1.9 11.2 13.7 7.3 1.9 2.61 134.65 37.95 − 0.88 111.00 75.12 − Switzerland 3.2 Ireland 31.7 99.2 3.1 41.9 134.5 66.97 62.92 4.00 − 1.11 108.98 83.16 − Chile 3.2 11.7 28.5 8.9 Iran, Islamic Rep. 50.3 4.3 2.41 30.78 89.48 0.74 China 12.86 84.86 − − 115.0 3.9 21.7 77.5 3.6 Israel 29.4 2.42 Colombia 30.85 76.50 − 1.00 29.20 89.97 − 2.8 11.1 32.7 2.9 13.6 38.1 Italy 71.42 54.93 − 0.90 137.42 74.15 − 2.89 Czech Republic Japan 7.8 25.7 3.3 7.1 21.4 3.0 Germany − 40.16 80.83 1.59 0.75 − 45.70 60.73 42.7 3.5 21.3 65.4 3.1 Korea, Republic of 12.3 79.21 50.01 − 0.84 75.38 75.61 − 2.01 Denmark Mexico 15.0 45.0 3.0 11.3 3.0 33.9 − Spain 51.99 63.98 − 2.41 53.40 83.71 0.94 219.0 3.2 46.8 74.1 22.8 Malaysia 4.7 2.37 − 67.97 80.35 0.90 − 52.60 62.64 Finland Nigeria 5.0 13.2 70.9 5.4 10.5 52.6 35.35 86.27 − France 39.19 66.26 − 0.80 2.05 2.8 Netherlands 52.1 26.2 79.8 3.0 18.8 − 0.75 31.80 85.06 United Kingdom 44.68 61.03 − 1.71 51.0 3.2 14.9 3.4 63.3 Norway 19.7 Greece − 3.36 0.98 121.91 78.89 − 72.64 57.37 New Zealand 11.7 30.6 2.6 11.5 32.3 2.8 Hungary 2.66 − 0.92 166.14 69.18 − 86.53 50.62 9.5 Pakistan 3.8 12.8 61.9 4.8 36.7 35.58 86.93 − 2.17 0.95 − Indonesia 25.16 79.04 Philippines 18.5 127.8 57.7 3.1 23.0 5.5 − 0.62 − 13.73 81.18 India 1.50 20.91 88.09 16.6 47.7 3.4 2.9 21.1 72.0 Poland Ireland 99.17 44.47 − − 2.03 0.85 134.52 65.75 3.9 59.6 Portugal 18.8 3.2 19.1 75.0 Iran, Islamic Rep. 28.48 77.76 − 50.29 84.97 − 2.50 1.00 3.4 15.3 44.1 2.9 20.5 70.0 Romania 114.97 46.86 Israel − 1.01 77.49 78.02 − 2.31 35.5 Rest of the qorld 16.3 2.2 21.9 56.6 2.6 − 32.70 73.27 Italy 2.26 − 38.05 86.69 0.91 Russian Federation 9.1 25.1 2.7 10.8 34.9 3.2 2.30 25.68 80.21 Japan 1.04 21.43 91.91 − − Saudi Arabia 3.2 68.1 3.3 21.1 49.6 14.9 2.72 Korea, Republic of 42.74 71.15 − 1.05 65.43 83.12 − 3.8 73.1 218.3 57.2 4.9 361.7 Singapore Mexico 1.81 − 33.92 85.14 0.90 − 44.99 66.28 Sweden 55.3 2.7 21.2 57.5 21.4 2.6 − Malaysia 0.92 219.00 70.58 − 3.33 74.13 54.57 Thailand 2.5 89.0 2.7 35.5 51.3 19.1 Nigeria − 1.44 70.91 76.39 − 1.99 52.59 74.54 3.3 3.0 12.3 37.5 41.0 Turkey 12.6 0.86 − Netherlands 79.77 50.39 − 52.10 81.01 1.99 Ukraine 22.3 86.7 3.9 31.4 174.3 5.6 2.18 − 51.03 82.77 Norway 0.93 − 63.33 58.92 United States 9.9 19.4 2.0 6.4 2.1 13.5 0.82 2.11 − 32.30 87.60 New Zealand 30.58 72.19 − Venezuela, RB 27.9 3.3 8.4 9.2 41.0 4.5 1.16 Pakistan 36.70 76.45 − 61.90 85.77 − 3.14 2.7 14.6 30.5 42.3 11.2 South Africa 2.9 Philippines 57.71 60.66 − 0.91 127.80 76.05 − 3.01 16.5 3.1 16.9 55.9 Median 48.6 3.3 Poland − 72.01 80.79 0.86 − 47.69 63.63 2.54 59.58 60.26 − 0.92 74.97 81.67 − Portugal 2.76 Note: This table summarizes the changes in real income resulting from a move from autar- 44.12 65.82 − 0.87 69.98 80.33 − 2.42 Romania ky to year 2007 levels of trade. The results under “ True gain are computed using the in- ” 0.68 − 1.87 56.56 78.65 Rest of the world 35.47 64.15 − “ are computed using the aggregate ” Naive gain dustry-level formulas, the results under − 34.86 88.00 0.87 − 25.08 77.31 Russian 2.34 formulas, and the results under “ Ratio ” simply compute the ratio of the two. Columns 1 – Federation 6 do. I include – 3 do not adjust for non-traded or intermediate goods while columns 4 Saudi Arabia − 0.99 68.06 72.71 − 49.60 66.51 1.63 fi Hong Kong in my de nition of China. 218.27 26.39 2.95 − 0.87 361.71 59.58 − Singapore 1.97 − Sweden 57.53 56.43 0.79 55.31 79.96 − 1.94 0.81 Thailand 51.27 59.82 − 88.97 72.09 − − Turkey 37.55 70.47 − 0.91 40.97 84.68 2.06 3. Data 1.05 174.28 76.04 Ukraine 86.75 55.28 − − 3.68 1.42 − − United States 19.38 75.73 13.47 91.47 0.64 I focus on the world's 49 largest economies and a residual Rest of Venezuela, RB 27.95 78.79 − 1.03 40.97 88.37 − 2.78 10 the World in the year 2007. To quantify the gains from trade using − South Africa 30.53 73.18 0.85 42.28 86.28 − 2.39 2.30 55.93 82.22 48.64 63.80 Median − 0.90 − (7) formula fl ows to , I need the full matrix of industry-level trade compute the statistics λ and as well as estimates of the consumption λ . Table 2 Note: This table provides more details on the calculation of the gains from trade in j js Inparticular, itagain lists the gains from tradecomputedusing the industry-levelformulas . Notice that the gains and and the exponent from formula λ and explicitly shows the (7) the λ are expressed as percentages so that “ True gain (%) ” =100*(( “ Lambda (%) ” / 10 “ 3 do not adjust for non-traded or intermediate 1). Columns 1 − ” Exponent – 100)^ I ranked countries by GDP as reported in the World Bank's World Development – goods while columns 4 6do. Indicators.

