Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Nuevos Desarrollos en Teoría del Comercio Internacional y su Análisis Empírico Timothy J. Kehoe University of Minnesota y Federal Reserve Bank of Minneapolis Universitat de Barcelona Mayo 2006 www.econ.umn.edu/~tkehoe/ub-06.html Outline: 1. Standard theory (hybrid Heckscher-Ohlin/New Trade Theory) does not well when matched with the data on the growth and composition of trade. 2. Applied general equilibrium models that put the standard theory to work do not well in predicting the impact of trade liberalization experiences like NAFTA. 3. Much of the growth of trade after a trade liberalization experience is growth on the extensive margin. Models need to allow for corner solutions or fixed costs. 4. Modeling the fixed costs may explain why real exchange rate data indicate that more arbitrage across countries that have a strong bilateral trade relationship. 5. The major determinant of large macroeconomic fluctuations in Mexico is fluctuations in total factor productivity. If trade liberalization and the 1995 financial crisis have had large macroeconomic effects, it is through their effects on TFP. 1. Standard theory (hybrid Heckscher-Ohlin/New Trade Theory) does not well when matched with the data on the growth and composition of trade. In the 1980s and 1990s trade economists reached a consensus that North-North trade — trade among rich countries — was driven by forces captured by the New Trade Theory and North-South trade — trade between rich countries and poor countries — was driven by forces captured by Heckscher-Ohlin theory. (South-South trade was negligible.) A. V. Deardorff, “Testing Trade Theories and Predicting Trade Flows,” in R. W. Jones and P. B. Kenen, editors, Handbook of International Economics, volume l, North-Holland, 1984, 467-517. J. Markusen, “Explaining the Volume of Trade: An Eclectic Approach,” American Economic Review, 76 (1986), 1002-1011. In fact, a calibrated version of this hybrid model does not match the data. R. Bergoeing and T. J. Kehoe, “Trade Theory and Trade Facts,” Federal Reserve Bank of Minneapolis, 2003. TRADE THEORY Traditional trade theory — Ricardo, Heckscher-Ohlin — says countries trade because they are different. In 1990 by far the largest bilateral trade relation in the world was U.S.-Canada. The largest two-digit SITC export of the United States to Canada was 78 Road Vehicles. The largest two-digit SITC export of Canada to the United States was 78 Road Vehicles. The New Trade Theory — increasing returns, taste for variety, monopolistic competition — explains how similar countries can engage in a lot of intraindustry trade. Helpman and Krugman (1985) Markusen (1986) TRADE THEORY AND TRADE FACTS x Some recent trade facts x A “New Trade Theory” model x Accounting for the facts x Intermediate goods? x Policy? How important is the quantitative failure of the New Trade Theory? Where should trade theory and applications go from here? SOME RECENT TRADE FACTS x The ratio of trade to product has increased. World trade/world GDP increased by 59.3 percent 1961-1990. OECD-OECD trade/OECD GDP increased by 111.5 percent 1961-1990. x Trade has become more concentrated among industrialized countries OECD-OECD trade/OECD-RW trade increased by 87.1 percent 1961-1990. x Trade among industrialized countries is mostly intraindustry trade Grubel-Lloyd index for OECD-OECD trade in 1990 is 68.4. Grubel-Lloyd index for OECD-RW trade in 1990 is 38.1. OECD-OECD Trade / OECD GDP 0.12 0.10 0.08 0.06 0.04 0.02 0.00 1960 1963 1966 1969 1972 1975 year 1978 1981 1984 1987 1990 OECD-OECD Trade / OECD-RW Trade 1.80 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1960 1963 1966 1969 1972 1975 year 1978 1981 1984 1987 1990 Helpman and Krugman (1985): “These....empirical weaknesses of conventional trade theory...become understandable once economies of scale and imperfect competition are introduced into our analysis.” Markusen, Melvin, Kaempfer, and Maskus (1995): “Thus, nonhomogeneous demand leads to a decrease in NorthSouth trade and to an increase in intraindustry trade among the northern industrialized countries. These are the stylized facts that were to be explained.” Goal: To measure how much of the increase in the ratio of trade to output in the OECD and of the concentration of world trade among OECD countries can be accounted for by the “New Trade Theory.” PUNCHLINE In a calibrated general equilibrium model, the New Trade Theory cannot account for the increase in the ratio of trade to output in the OECD. Back-of-the-envelope calculations: Suppose that the world consists of the OECD and the only trade is manufactures. With Dixit-Stiglitz preferences, country j exports all of its production of manufactures Ymj except for the fraction s j Y j / Y oe that it retains for domestic consumption. World imports: M World trade/GDP: M Y oe World trade/GDP: n ¦j M Ymoe Ymoe Y oe j j 1 s Y m. 1 1 ¦ n j oe Y j 2 m s ( ) . oe 1 Y M Y oe n M Ymoe Ymoe Y oe 1 ¦ n j oe Y j 2 m s ( ) . oe 1 Y 1 ¦ j 1 ( s j ) 2 goes from 0.663 in 1961 to 0.827 in 1990. Ymoe / Y oe goes from 0.295 in 1961 to 0.222 in 1990. 