8 277 – R. Ossa / Journal of International Economics 97 (2015) 266 273 300 BEL 250 SGP 200 150 CHE ARE ISR Gains from trade in % AUT 100 IRL UKR HUN NLD DNK MYS GRC CZE CHL NOR PRT PHL SWE CAN FIN NGA ESP THA SAU 50 POL DEU MEX GBR ROU KOR FRA TUR PAK AUS ROW ITA COL NZL ZAF IRN ARG VEN JPN IDN RUS USA IND CHN BRA 0 40 50 20 30 80 70 60 90 Lambda (share of trade with oneself) in % Gains from trade without non-traded and intermediate goods. Fig. 1. share of each bilateral GTAP industry trade fl ow should be attributed covering 129 countries and 57 industries which span all sectors of 11 UN data ow from the NBER fl to each bilateral SITC-Rev3 3-digit trade – the economy. and then superimpose these shares onto the GTAP 8 data so that every- It is not obvious at what level of aggregation my analysis should thing aggregates back to the GTAP 8 data in the end. Since internal trade be performed. On the one hand, the main point of the paper is that UN data,this strategy only worksfor – ows are not reported in the NBER fl excessive aggregation is likely to introduce biases which suggests ows to international trade ows and I simply apportion internal trade fl fl that a low level of aggregation should be preferred. On the other SITC-Rev3 3-digit sectors uniformly. Douglas assumptions in consumption ( – hand, my Cobb α is constant) js t – The GTAP 8 data includes input output accounts for all included and production ( γ is constant) seem less reasonable the narrower js t γ countries which I use to calculate the industry classi fi cation which suggests that disaggregating too fi nely . One problem with β and js js is problematic as well. Since departing from the Cobb Douglas assump- – sesisthattheyseparate fi input rms' – output accounts for my purpo tion seems challenging particularly on the production side where it is purchases into intermediate consumption (which is reported in the – output accounts, I choose the natural interpretation of national input – output tables for each upstream – down- main body of the input the SITC-Rev3 3-digit level as a compromise but also report results at xed investment (which is reported in a stream industry pair) and fi a higher level of aggregation as a sensitivity check. After constructing separate column of the input output tables for each upstream industry – a cross-walk between the GTAP 8 data and the NBER UN data, I am – only) depending on how fi rms treat these purchases in their balance left with 251 industries from agriculture, mining, and manufacturing sheets. Since I do not explicitly allow for investment in my model, I and a residual one aggregating all other industries available in the rms' intermediate consumption by the fi scale all entries referring to GTAP database. total investment to intermediate consumption ratio of the correspond- UN data is originally at the SITC-Rev2 4-digit level and I The NBER – ing upstream industry to obtain a more accurate picture of what fi rms convert it to the SITC-Rev3 3-digit level using a concordance from the actually buy. Center for International Data at UC Davis. I then match the SITC-Rev3 For example, for each piece of “ ” other machinery and equipment 3-digit industries to the GTAP industries using a concordance which I classi fi ed as intermediate consumption in the US, there are 0.8 addi- manually constructed with the help of various concordances available xed investment on average, and I scale all fi ed as fi tional piece classi cation focuses on fi from the GTAP website. By design, the SITC classi intermediate consumption values in the input output matrix by – traded goods only so that the residual industry aggregates over the re- 1.8 to account for this. Using this scaled data, I then simply read off maining industries of the economy which have relatively little trade the share of intermediate consumption spending of downstream in- t λ (the residual industry has an average t on upstream industry dustry γ , s of 0.94 compared to an average , as well as the associated share of is js value added in gross production, λ β . Finally, I disaggregate to the of 0.63 elsewhere and includes sectors such as construction and ser- is js vices). I will therefore refer to the residual industry as the non-traded SITC-Rev3 3-digit level by applying all shares uniformly across sub-industries. industry in the following even though I will actually treat it as a traded S F F I calculate α industry with little trade. = E α E from the relationship / , where ∑ js js =1 jt js t F F To construct λ E s goods in country and .Ofcourse, E λ is is fi nal expenditure on industry , I disaggregate the GTAP 8 data using bilateral j j js js is simply the difference between total expenditure and intermediate ex- trade shares from the NBER – UN data. In particular, I calculate what S F t I , N t penditure, E X −∑ ∑ X = γ , where the total expendi- =1 j m js js t =1 mjs I , t t ture of downstream industry X , , can be calculated from the j I , t N 11 equilibrium relationship X − ∑ =(1 ) X β . One problem with jt n =1 jnt j The database is documented in Narayanan et al. (2012) which can be accessed directly from the GTAP website under https://www.gtap.agecon.purdue.edu . α this approach is that some turn out to be negative, essentially implying js

9 277 274 – R. Ossa / Journal of International Economics 97 (2015) 266 600 BEL 500 400 SGP 300 Gains from trade in % MYS 200 UKR HUN ARE CZE IRL PHL GRC CHE CHL 100 AUT THA ISR DNK PRT POL NGA ROU SAU FIN KOR PAK ROW SWE ESP NLD NOR IRN CAN ZAF VEN TUR DEU ITA IDN FRA RUS MEX NZL GBR ARG CHN COL AUS JPN IND USA BRA 0 60 90 55 95 65 85 80 75 70 Lambda (share of trade with oneself) in % Gains from trade with non-traded and intermediate goods. Fig. 2. 4. Estimation that the abovementioned strategy of uniformly applying all GTAP- t industry-level β to the corresponding SITC-Rev3 3-digit level and γ js js t γ sub-industries does not always work. In those cases, I scale such js – Using the abovementioned NBER UN bilateral trade data for the N S t t ; I t t t ~ that α Þ γ γ ∑ with , X X = 0 by replacing = ∑ ¼ð γ γ using the 2008, I estimate the elasticities of substitution σ years 1994 – js js mjs ¼ ¼ t m 1 1 js s js js j t S t Broda and fi and re ned by method developed by Feenstra (1994) ∑ = 1, and repeat this process again to ensure γ γ then scale js =1 s js for all 251 matched SITC-Rev3 3-digit traded indus- Weinstein (2006) until all α ≥ 0. Overall, this only leads to minor corrections with the js t σ tries (I simply use the average for the residual non-traded industry). correlation between the original and the adjusted γ being 99.9%. s js 2 BEL 1.8 1.6 SGP 1.4 1.2 MYS UKR 1 HUN ARE CZE IRL PHL GRC 0.8 CHE CHL AUT THA 0.6 ISR DNK PRT POL NGA ROU SAU FIN KOR PAK Log gains from trade with non-traded and intermediate goods ROW SWE ESP NLD NOR IRN 0.4 CAN ZAF TUR VEN DEU ITA IDN FRA RUS MEX NZL GBR ARG CHN COL AUS 0.2 JPN IND USA BRA 0 0.2 1.6 2 1.8 0 1.4 1.2 1 0.8 0.6 0.4 Log gains from trade without non-traded and intermediate goods Fig. 3. Gains from trade with and without non-traded and intermediate goods.