0.663 u 0.295 0.196 | 0.184 0.827 u 0.222 . Effects cancel! A “NEW TRADE THEORY” MODEL Environment: x Static: endowments of factors are exogenous x 2 regions: OECD and rest of world x 2 traded goods: homogeneous — primaries (CRS) and differentiated — manufactures (IRS) x 1 nontraded good — services (CRS) x 2 factors: (effective) labor and capital x Identical technologies and preferences (love for variety) across regions x Primaries are inferior to manufactures We only consider merchandise trade in both the data and in the model. Key Features of the Model Consumers' problem: max s.t. E p (c pj J p )K E m ( ³ cmj ( z ) U dz p )K / U E s (csj J s )K 1 Dw K q p c pj ³ w qm ( z )cmj ( z )dz p qsj csj d r j k j w j h j . D Firms' problems Primaries and Services: Standard CRS problems. Ypj T p ( K pj )D p ( H pj )1D p Ys j Ts ( K sj )D s ( H sj )1D s Manufactures: Standard (Dixit-Stiglitz) monopolistically competitive problem: x Fixed cost. Ym ( z ) max ª¬Tm K m ( z )Dm H m ( z )1Dm F ,0 º¼ x Firm z sets its price qm ( z ) to max profits given all of the other prices. n Ym ( z ) ¦ j 1 Cmj ( z ) Cmrw ( z ) . 1 1K Cmj ( z ) Em (r j K j w j H j q p J p N j qsj J s N j ) 1 1U U 1U qm ( z ) ª ³ D w qm ( z c) dz cº ¬ ¼ UK U (1K ) (1U ) U K 1K ' U K 1 ª º 1U 1K 1K ' Em « ³ w qm ( z c) dz c » E s qs D ¬ ¼ x Every firm is uniquely associated with only one variety (symmetry). x Free entry. [0, d w ] with d w finite and endogenously determined. x Dw K 1K E p qp 1 1K 1 1K Volume of Trade Let s j be the share of country j , j production of manufactures, sj ³D j 1,..., n, rw , in the world Ym ( z )dz / ³D w Ym ( z )dz Ymj / Ymw . The imports by country j from the OECD are M oej (1 s rw s j )Cmj M oerw (1 s rw )Cmrw . Total imports in the OECD from the other OECD countries are n M oe oe n j 2 rw oe M (1 s ( s ) /(1 s )) C ¦ ¦ m . j oe j 1 rw j 1 OECD in 1990 Country Australia Austria Belgium-Lux Canada Denmark Finland France Germany Greece Iceland Ireland Italy Share of GDP % 1.79 0.97 1.26 3.45 0.78 0.81 7.26 9.96 0.50 0.04 0.28 6.64 Country Japan Netherlands New Zealand Norway Portugal Spain Sweden Switzerland Turkey United Kingdom United States Share of GDP % 18.04 1.72 0.26 0.70 0.41 3.00 1.40 0.17 0.91 5.92 33.72 ACCOUNTING FOR THE FACTS Compare the changes that the model predicts for 1961-1990 with what actually took place. Focus on key variables: OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manfacturing GDP/OECD GDP Calibrate to 1990 data. Backcast to 1961 by imposing changes in parameters: relative sizes of countries in the OECD populations sectoral productivities endowments ACCOUNTING FOR THE FACTS Benchmark 1990 OECD Data Set (Billion U.S. dollars) H ioe K ioe Yi oe Cioe Yi oe Cioe Primaries Manufactures 228 2,884 441 775 669 3,659 862 3,466 -193 193 Services 8,644 3,497 12,141 12,141 0 Total 11,756 4,713 16,469 16,469 0 ACCOUNTING FOR THE FACTS Benchmark 1990 Rest of the World Data Set (Billion U.S. dollars) Yi rw Cirw Yi rw Cirw Primaries Manufactures 1,223 1,159 1,030 1,352 193 -193 Services 3,447 3,447 0 Total 5,829 5,829 0 ACCOUNTING FOR THE FACTS x N oe x ¦ i 854 , N rw rw Y p ,m , s i x Set q p qm ( z ) values). xU 4, 428 . ¦ i rw C p ,m , s i qs w 5,829 . r 1 (quantities are 1990 1/1.2 (Morrison 1990, Martins, Scarpetta, and Pilat 1996). x Normalize d w 100 . x Calibrate H rw , K rw so that benchmark data set is an equilibrium. x Alternative calibrations of utility parameters J p , J s , and K . OECD in 1961 Country Austria Belgium-Lux Canada Denmark France Germany Greece Iceland Ireland Italy Share of GDP % 0.75 1.25 4.22 0.70 6.99 9.71 0.50 0.03 0.21 4.64 Country Share of GDP % Netherlands 1.37 Norway 0.60 Portugal 0.32 Spain 1.38 Sweden 1.62 Switzerland 1.07 Turkey 0.83 United Kingdom 8.08 United States 55.74 Numerical Experiments Calculate equilibrium in 1961: T p ,1961 T m ,1961 T s ,1961 N oe T p ,1990 T m ,1990 /1.01429 , F1961 F1990 /1.01429 T s ,1990 /1.00529 (Echevarria 1997) 536, N rw 2,545 Numerical Experiments oe oe rw rw Choose H1961 , K1961 , H1961 , K1961 so that ¦ ¦ ¦ ¦ i oe oe Y / N 1990 p ,m , s i ,1990 oe i p ,m , s i ,1961 Y i oe 1961 /N rw rw Y / N 1990 p ,m , s i ,1990 rw i p ,m , s i ,1961 Y rw 1961 /N oe oe K1961 K1990 oe oe H1961 H1990 q p ,1961 (Yprw,1961 C prw,1961 ) ¦ i p ,m , s rw i ,1961 i ,1961 q Y 2.400 2.055 0.050 How Can the Model Work in Matching the Facts? x The ratio of trade to product has increased: The size distribution of countries has become more equal (Helpman-Krugman). x Trade has become more concentrated among industrialized countries: OECD countries have comparative advantage in manufactures, while the RW has comparative advantage in primaries. Because they are inferior to manufactures, primaries become less important in trade as the world becomes richer (Markusen). How Can the Model Work in Matching the Facts? x Trade among industrialized countries is largely intraindustry trade: OECD countries export manufactures. Because of taste for variety, every country consumes some manufactures from every other country (Dixit-Stiglitz). x The different total factor productivity growth rates across sectors imply that the price of manufactures relative to primaries and services has fallen sharply between 1961 and 1990. If price elasticities of demand are not equal to one, a lot can happen. Experiment 1 Jp Jp K 0 Data OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 1. Jp = 0, Js = 0, K = 0 OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 1961 1990 Change 0.053 0.844 0.295 0.112 1.579 0.222 111.5% 87.1% 24.6% 0.108 0.893 0.223 0.136 1.169 0.222 25.8% 30.9% 0.4% Experiment 2 Jp = 169.5, Js = 314.7 to match consumption in RW in 1990, K=0 1961 Data OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 