10 277 – R. Ossa / Journal of International Economics 97 (2015) 266 275 not change the gains from trade estimates that much. On the one This method identi fi es the elasticities from variation in the variances hand, including non-traded industries raises the median own trade and covariances of demand and supply shocks across countries and share from 63.8% to 82.2% which tends to dampen the gains from Feenstra (2010) over time. I base my estimation on the instructions in trade. On the other hand, including intermediate goods increases the in which the method is particularly clearly explained. My estimating − 2.3 which tends to magnify the median exponent from − 0.9 to Feenstra (2010) which I estimate using equation is equation (2.21) in gains from trade. On average, these two forces are roughly offsetting weighted least squares following the code provided in Appendix 2.2 of fi so that the unadjusted special case provides a reasonable rst pass. Col- . However, I do not focus on a single importer, but Feenstra (2010) umns 2 and 3 further reveal that most of the variation in the unadjusted pool across the 49 importers considered in my analysis (I keep all gains from trade is due to variation in γ exporters available in the data). This is not only consistent with my the- , while columns 5 and 6 point j σ oretical assumption that out that variation in the exponent is more pronounced in the presence does not vary by country but also gives me a s of non-traded and intermediate goods. much larger dataset with over 5 million price quantity pairs. – – relates the unadjust- Figs. 1 This point is further explored in 3 . Fig. 1 lists the resulting elasticity estimates in increasing order Table 1 ed gains from trade to the corresponding own trade shares and shows together with the SITC-Rev3 code and an abbreviated description of Fig. 2 does the same for the adjusted that the correlation is very tight. the corresponding industry. As can be seen, they range from 1.54 to trade shares and it is clear that variation in the exponent now plays a 25.05 and have a mean of 3.63 which is within the range of other es- timates in the literature. also reports the associated 95% con- Table 1 fi dence intervals which I obtained by bootstrapping with 1000 Table 4 repetitions per industry. When resampling, I always clustered by ex- Con dence intervals. fi porter and importer to ensure that it is conducted separately for each Unadjusted Adjusted importer pair. As can be seen, the con – dence intervals vary fi exporter True gain (%) 95% CI True gain (%) 95% CI widely by industry and are quite large on average. In particular, the average lower bound is 2.32 and the average upper bound is 8.16 105.1 159.3 92.9 142.3 148.8 United Arab Emirates 133.2 Argentina 26.5 34.2 23.8 31.3 31.5 28.3 suggesting that it might be important to account for estimation error 26.8 39.1 28.7 35.9 Australia 21.9 30.7 σ in when assessing the reliability of estimates of the gains from s 103.4 Austria 81.4 105.2 88.2 117.0 95.5 12 trade. Belgium 259.9 207.9 318.6 505.2 387.0 622.1 Brazil 9.8 8.7 10.4 9.5 8.3 9.8 Canada 36.6 48.7 44.3 60.3 44.0 53.6 5. Results 86.5 127.6 134.6 Switzerland 105.6 156.8 111.0 76.4 138.4 47.5 72.2 109.0 67.0 Chile 11.4 14.7 30.8 26.8 35.2 China 12.9 Table 2 summarizes the changes in real income resulting from a 25.1 32.3 29.2 30.8 23.6 30.3 Colombia – Unad- “ move from autarky to year 2007 levels of trade. Columns 1 3( Czech Republic 71.4 63.3 80.8 137.4 119.6 159.0 justed ” ) focus on the special case without non-traded and intermediate 35.9 43.7 45.7 Germany 41.7 51.1 40.2 ) adjust for these effects. Recall – 6( “ Adjusted ” goods while columns 4 60.6 75.2 65.1 81.2 75.4 79.2 Denmark Spain 43.7 63.1 42.3 62.1 53.4 52.0 that the special case without non-traded and intermediate goods in- Finland 53.2 77.0 41.7 61.1 68.0 52.6 volves dropping the residual non-traded industry as well as setting 29.8 38.7 32.8 43.6 35.3 39.2 France β Naive gain “ j = 1 for all ” are computed s . The results under and js 44.7 36.6 50.0 31.8 United Kingdom 26.0 34.2 using the aggregate formulas, the entries under ” True gain “ are comput- Greece 72.6 54.8 83.2 121.9 91.6 157.8 137.8 193.8 74.5 101.4 166.1 86.5 Hungary are ed using the industry-level formulas, and the entries under ” Ratio “ Indonesia 25.2 20.1 26.7 35.6 28.1 37.6 simply the ratio of the two. When using the aggregate formulas, I India 13.7 12.5 18.6 20.9 18.8 26.6 σ = 3.94 which is thetrade-weighted cross-industry average work with 99.2 Ireland 80.2 106.3 134.5 102.2 142.7 σ of all . When allowing for non-traded and intermediate goods, I fur- s Iran, Islamic Rep. 28.5 41.5 50.0 24.0 28.7 50.3 β ther construct aggregate by calculating the economy-wide share of Israel 115.0 93.0 139.1 77.5 61.3 90.3 j Italy 24.7 37.8 38.1 28.6 42.9 32.7 value added in gross production. Japan 25.7 17.7 29.0 21.2 35.3 21.4 As can be seen, allowing for cross-industry heterogeneity in the Korea, Republic of 53.1 88.7 36.1 59.2 65.4 42.7 trade elasticities substantially increases the estimated gains from trade 40.3 48.4 33.9 Mexico 45.0 30.4 36.1 for all countries in the sample. While the unadjusted median ” naive “ Malaysia 58.0 92.2 219.0 154.3 293.2 74.1 55.1 70.9 Nigeria 52.6 41.2 53.5 70.9 ” gains are actually gains are only 16.5%, the unadjusted median “ true Netherlands 79.8 70.9 92.3 52.1 45.2 56.1 48.6% so that accounting for cross-industry heterogeneity multiplies 40.3 52.3 Norway 63.3 49.9 67.4 51.0 the median gains from trade by a factor of 3.1. Similarly, the adjusted 30.6 New Zealand 24.5 33.8 23.2 32.5 32.3 true gains are only 16.9% while the adjusted median ” naive “ median ” “ 31.4 39.7 61.9 52.6 67.9 Pakistan 36.7 gains are actually 55.9%, representing an increase by a factor of 3.3. Philippines 57.7 103.0 271.8 45.8 72.8 127.8 42.7 52.2 72.0 47.7 Poland 64.1 78.9 cation effect from havingmultiple industries is similar While themagni fi 63.3 89.1 50.3 70.5 75.0 59.6 Portugal ,myesti- Costinot and Rodriguez-Clare (2014) to the one estimated by 44.1 57.0 73.4 Romania 36.3 46.8 70.0 mates of the absolute gains from trade are quite a bit larger than Rest of the world 35.5 32.2 37.4 56.6 49.8 58.4 13 theirs. Russian Federation 25.1 19.1 27.5 34.9 26.5 37.4 true gains from trade into the own trade Table 3 ” “ decomposes the Saudi Arabia 49.6 37.0 52.1 68.1 49.3 69.3 274.1 439.3 Singapore 218.3 175.6 330.9 361.7 . This decomposition helps to (7) share and the exponent from formula Sweden 57.5 52.6 67.1 55.3 49.2 62.0 understand why allowing for non-traded and intermediate goods does 80.4 103.2 47.8 60.5 89.0 51.3 Thailand 37.5 34.9 45.8 31.8 42.0 41.0 Turkey 61.6 101.0 174.3 86.7 Ukraine 121.2 201.9 12 Broda and Weinstein (2006) Recall that the Feenstra (1994) – method assumes that all 16.5 22.0 13.5 United States 19.4 11.5 14.9 dence intervals in fi varieties are substitutes which is why all elasticity estimates and con 21.5 30.5 41.0 27.9 32.0 44.5 Venezuela, RB σ Table 1 imply N 1. s South Africa 30.5 25.8 35.1 42.3 35.4 47.4 13 Costinot and Rodriguez-Clare(2014) Caliendoand Parro use the elasticity estimates of Median 48.6 41.4 52.8 55.9 49.3 62.5 which have a higher variance, a higher mean, and a higher minimum value than (2015) ” “ gains from Note: This table summarizes the 95% con fi dence intervals around the true fi the ones I use. The higher variance explains why they cation effect fi nd a similar magni 3 do not adjust for non-traded or intermediate trade reported in Table 2 . Columns 1 – despite using a higher level of aggregation. The higher mean and higher minimum value 6do. – goods while columns 4 explain why they estimate lower gains from trade.