2. Jp = 169.5, Js = 314.7, K = 0 OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 1990 Change 0.053 0.844 0.295 0.112 111.5% 1.579 87.1% 0.222 24.6% 0.103 0.739 0.225 0.132 1.060 0.222 28.1% 43.6% 1.4% Experiment 3 Jp = 169.5, Js = 314.7, K = 0.559 to match growth in OECD-OECD Trade/OECD GDP 1961 Data OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 3. Jp = 169.5, Js = 314.7, K = 0.559 OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 1990 Change 0.053 0.844 0.295 0.112 111.5% 1.579 87.1% 0.222 24.6% 0.063 0.738 0.137 0.132 111.5% 1.060 43.7 % 0.222 62.7% Experiments 4 and 5 Jp = 169.5, Js = 314.7, reasonable values of K ( 0.5 t 1/(1 K ) t 0.1) 1961 1990 Change Data OECD-OECD Trade/OECD GDP 0.053 0.112 111.5% OECD-OECD Trade/OECD-RW Trade 0.844 1.579 87.1% OECD Manf GDP/OECD GDP 0.295 0.222 24.6% 4. Jp = 169.5, Js = 314.7, K = 1 OECD-OECD Trade/OECD GDP 0.118 0.132 11.7% OECD-OECD Trade/OECD-RW Trade 0.739 1.060 43.5% OECD Manf GDP/OECD GDP 0.259 0.222 14.1% 5. Jp = 169.5, Js = 314.7, K = 9 OECD-OECD Trade/OECD GDP 0.118 0.132 1.6% OECD-OECD Trade/OECD-RW Trade 0.739 1.060 43.5% OECD Manf GDP/OECD GDP 0.284 0.222 21.8% Sensitivity Analysis: Alternative Calibration Methodologies x Alternative specifications of nonhomogeneity x Gross imports calibration x Alternative RW endowment calibration x Alternative RW growth calibration x Intermediate goods INTERMEDIATE GOODS? Ypj j j ª X j ³ w X mp º ( z ) dz X D 1 D p p pp sp » min « , D , , T p K pj H pj amp asp «¬ a pp »¼ Ym ( z ) Ys j j ª X j ( z ) ³ w X mm ( z , z ')dz ' X j ( z ) º pm « , D , sm , » amm asm » min « a pm « » Dm 1D m F «¬Tm K m ( z ) H m ( z ) »¼ j ª X j ³ w X ms º j ( z ) dz D 1 D X s s ps min « , D , ss , Ts K sj H sj » ams ass «¬ a ps »¼ Results for Model with Intermediate Goods 1961 Data OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 4. Jp = 307.8, Js = 262.2, K = 1 OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 5. Jp = 307.8, Js = 262.2, K = 9 OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 1990 Change 0.053 0.112 0.844 1.579 0.295 0.222 111.5% 87.1% 24.6% 0.323 0.370 0.994 1.305 0.263 0.222 14.5% 31.3% 15.6% 0.337 0.370 0.933 1.305 0.307 0.222 9.7% 39.9% 27.5% POLICY? In a version of our model with n OECD countries, a manufacturing sector, and a uniform ad valorem tariff W, the ratio of exports to income is given by M Y (n 1)C f Y n 1 n 1 (1 W )1/(1 U ) Fixing n to replicate the size distribution of national incomes in the OECD, and setting U 1/1.2 , a fall in W from 0.45 to 0.05 produces an increase in the ratio of trade to output as seen in the data. World Trade / World GDP 0.90 0.80 0.70 rho=1/2.0 0.60 0.50 0.40 rho=1/1.5 0.30 rho=1/1.1 0.20 rho=1/1.2 0.10 0.00 0 0.1 0.2 0.3 0.4 0.5 tau 0.6 0.7 0.8 0.9 1 2. Applied general equilibrium models that put the standard theory to work do not well in predicting the impact of trade liberalization experiences like NAFTA. Applied general equilibrium models were the only analytical game in town when it came to analyzing the impact of NAFTA in 1992-1993. Typical sort of model: Static applied general equilibrium model with large number of industries and imperfect competition (Dixit-Stiglitz or Eastman-Stykolt) and finite number of firms in some industries. In some numerical experiments, new capital is placed in Mexico owned by consumers in the rest of North America to account for capital flows. Examples: Brown-Deardorff-Stern model of Canada, Mexico, and the United States Cox-Harris model of Canada Sobarzo model of Mexico T. J. Kehoe, “An Evaluation of the Performance of Applied General Equilibrium Models of the Impact of NAFTA,” in T. J. Kehoe, T. N. Srinivasan, and J. Whalley, editors, Frontiers in Applied General Equilibrium Modeling: Essays in Honor of Herbert Scarf, Cambridge University Press, 2005, 341-77. Research Agenda: x Compare results of numerical experiments of models with data. x Determine what shocks — besides NAFTA policies — were important. x Construct a simple applied general equilibrium model and perform experiments with alternative specifications to determine what was wrong with the 1992-1993 models. Applied GE Models Can Do a Good Job! Spain: Kehoe-Polo-Sancho (1992) evaluation of the performance of the Kehoe-Manresa-Noyola-Polo-Sancho-Serra MEGA model of the Spanish economy: A Shoven-Whalley type model with perfect competition, modified to allow government and trade deficits and unemployment (Kehoe-Serra). Spain’s entry into the European Community in 1986 was accompanied by a fiscal reform that introduced a value-added tax (VAT) on consumption to replace a complex range of indirect taxes, including a turnover tax applied at every stage of the production process. What would happen to tax revenues? Trade reform was of secondary importance. Canada-U.S.: Fox (1999) evaluation of the performance of the Brown-Stern (1989) model of the 1989 Canada-U.S. FTA. Other changes besides policy changes are important! Changes in Consumer Prices in the Spanish Model (Percent) sector food and nonalcoholic beverages tobacco and alcoholic beverages clothing housing household articles medical services transportation recreation other services data model model model 1985-1986 policy only shocks only policy&shocks 1.8 -2.3 4.0 1.7 3.9 2.5 3.1 5.8 2.1 5.6 0.9 6.6 -3.3 -2.2 -2.7 -4.8 0.1 2.2 0.7 2.9 -0.7 -4.8 0.6 -4.2 -4.0 2.6 -8.8 -6.2 -1.4 -1.3 1.5 0.1 2.9 1.1 1.7 2.8 weighted correlation with data variance decomposition of change -0.08 0.30 0.87 0.77 0.94 0.85 regression coefficient a regression coefficient b 0.00 -0.08 0.00 0.54 0.00 0.67 Measures of Accuracy of Model Results 1. Weighted correlation coefficient. 2. Variance decomposition of the (weighted) variance of the changes in the data: vardec( y data , y model ) var ( y model ) . model data model var ( y ) var ( y y ) 3, 4. Estimated coefficients a and b from the (weighted) regression xidata a bximodel ei . Changes in Value of Gross Output/GDP in the Spanish Model (Percent) sector agriculture energy basic industry machinery automobile industry food products other manufacturing construction commerce transportation services government services data 1985-1986 -0.4 -20.3 -9.0 3.7 1.1 -1.8 0.5 5.7 6.6 -18.4 8.7 7.6 weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b model policy only -1.1 -3.5 1.6 3.8 3.9 -2.4 -1.7 8.5 -3.6 -1.5 -1.1 3.4 model model shocks only policy&shocks 8.3 6.9 -29.4 -32.0 -1.8 -0.1 1.0 5.0 4.7 8.6 4.7 2.1 2.3 0.5 1.4 10.3 4.4 0.4 1.0 -0.7 5.8 4.5 0.9 4.3 0.16 0.11 0.80 0.73 0.77 0.71 -0.52 0.44 -0.52 0.75 -0.52 0.67 Changes in Trade/GDP in the Spanish Model (Percent) direction of exports Spain to rest of E.C. Spain to rest of world rest of E.C. to Spain rest of world to Spain data model model model 1985-1986 policy only shocks only policy&shocks -6.7 -3.2 -4.9 -7.8 -33.2 -3.6 -6.1 -9.3 14.7 4.4 -3.9 0.6 -34.1 -1.8 -16.8 -17.7 weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b 0.69 0.02 0.77 0.17 0.90 0.24 -12.46 5.33 2.06 2.21 5.68 2.37 Changes in Composition of GDP in the Spanish Model (Percent of GDP) variable wages and salaries business income net indirect taxes and tariffs data model 1985-1986 policy only -0.53 -0.87 -1.27 -1.63 1.80 2.50 model shocks only -0.02 0.45 -0.42 model policy&shocks -0.91 -1.24 2.15 correlation with data variance decomposition of change 0.998 0.93 -0.94 0.04 0.99 0.96 regression coefficient a regression coefficient b private consumption private investment government consumption government investment exports -imports 0.00 0.73 -1.23 1.81 -0.06 -0.06 -0.42 -0.03 0.00 -3.45 -0.51 -0.58 -0.38 -0.07 -0.69 2.23 0.00 0.85 -1.78 1.32 -0.44 -0.13 -1.07 2.10 correlation with data variance decomposition of change 0.40 0.20 0.77 0.35 0.83 0.58 regression coefficient a regression coefficient b 0.00 0.87 0.00 1.49 0.00 1.24 -0.81 1.09 -0.02 -0.06 -3.40 3.20 Public Finances in the Spanish Model (Percent of GDP) data model model model variable 1985-1986 policy only shocks only policy&shocks indirect taxes and subsidies 2.38 3.32 -0.38 2.98 tariffs -0.58 -0.82 -0.04 -0.83 social security payments 0.04 -0.19 -0.03 -0.22 direct taxes and transfers -0.84 -0.66 0.93 0.26 government capital income -0.13 -0.06 0.02 -0.04 correlation with data variance decomposition of change regression coefficient a regression coefficient b 0.99 0.93 -0.70 0.08 0.92 0.86 -0.06 0.74 0.35 -1.82 -0.17 0.80 Models of NAFTA Did Not Do a Good Job! Ex-post evaluations of the performance of applied GE models are essential if policy makers are to have confidence in the results produced by this sort of model. Just as importantly, they help make applied GE analysis a scientific discipline in which there are well-defined puzzles and clear successes and failures for alternative hypotheses. Changes in Trade/GDP in Brown-Deardorff-Stern Model (Percent) variable Canadian exports Canadian imports Mexican exports Mexican imports U.S. exports U.S. imports data 1988-1999 52.9 57.7 240.6 50.5 19.1 29.9 weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b model 4.3 4.2 50.8 34.0 2.9 2.3 0.64 0.08 23.20 2.43 Changes in Canadian Exports/ GDP in the Brown-Deardorff-Stern Model (Percent) sector agriculture mining and quarrying food textiles clothing leather products footwear wood products furniture and fixtures paper products printing and publishing chemicals petroleum and products rubber products nonmetal mineral products glass products iron and steel nonferrous metals metal products nonelectrical machinery electrical machinery transportation equipment miscellaneous manufactures weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b exports to Mexico 1988–1999 model 122.5 3.1 -34.0 -0.3 89.3 2.2 268.2 -0.9 1544.3 1.3 443.0 1.4 517.0 3.7 232.6 4.7 3801.7 2.7 240.7 -4.3 6187.4 -2.0 37.1 -7.8 678.1 -8.5 647.4 -1.0 333.5 -1.8 264.4 -2.2 195.2 -15.0 38.4 -64.7 767.0 -10.0 376.8 -8.9 633.9 -26.2 305.8 -4.4 1404.5 -12.1 exports to United States 1988–1999 model 106.1 3.4 75.8 0.4 91.7 8.9 97.8 15.3 237.1 45.3 -14.4 11.3 32.8 28.3 36.5 0.1 282.6 12.5 113.7 -1.8 37.2 -1.6 109.4 -3.1 -42.5 0.5 113.4 9.5 20.5 1.2 74.5 30.4 92.1 12.9 34.7 18.5 102.2 15.2 28.9 3.3 88.6 14.5 30.7 10.7 100.0 -2.1 -0.91 0.003 -0.43 0.02 249.24 -15.48 79.20 -2.80 Changes in Mexican Exports/GDP in the Brown-Deardorff-Stern Model (Percent) sector agriculture mining and quarrying food textiles clothing leather products footwear wood products furniture and fixtures paper products printing and publishing chemicals petroleum and products rubber products nonmetal mineral products glass products iron and steel nonferrous metals metal products nonelectrical machinery electrical machinery transportation equipment miscellaneous manufactures weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b exports to Canada 1988–1999 model -20.5 -4.1 -35.5 27.3 70.4 10.8 939.7 21.6 1847.0 19.2 1470.3 36.2 153.0 38.6 4387.6 15.0 4933.2 36.2 23.9 32.9 476.3 