11 277 R. Ossa / Journal of International Economics 97 (2015) 266 276 – 100 90 80 70 60 50 40 Average share of (log) gains from trade realized in % 30 20 90 20 100 10 80 70 60 50 40 30 Share of individually most important industries liberalized in % Industry contributions to gains from trade. Fig. 4. countries. As can be seen, the 10% most important industries account plots the unadjusted log gains from trade against Fig. 3 larger role. for roughly 90% of the log gains from trade on average. the adjusted log gains from trade and also includes a 45° line for explores the sensitivity of the gains from trade estimates Table 5 ease of comparison. As can be seen, allowing for non-traded and in- to the level of industry aggregation. In particular, it repli- Table 2 from termediate goods tends to lower the gains from trade for richer fi cates rst aggregating all data back to the GTAP level after Table 2 countries but increase the gains from trade for poorer ones. The rea- Table 1 . using trade-weighted averages of the elasticity estimates from son is that richer countries tend to have higher expenditure shares At this level of aggregation, there are only 28 traded industries instead on non-traded industries and are also typically less dependent on 14 of the 251 traded industries used before. – output imports for inputs that feature prominently in their input By construction, the accounts. “ naive ” . However, the gains from trade are the same in Tables 2 and 5 dence intervals for the ” gains true “ fi reports the 95% con Table 4 ,asone than in gains from trade are lower in “ true Table 5 Table 2 ” . These con Table 2 reported in fi dence intervals are constructed by would expect given the higher level of aggregation.For example, the ad- re-calculating the gains from trade for each of the 1000 sets of gains fall from 55.9% to 35.2% when the analysis is ” true “ justed median bootstrapped elasticity estimates. Despite the considerable noise in conducted at the GTAP level instead of the 3-digit level. “ dence intervals around the fi true the elasticity estimates, the con ” gains from trade are actually tighter than one might have thought. In 6. Conclusion dence intervals of the fi particular, the median lower bound of the con unadjusted ” “ gains from trade is 41.4% while the median upper true In this paper, I argued that accounting for cross-industry variation in bound of these gains is 52.8%. Similarly, the median lower bound of trade elasticities greatly magni fi es the estimated gains from trade. The fi dence intervals of the adjusted “ true gains from trade is ” the con main idea was that a complete shutdown of international trade is very 49.3% while the median upper bound of these gains is 62.5%. This hap- costly even though imports in the average industry do not matter too pens because most of the variation in the bootstrapped elasticity esti- much since imports in some industries are critical to the functioning mates is in the right tail which is exactly where the gains from trade of the economy. While I have made this point in the context of a simple do not respond to elasticity changes that much. Armington (1969) model, it should be clear that it extends to other “ gains ” Fig. 4 illustrates that a large share of the adjusted true Eaton and Kortum commonly used quantitative trade models. In an from trade can be attributed to a small share of critical industries. (2002) model, for example, the interpretation would be that interna- tional productivity differences are so large in some industries that re- b w j gure based on the relationship ln fi I construct this − ¼ P j ciently produced fi placing ef fi ciently produced imports with inef 1 domestic substitutes in these industries would imply extreme costs. S S s α ∑ δ ð ln ÞÞ which follows immediately from λ ∑ ð js jt t 1 s ¼ 1 ¼ jt σ − 1 t the above formulas for the gains from trade. First, I rank all 14 The original GTAP data actually features 42 traded industries. I aggregate them into 28 by their contribution to the overall log gains from t industries , , ” traded industries by combining cereal grains nec “ , paddy rice ” wheat “ ” vegetables, “ “ 1 oil seeds “ , prod- fruits, nuts into ” processed rice “ ,and ” crops nec “ , ” bres fi plant-based “ “ , ” ” S s α ð λ δ Þ for each country. Then, I compute ln trade −∑ js jt s ¼ 1 jt ucts of agriculture, etc “ , , ” animal products nec ” bovine cattle, sheep and goats, horses “ ” , σ 1 − t “ wool,silk-worm cocoons “ and ” dairy products “ and ” raw milk in- , ” liveanimals,etc “ into ” the shares of the log gains from trade due to shares of most important ” vegetable “ ” meat products nec “ , ,and bovine meat products “ , ” milkand dairy products “ to 1 S s ,and ” sugar food products “ , ” ,and “ sugar cane, sugar beet “ ” meat,oil,etc “ oils and fats ” into α δ for each ð ln λ Þ −∑ industries by cumulating over js jt s ¼ 1 jt σ 1 − nec ” food products nec ” . This is necessary to ensure that each SITC-Rev3 3-digit sec- into “ t tor uniquely maps into one GTAP sector. country. Finally, I take the simple average of these shares across