15.0 204.6 36.0 -10.6 32.9 2366.2 -6.7 1396.1 5.7 676.8 13.3 32.5 19.4 -35.4 138.1 610.4 41.9 570.6 17.3 1349.2 137.3 2303.4 3.3 379.4 61.1 exports to United States 1988–1999 model -15.0 2.5 -22.9 26.9 9.4 7.5 832.3 11.8 829.6 18.6 618.3 11.7 111.1 4.6 145.6 -2.7 181.2 7.6 70.3 13.9 122.1 3.9 70.4 17.0 66.4 34.1 783.8 -5.3 222.3 3.7 469.8 32.3 40.9 30.8 111.2 156.5 477.2 26.8 123.6 18.5 744.9 178.0 349.0 6.2 181.5 43.2 0.19 0.01 0.71 0.04 120.32 2.07 38.13 3.87 Changes in U.S. Exports/GDP in the Brown-Deardorff-Stern Model (Percent) sector agriculture mining and quarrying food textiles clothing leather products footwear wood products furniture and fixtures paper products printing and publishing chemicals petroleum and products rubber products nonmetal mineral products glass products iron and steel nonferrous metals metal products nonelectrical machinery electrical machinery transportation equipment miscellaneous manufactures weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b exports to Canada 1988–1999 model -24.1 5.1 -23.6 1.0 62.4 12.7 177.2 44.0 145.5 56.7 29.9 7.9 48.8 45.7 76.4 6.7 83.8 35.6 -20.5 18.9 50.8 3.9 49.8 21.8 -6.9 0.8 95.6 19.1 56.5 11.9 50.5 4.4 0.6 11.6 -20.7 -6.7 66.7 18.2 36.2 9.9 154.4 14.9 36.5 -4.6 117.3 11.5 exports to Mexico 1988–1999 model 6.5 7.9 -19.8 0.5 37.7 13.0 850.5 18.6 543.0 50.3 87.7 15.5 33.1 35.4 25.7 7.0 224.1 18.6 -41.9 -3.9 507.9 -1.1 61.5 -8.4 -41.1 -7.4 165.6 12.8 55.9 0.8 112.9 42.3 144.5 -2.8 -28.7 -55.1 301.4 5.4 350.8 -2.9 167.8 -10.9 290.3 9.9 362.3 -9.4 -0.01 0.14 37.27 -0.02 0.50 0.02 190.89 3.42 Changes in Canadian Trade/GDP in Cox-Harris Model (Percent) variable total trade trade with Mexico trade with United States data 1988-2000 57.2 280.0 76.2 weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b model 10.0 52.2 20.0 0.99 0.52 38.40 1.93 Changes in Canadian Trade/GDP in the Cox-Harris Model (Percent) sector agriculture forestry fishing mining food, beverages, and tobacco rubber and plastics textiles and leather wood and paper steel and metal products transportation equipment machinery and appliances nonmetallic minerals refineries chemicals and misc. manufactures total exports 1988-2000 model -13.7 -4.1 215.5 -11.5 81.5 -5.4 21.7 -7.0 50.9 18.6 194.4 24.5 201.1 108.8 31.9 7.3 30.2 19.5 66.3 3.5 112.9 57.1 102.7 31.8 20.3 -2.7 53.3 28.1 total imports 1988-2000 model 4.6 7.2 -21.5 7.1 107.3 9.5 32.1 4.0 60.0 3.8 87.7 13.8 24.6 18.2 97.3 7.2 52.2 10.0 29.7 3.0 65.0 13.3 3.6 7.3 5.1 1.5 92.5 10.4 weighted correlation with data variance decomposition of change 0.49 0.32 0.85 0.08 41.85 0.81 22.00 3.55 regression coefficient a regression coefficient b What Do We Learn from these Evaluations? The Spanish model seems to have been far more successful in predicting the consequences of policy changes than the three models of NAFTA, but x Kehoe, Polo, and Sancho (KPS) knew the structure of their model well enough to precisely identify the relationships between the variables in their model with those in the data; x KPS were able to use the model to carry out numerical exercises to incorporate the impact of exogenous shocks. KPS had an incentive to show their model in the best possible light. 3. Much of the growth of trade after a trade liberalization experience is growth on the extensive margin. Models need to allow for corner solutions or fixed costs. T. J. Kehoe and K. J. Ruhl, “How Important is the New Goods Margin in International Trade?” Federal Reserve Bank of Minneapolis, 2002. K. J. Ruhl, “Solving the Elasticity Puzzle in International Economics,” University of Texas at Austin, 2005. What happens to the least-traded goods: Over the business cycle? During trade liberalization? Indirect evidence on the extensive margin 4. Modeling the fixed costs may explain why real exchange rate data indicate that more arbitrage across countries that have a strong bilateral trade relationship. C. M. Betts and T. J. Kehoe, “Real Exchange Rate Movements and the Relative Price of Nontraded Goods,” University of Minnesota and University of Southern California, 2003. C. M. Betts and T. J. Kehoe, “U.S. Real Exchange Rate Fluctuations and Relative Price Fluctuations,” University of Minnesota and University of Southern California, 2003. 5. The major determinant of large macroeconomic fluctuations in Mexico is fluctuations in total factor productivity. If trade liberalization and the 1995 financial crisis have had large macroeconomic effects, it is through their effects on TFP. R. Bergoeing, P. J. Kehoe, T. J. Kehoe, and R. Soto “A Decade Lost and Found: Mexico and Chile in the 1980s,” Review of Economic Dynamics, 5 (2002), 166-205. T. J. Kehoe and E. C. Prescott, “Great Depressions of the Twentieth Century,” Review of Economic Dynamics, 5 (2002), 118. T. J. Kehoe and K. J. Ruhl, “Sudden Stops, Sectoral Reallocations, and Productivity Drops,” University of Minnesota, 2006. 1. Standard theory (hybrid Heckscher-Ohlin/New Trade Theory) does not well when matched with the data on the growth and composition of trade. In the 1980s and 1990s trade economists reached a consensus that North-North trade — trade among rich countries — was driven by forces captured by the New Trade Theory and North-South trade — trade between rich countries and poor countries — was driven by forces captured by Heckscher-Ohlin theory. (South-South trade was negligible.) A. V. Deardorff, “Testing Trade Theories and Predicting Trade Flows,” in R. W. Jones and P. B. Kenen, editors, Handbook of International Economics, volume l, North-Holland, 1984, 467-517. J. Markusen, “Explaining the Volume of Trade: An Eclectic Approach,” American Economic Review, 76 (1986), 1002-1011. In fact, a calibrated version of this hybrid model does not match the data. R. Bergoeing and T. J. Kehoe, “Trade Theory and Trade Facts,” Federal Reserve Bank of Minneapolis, 2003. TRADE THEORY Traditional trade theory — Ricardo, Heckscher-Ohlin — says countries trade because they are different. In 1990 by far the largest bilateral trade relation in the world was U.S.-Canada. The largest two-digit SITC export of the United States to Canada was 78 Road Vehicles. The largest two-digit SITC export of Canada to the United States was 78 Road Vehicles. The New Trade Theory — increasing returns, taste for variety, monopolistic competition — explains how similar countries can engage in a lot of intraindustry trade. Helpman and Krugman (1985) Markusen (1986) TRADE THEORY AND TRADE FACTS x Some recent trade facts x A “New Trade Theory” model x Accounting for the facts x Intermediate goods? x Policy? How important is the quantitative failure of the New Trade Theory? Where should trade theory and applications go from here? SOME RECENT TRADE FACTS x The ratio of trade to product has increased. World trade/world GDP increased by 59.3 percent 1961-1990. OECD-OECD trade/OECD GDP increased by 111.5 percent 1961-1990. x Trade has become more concentrated among industrialized countries OECD-OECD trade/OECD-RW trade increased by 87.1 percent 1961-1990. x Trade among industrialized countries is mostly intraindustry trade Grubel-Lloyd index for OECD-OECD trade in 1990 is 68.4. Grubel-Lloyd index for OECD-RW trade in 1990 is 38.1. OECD-OECD Trade / OECD GDP 0.12 0.10 0.08 0.06 0.04 0.02 0.00 1960 1963 1966 1969 1972 1975 year 1978 1981 1984 1987 1990 OECD-OECD Trade / OECD-RW Trade 1.80 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1960 1963 1966 1969 1972 1975 year 1978 1981 1984 1987 1990 Helpman and Krugman (1985): “These....empirical weaknesses of conventional trade theory...become understandable once economies of scale and imperfect competition are introduced into our analysis.” Markusen, Melvin, Kaempfer, and Maskus (1995): “Thus, nonhomogeneous demand leads to a decrease in NorthSouth trade and to an increase in intraindustry trade among the northern industrialized countries. These are the stylized facts that were to be explained.” Goal: To measure how much of the increase in the ratio of trade to output in the OECD and of the concentration of world trade among OECD countries can be accounted for by the “New Trade Theory.” PUNCHLINE In a calibrated general equilibrium model, the New Trade Theory cannot account for the increase in the ratio of trade to output in the OECD. Back-of-the-envelope calculations: Suppose that the world consists of the OECD and the only trade is manufactures. With Dixit-Stiglitz preferences, country j exports all of its production of manufactures Ymj except for the fraction s j Y j / Y oe that it retains for domestic consumption. World imports: M World trade/GDP: M Y oe World trade/GDP: n ¦j M Ymoe Ymoe Y oe j j 1 s Y m. 1 1 ¦ n j oe Y j 2 m s ( ) . oe 1 Y M Y oe n M Ymoe Ymoe Y oe 1 ¦ n j oe Y j 2 m s ( ) . oe 1 Y 1 ¦ j 1 ( s j ) 2 goes from 0.663 in 1961 to 0.827 in 1990. Ymoe / Y oe goes from 0.295 in 1961 to 0.222 in 1990. 0.663 u 0.295 0.196 | 0.184 0.827 u 0.222 . Effects cancel! A “NEW TRADE THEORY” MODEL Environment: x Static: endowments of factors are exogenous x 2 regions: OECD and rest of world x 2 traded goods: homogeneous — primaries (CRS) and differentiated — manufactures (IRS) x 1 nontraded good — services (CRS) x 2 factors: (effective) labor and capital x Identical technologies and preferences (love for variety) across regions x Primaries are inferior to manufactures We only consider merchandise trade in both the data and in the model. Key Features of the Model Consumers' problem: max s.t. E p (c pj J p )K E m ( ³ cmj ( z ) U dz p )K / U E s (csj J s )K 1 Dw K q p c pj ³ w qm ( z )cmj ( z )dz p qsj csj d r j k j w j h j . D Firms' problems Primaries and Services: Standard CRS problems. Ypj T p ( K pj )D p ( H pj )1D p Ys j Ts ( K sj )D s ( H sj )1D s Manufactures: Standard (Dixit-Stiglitz) monopolistically competitive problem: x Fixed cost. Ym ( z ) max ª¬Tm K m ( z )Dm H m ( z )1Dm F ,0 º¼ x Firm z sets its price qm ( z ) to max profits given all of the other prices. n Ym ( z ) ¦ j 1 Cmj ( z ) Cmrw ( z ) . 1 1K Cmj ( z ) Em (r j K j w j H j q p J p N j qsj J s N j ) 1 1U U 1U qm ( z ) ª ³ D w qm ( z c) dz cº ¬ ¼ UK U (1K ) (1U ) U K 1K ' U K 1 ª º 1U 1K 1K ' Em « ³ w qm ( z c) dz c » E s qs D ¬ ¼ x Every firm is uniquely associated with only one variety (symmetry). x Free entry. [0, d w ] with d w finite and endogenously determined. x Dw K 1K E p qp 1 1K 1 1K Volume of Trade Let s j be the share of country j , j production of manufactures, sj ³D j 1,..., n, rw , in the world Ym ( z )dz / ³D w Ym ( z )dz Ymj / Ymw . The imports by country j from the OECD are M oej (1 s rw s j )Cmj M oerw (1 s rw )Cmrw . Total imports in the OECD from the other OECD countries are n M oe oe n j 2 rw oe M (1 s ( s ) /(1 s )) C ¦ ¦ m . j oe j 1 rw j 1 OECD in 1990 Country Australia Austria Belgium-Lux Canada Denmark Finland France Germany Greece Iceland Ireland Italy Share of GDP % 1.79 0.97 1.26 3.45 0.78 0.81 7.26 