12 277 – R. Ossa / Journal of International Economics 97 (2015) 266 277 Table 5 References Gains from trade with GTAP instead of 3-digit industry aggregation. Arkolakis, K., Costinot, A., Rodriguez-Clare, A., 2012. New trade models, same old gains? Unadjusted Adjusted 130. – Am. Econ. Rev. 102 (1), 94 Armington, P., 1969. A theory of demand for products distinguished by place of produc- Ratio True gain Ratio Naive gain True gain Naive gain tion. IMF Staff. Pap. 16, 159 – 176. (%) (%) (%) (%) Globalization and the gains from variety. Q. J. Econ. 121 (2), Broda, C., Weinstein, D., 2006. 585. – 541 1.9 1.5 35.9 United Arab 68.8 58.8 39.8 Caliendo, L., Parro, F., 2015. Estimates of the trade and welfare effects of NAFTA. Rev. Econ. Emirates – Stud. 82 (1), 1 44. Argentina 14.2 1.5 9.6 16.6 1.7 9.2 Trade theory with numbers: quantifying the con- Costinot, A., Rodriguez-Clare, A., 2014. 13.1 Australia 1.4 9.7 15.5 1.6 17.7 sequences of globalization. In: Gopinath, G., Helpman, E., Rogoff, K. (Eds.), Handbook 48.7 32.1 Austria 1.8 48.2 1.5 27.1 of International Economics. Belgium 53.3 132.3 2.5 59.5 259.6 4.4 Eaton, J., Kortum, S., 2002. Technology, geography, and trade. Econometrica 70 (5), 6.3 Brazil 4.7 1.3 1.3 6.4 4.9 1741 – 1779. 21.9 1.5 24.8 Canada 1.3 14.4 19.0 Edmond, C., Midrigan, V., Xu, D., 2012. Competition, Markups, and the Gains from 55.6 Switzerland 39.0 72.7 1.9 24.1 2.3 International Trade. NBER Working Paper 18041. 17.0 31.9 Chile 3.7 60.0 1.9 16.0 Feenstra, Robert C., 1994. New product varieties and the measurement of international 5.7 7.7 1.3 13.8 17.8 1.3 China prices. Am. Econ. Rev. 84 (1), 157 – 177. Colombia 15.9 9.5 15.7 2.1 1.7 7.6 Feenstra, R., 2010. Product Variety and the Gains from International Trade. MIT Press, Cambridge, MA. 80.5 1.9 38.0 42.8 22.6 Czech Republic 2.1 Hsieh, C., Ossa, R., 2012. A Global View of Productivity Growth in China. NBER Working 1.4 25.7 1.6 17.7 28.7 18.5 Germany Paper 16778. 40.3 Denmark 1.5 25.4 1.7 42.1 26.5 Imbs, J., Mejean, I., 2015. – 83. Elasticity optimism. Am. Econ. J. Macroecon. 7 (3), 43 2.5 38.5 2.1 15.4 35.0 16.4 Spain Intermediate goods and weak links in the theory of economic develop- Jones, C., 2011. Finland 1.9 22.0 33.4 17.2 42.4 1.9 ment. Am. Econ. J. Macroecon. 3 (2), 1 – 28. 1.9 1.7 13.1 24.2 France 15.0 26.0 Scale economies, product differentiation, and the pattern of trade. Am. Krugman, P., 1980. United Kingdom 18.3 24.3 1.3 12.6 19.2 1.5 959. – Econ.Rev.70(5),950 40.4 85.7 1.9 19.4 4.4 Greece 20.8 Melitz, M., 2003. The impact of trade on intra-industry reallocations and aggregate indus- 1.8 45.4 Hungary 88.8 47.9 2.0 26.0 try productivity. Econometrica 71 (6), 1695 – 1725. 8.3 1.7 18.7 Indonesia 12.8 1.5 11.3 Monetary non-neutrality in a multisector menu cost Nakamura, E., Steinsson, J., 2010. 1.5 India 7.3 11.4 1.6 11.2 17.4 – model. Q. J. Econ. 125 (3), 961 1013. 70.7 31.7 51.3 1.6 41.9 1.7 Ireland Narayanan, B., Aguiar, A., McDougall, R. (Eds.), 2012. Global Trade, Assistance, and 8.9 31.1 1.8 11.7 Iran, Islamic Rep. 2.7 15.9 Production: The GTAP 8 Data Base. Center for Global Trade Analysis, Purdue University. 29.4 Israel 2.0 21.7 41.1 1.9 59.0 4146. Trade wars and trade talks with data. Am. Econ. Rev. 104 (2), 4104 Ossa, R., 2014. – 27.1 2.0 23.1 11.1 Italy 2.1 13.6 19.5 7.1 3.0 2.7 23.2 7.8 Japan Korea, Republic of 12.3 35.4 2.9 21.3 52.4 2.5 23.4 18.1 1.6 15.0 1.6 11.3 Mexico Malaysia 22.8 31.7 1.4 46.8 64.5 1.4 2.8 37.1 2.3 13.2 24.2 10.5 Nigeria 1.6 Netherlands 26.2 48.7 1.9 18.8 30.6 1.8 28.8 Norway 19.7 1.5 14.9 26.8 New Zealand 1.6 1.4 11.5 16.1 11.7 18.1 3.6 9.5 Pakistan 45.5 2.3 12.8 22.0 Philippines 18.5 28.5 1.5 23.0 97.0 4.2 44.1 2.1 1.7 21.1 28.0 16.6 Poland 50.4 2.6 Portugal 18.8 37.9 2.0 19.1 37.3 1.8 22.8 15.3 Romania 1.5 20.5 Rest of the world 16.3 1.7 36.2 1.4 21.9 23.0 Russian Federation 9.1 12.3 17.8 1.6 1.4 10.8 33.9 1.6 Saudi Arabia 14.9 23.6 1.6 21.1 134.4 1.8 113.1 57.2 Singapore 2.0 73.1 Sweden 1.6 1.6 21.2 35.1 21.4 34.3 1.4 19.1 Thailand 49.1 1.7 35.5 31.8 Turkey 12.6 24.9 2.0 12.3 28.6 2.3 101.8 3.2 2.4 31.4 53.4 22.3 Ukraine 6.4 8.9 1.4 United States 9.9 12.2 1.2 9.2 2.3 1.6 Venezuela, RB 8.4 13.2 20.7 24.2 1.7 South Africa 11.2 17.7 1.6 14.6 35.2 27.0 16.5 1.8 Median 1.6 16.9 Note: This table summarizes the changes in real income resulting from a move from autar- ky to year 2007 levels of trade using a 2-digit instead of a 3-digit industry aggregation. The are computed using the industry-level formulas, the results True gain “ results under ” are computed using the aggregate formulas, and the results under under “ Naive gain ” – Ratio “ ” simply compute the ratio of the two. Columns 1 3 do not adjust for non-traded fi 6 do. I include Hong Kong in my de – or intermediate goods while columns 4 nition of China. The GTAP aggregation features 28 traded and one non-traded industry while the earlier 3-digit aggregation features 251 traded and one non-traded industry.