9.96 0.50 0.04 0.28 6.64 Country Japan Netherlands New Zealand Norway Portugal Spain Sweden Switzerland Turkey United Kingdom United States Share of GDP % 18.04 1.72 0.26 0.70 0.41 3.00 1.40 0.17 0.91 5.92 33.72 ACCOUNTING FOR THE FACTS Compare the changes that the model predicts for 1961-1990 with what actually took place. Focus on key variables: OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manfacturing GDP/OECD GDP Calibrate to 1990 data. Backcast to 1961 by imposing changes in parameters: relative sizes of countries in the OECD populations sectoral productivities endowments ACCOUNTING FOR THE FACTS Benchmark 1990 OECD Data Set (Billion U.S. dollars) H ioe K ioe Yi oe Cioe Yi oe Cioe Primaries Manufactures 228 2,884 441 775 669 3,659 862 3,466 -193 193 Services 8,644 3,497 12,141 12,141 0 Total 11,756 4,713 16,469 16,469 0 ACCOUNTING FOR THE FACTS Benchmark 1990 Rest of the World Data Set (Billion U.S. dollars) Yi rw Cirw Yi rw Cirw Primaries Manufactures 1,223 1,159 1,030 1,352 193 -193 Services 3,447 3,447 0 Total 5,829 5,829 0 ACCOUNTING FOR THE FACTS x N oe x ¦ i 854 , N rw rw Y p ,m , s i x Set q p qm ( z ) values). xU 4, 428 . ¦ i rw C p ,m , s i qs w 5,829 . r 1 (quantities are 1990 1/1.2 (Morrison 1990, Martins, Scarpetta, and Pilat 1996). x Normalize d w 100 . x Calibrate H rw , K rw so that benchmark data set is an equilibrium. x Alternative calibrations of utility parameters J p , J s , and K . OECD in 1961 Country Austria Belgium-Lux Canada Denmark France Germany Greece Iceland Ireland Italy Share of GDP % 0.75 1.25 4.22 0.70 6.99 9.71 0.50 0.03 0.21 4.64 Country Share of GDP % Netherlands 1.37 Norway 0.60 Portugal 0.32 Spain 1.38 Sweden 1.62 Switzerland 1.07 Turkey 0.83 United Kingdom 8.08 United States 55.74 Numerical Experiments Calculate equilibrium in 1961: T p ,1961 T m ,1961 T s ,1961 N oe T p ,1990 T m ,1990 /1.01429 , F1961 F1990 /1.01429 T s ,1990 /1.00529 (Echevarria 1997) 536, N rw 2,545 Numerical Experiments oe oe rw rw Choose H1961 , K1961 , H1961 , K1961 so that ¦ ¦ ¦ ¦ i oe oe Y / N 1990 p ,m , s i ,1990 oe i p ,m , s i ,1961 Y i oe 1961 /N rw rw Y / N 1990 p ,m , s i ,1990 rw i p ,m , s i ,1961 Y rw 1961 /N oe oe K1961 K1990 oe oe H1961 H1990 q p ,1961 (Yprw,1961 C prw,1961 ) ¦ i p ,m , s rw i ,1961 i ,1961 q Y 2.400 2.055 0.050 How Can the Model Work in Matching the Facts? x The ratio of trade to product has increased: The size distribution of countries has become more equal (Helpman-Krugman). x Trade has become more concentrated among industrialized countries: OECD countries have comparative advantage in manufactures, while the RW has comparative advantage in primaries. Because they are inferior to manufactures, primaries become less important in trade as the world becomes richer (Markusen). How Can the Model Work in Matching the Facts? x Trade among industrialized countries is largely intraindustry trade: OECD countries export manufactures. Because of taste for variety, every country consumes some manufactures from every other country (Dixit-Stiglitz). x The different total factor productivity growth rates across sectors imply that the price of manufactures relative to primaries and services has fallen sharply between 1961 and 1990. If price elasticities of demand are not equal to one, a lot can happen. Experiment 1 Jp Jp K 0 Data OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 1. Jp = 0, Js = 0, K = 0 OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 1961 1990 Change 0.053 0.844 0.295 0.112 1.579 0.222 111.5% 87.1% 24.6% 0.108 0.893 0.223 0.136 1.169 0.222 25.8% 30.9% 0.4% Experiment 2 Jp = 169.5, Js = 314.7 to match consumption in RW in 1990, K=0 1961 Data OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 2. Jp = 169.5, Js = 314.7, K = 0 OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 1990 Change 0.053 0.844 0.295 0.112 111.5% 1.579 87.1% 0.222 24.6% 0.103 0.739 0.225 0.132 1.060 0.222 28.1% 43.6% 1.4% Experiment 3 Jp = 169.5, Js = 314.7, K = 0.559 to match growth in OECD-OECD Trade/OECD GDP 1961 Data OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 3. Jp = 169.5, Js = 314.7, K = 0.559 OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 1990 Change 0.053 0.844 0.295 0.112 111.5% 1.579 87.1% 0.222 24.6% 0.063 0.738 0.137 0.132 111.5% 1.060 43.7 % 0.222 62.7% Experiments 4 and 5 Jp = 169.5, Js = 314.7, reasonable values of K ( 0.5 t 1/(1 K ) t 0.1) 1961 1990 Change Data OECD-OECD Trade/OECD GDP 0.053 0.112 111.5% OECD-OECD Trade/OECD-RW Trade 0.844 1.579 87.1% OECD Manf GDP/OECD GDP 0.295 0.222 24.6% 4. Jp = 169.5, Js = 314.7, K = 1 OECD-OECD Trade/OECD GDP 0.118 0.132 11.7% OECD-OECD Trade/OECD-RW Trade 0.739 1.060 43.5% OECD Manf GDP/OECD GDP 0.259 0.222 14.1% 5. Jp = 169.5, Js = 314.7, K = 9 OECD-OECD Trade/OECD GDP 0.118 0.132 1.6% OECD-OECD Trade/OECD-RW Trade 0.739 1.060 43.5% OECD Manf GDP/OECD GDP 0.284 0.222 21.8% Sensitivity Analysis: Alternative Calibration Methodologies x Alternative specifications of nonhomogeneity x Gross imports calibration x Alternative RW endowment calibration x Alternative RW growth calibration x Intermediate goods INTERMEDIATE GOODS? Ypj j j ª X j ³ w X mp º ( z ) dz X D 1 D p p pp sp » min « , D , , T p K pj H pj amp asp «¬ a pp »¼ Ym ( z ) Ys j j ª X j ( z ) ³ w X mm ( z , z ')dz ' X j ( z ) º pm « , D , sm , » amm asm » min « a pm « » Dm 1D m F «¬Tm K m ( z ) H m ( z ) »¼ j ª X j ³ w X ms º j ( z ) dz D 1 D X s s ps min « , D , ss , Ts K sj