Related documents

PB

PB

OFFICIAL 2019 CONNECTICUT PRACTICE BOOK (Revision of 1998) CONTAINING RULES OF PROFESSIONAL CONDUCT CODE OF JUDICIAL CONDUCT RULES FOR THE SUPERIOR COURT RULES OF APPELLATE PROCEDURE APPENDIX OF FORMS...

More info »
June2018CUR

June2018CUR

CHANCELLOR'S UNIVERSITY REPORT JUNE 25 2018

More info »
TLA Referral Dir

TLA Referral Dir

2019–2020 REFERRAL DIRECTORY Legal Services and Other Resources for Low-Income Texans Published by the LEGAL ACCESS DIVISION OF THE STATE BAR OF TEXAS . Austin, Texas 78711-2487 P.O. Box 12487 . Fax: ...

More info »
TLA Referral Dir

TLA Referral Dir

2019–2020 REFERRAL DIRECTORY Legal Services and Other Resources for Low-Income Texans Published by the LEGAL ACCESS DIVISION OF THE STATE BAR OF TEXAS . Austin, Texas 78711-2487 P.O. Box 12487 . Fax: ...

More info »
DEC CUR

DEC CUR

CHANCELLOR'S UNIVERSITY REPORT DECEMBER 10, 2018

More info »
Building Soils For Better Crops

Building Soils For Better Crops

third EditiON Building SoilS FoR Building SoilS BetteR CRopS SuS tainable Soil ManageM ent SuStainable Soil ManageMent third EditiON FoR Bet teR CRopS “...it is our work with living soil that provides...

More info »
Department of Defense   Law of War Manual (June 2015)

Department of Defense Law of War Manual (June 2015)

D E A R T M E N T O F D E F E N S E P N A L O F W A R M A W U A L J U N E 2 0 1 5 O F F I C E O F G E N ER A L C O U N S E L D P A R T M E N T E O F D E F E N S E

More info »
Implementation Handbook For The Convention On The Rights Of The Child

Implementation Handbook For The Convention On The Rights Of The Child

IMPLEMENTATION HANDBOOK FOR THE CONVENTION ON THE RIGHTS OF THE CHILD FULLY REVISED THIRD EDITION IMPLEMENTATION HANDBOOK IMPLEMENTATION HANDBOOK FOR THE CONVENTION ON THE FOR THE CONVENTION ON THE RI...

More info »
a i5199e

a i5199e

Status of the Main Report World’s Soil Resources © FAO | Giuseppe Bizzarri INTERGOVERNMENTAL INTERGOVERNMENTAL TECHNICAL PANEL ON SOILS TECHNICAL PANEL ON SOILS

More info »
C:\Users\aholmes4\AppData\Roaming\SoftQuad\XMetaL\7.0\gen\c\H5515 ~1.XML

C:\Users\aholmes4\AppData\Roaming\SoftQuad\XMetaL\7.0\gen\c\H5515 ~1.XML

H. R. 5515 One Hundred Fifteenth Congress of the United States of America AT THE SECOND SESSION Begun and held at the City of Washington on Wednesday, the third day of January, two thousand and eighte...

More info »
435 441 458 467r e

435 441 458 467r e

WT/DS435/R, WT/DS441/R WT/DS458/R, WT/DS467/R 28 June 2018 Page: (18 - 1/884 4061 ) Original: English AUSTRALIA CERTAIN MEASURES CON CERNING TRADEMARKS, – PACKAGING IONS AND OTHER PLAIN GEOGRAPHICAL I...

More info »
Microsoft Word   Teacher Shortage Areas Report 2017 18.docx

Microsoft Word Teacher Shortage Areas Report 2017 18.docx

Teacher Shortage Areas Nationwide Listing 1990–1991 through 2017–2018 June 2017 U.S. Department of Education Office of Postsecondary Education Freddie Cross Senior Statistician U.S. Dept. of Education...

More info »
PCT INTERNATIONAL SEARCH AND PRELIMINARY EXAMINATION GUIDELINES

PCT INTERNATIONAL SEARCH AND PRELIMINARY EXAMINATION GUIDELINES

E PCT/GL/ISPE/ 8 ENGLISH ORIGINAL: DATE: December 19, 2018 PATENT COOPERATION TR E AT Y (PCT) PCT INTERNATIONAL SEARCH AND PRELIMINARY EXAMINAT ION GUIDELINES (Guidelines for the Processing by Interna...

More info »
title17

title17

c i r c u l a r 9 2 Copyright Law of the United States and Related Laws Contained in Tıtle 17 of the United States Code december 2016 t

More info »
Quantitative Literacy: Why Numeracy Matters for Schools and Colleges

Quantitative Literacy: Why Numeracy Matters for Schools and Colleges

NATIONAL COUNCIL ON EDUCATION AND THE DISCIPLINES The goal of the National Council on Education and the Disciplines (NCED) is to advance a vision that will unify and guide efforts to strengthen K-16 e...

More info »
towards the other america ebook

towards the other america ebook

h To w a r d s t e Other America ” “ cist ResouRces foR White PeoPle Anti-RA ck lives MA tAking Action foR BlA tteR Chris Crass “A clarion call for conscious whites who seek genuine solidarity with th...

More info »
AndersBehringBreivikManifesto

AndersBehringBreivikManifesto

2011 , London – By Andrew Berwick

More info »
meinkampf

meinkampf

Mein Kampf Adolf Hitler Translated into English by James Murphy

More info »