H sj » ams ass «¬ a ps »¼ Results for Model with Intermediate Goods 1961 Data OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 4. Jp = 307.8, Js = 262.2, K = 1 OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 5. Jp = 307.8, Js = 262.2, K = 9 OECD-OECD Trade/OECD GDP OECD-OECD Trade/OECD-RW Trade OECD Manf GDP/OECD GDP 1990 Change 0.053 0.112 0.844 1.579 0.295 0.222 111.5% 87.1% 24.6% 0.323 0.370 0.994 1.305 0.263 0.222 14.5% 31.3% 15.6% 0.337 0.370 0.933 1.305 0.307 0.222 9.7% 39.9% 27.5% POLICY? In a version of our model with n OECD countries, a manufacturing sector, and a uniform ad valorem tariff W, the ratio of exports to income is given by M Y (n 1)C f Y n 1 n 1 (1 W )1/(1 U ) Fixing n to replicate the size distribution of national incomes in the OECD, and setting U 1/1.2 , a fall in W from 0.45 to 0.05 produces an increase in the ratio of trade to output as seen in the data. World Trade / World GDP 0.90 0.80 0.70 rho=1/2.0 0.60 0.50 0.40 rho=1/1.5 0.30 rho=1/1.1 0.20 rho=1/1.2 0.10 0.00 0 0.1 0.2 0.3 0.4 0.5 tau 0.6 0.7 0.8 0.9 1 2. Applied general equilibrium models that put the standard theory to work do not well in predicting the impact of trade liberalization experiences like NAFTA. Applied general equilibrium models were the only analytical game in town when it came to analyzing the impact of NAFTA in 1992-1993. Typical sort of model: Static applied general equilibrium model with large number of industries and imperfect competition (Dixit-Stiglitz or Eastman-Stykolt) and finite number of firms in some industries. In some numerical experiments, new capital is placed in Mexico owned by consumers in the rest of North America to account for capital flows. Examples: Brown-Deardorff-Stern model of Canada, Mexico, and the United States Cox-Harris model of Canada Sobarzo model of Mexico T. J. Kehoe, “An Evaluation of the Performance of Applied General Equilibrium Models of the Impact of NAFTA,” in T. J. Kehoe, T. N. Srinivasan, and J. Whalley, editors, Frontiers in Applied General Equilibrium Modeling: Essays in Honor of Herbert Scarf, Cambridge University Press, 2005, 341-77. Research Agenda: x Compare results of numerical experiments of models with data. x Determine what shocks — besides NAFTA policies — were important. x Construct a simple applied general equilibrium model and perform experiments with alternative specifications to determine what was wrong with the 1992-1993 models. Applied GE Models Can Do a Good Job! Spain: Kehoe-Polo-Sancho (1992) evaluation of the performance of the Kehoe-Manresa-Noyola-Polo-Sancho-Serra MEGA model of the Spanish economy: A Shoven-Whalley type model with perfect competition, modified to allow government and trade deficits and unemployment (Kehoe-Serra). Spain’s entry into the European Community in 1986 was accompanied by a fiscal reform that introduced a value-added tax (VAT) on consumption to replace a complex range of indirect taxes, including a turnover tax applied at every stage of the production process. What would happen to tax revenues? Trade reform was of secondary importance. Canada-U.S.: Fox (1999) evaluation of the performance of the Brown-Stern (1989) model of the 1989 Canada-U.S. FTA. Other changes besides policy changes are important! Changes in Consumer Prices in the Spanish Model (Percent) sector food and nonalcoholic beverages tobacco and alcoholic beverages clothing housing household articles medical services transportation recreation other services data model model model 1985-1986 policy only shocks only policy&shocks 1.8 -2.3 4.0 1.7 3.9 2.5 3.1 5.8 2.1 5.6 0.9 6.6 -3.3 -2.2 -2.7 -4.8 0.1 2.2 0.7 2.9 -0.7 -4.8 0.6 -4.2 -4.0 2.6 -8.8 -6.2 -1.4 -1.3 1.5 0.1 2.9 1.1 1.7 2.8 weighted correlation with data variance decomposition of change -0.08 0.30 0.87 0.77 0.94 0.85 regression coefficient a regression coefficient b 0.00 -0.08 0.00 0.54 0.00 0.67 Measures of Accuracy of Model Results 1. Weighted correlation coefficient. 2. Variance decomposition of the (weighted) variance of the changes in the data: vardec( y data , y model ) var ( y model ) . model data model var ( y ) var ( y y ) 3, 4. Estimated coefficients a and b from the (weighted) regression xidata a bximodel ei . Changes in Value of Gross Output/GDP in the Spanish Model (Percent) sector agriculture energy basic industry machinery automobile industry food products other manufacturing construction commerce transportation services government services data 1985-1986 -0.4 -20.3 -9.0 3.7 1.1 -1.8 0.5 5.7 6.6 -18.4 8.7 7.6 weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b model policy only -1.1 -3.5 1.6 3.8 3.9 -2.4 -1.7 8.5 -3.6 -1.5 -1.1 3.4 model model shocks only policy&shocks 8.3 6.9 -29.4 -32.0 -1.8 -0.1 1.0 5.0 4.7 8.6 4.7 2.1 2.3 0.5 1.4 10.3 4.4 0.4 1.0 -0.7 5.8 4.5 0.9 4.3 0.16 0.11 0.80 0.73 0.77 0.71 -0.52 0.44 -0.52 0.75 -0.52 0.67 Changes in Trade/GDP in the Spanish Model (Percent) direction of exports Spain to rest of E.C. Spain to rest of world rest of E.C. to Spain rest of world to Spain data model model model 1985-1986 policy only shocks only policy&shocks -6.7 -3.2 -4.9 -7.8 -33.2 -3.6 -6.1 -9.3 14.7 4.4 -3.9 0.6 -34.1 -1.8 -16.8 -17.7 weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b 0.69 0.02 0.77 0.17 0.90 0.24 -12.46 5.33 2.06 2.21 5.68 2.37 Changes in Composition of GDP in the Spanish Model (Percent of GDP) variable wages and salaries business income net indirect taxes and tariffs data model 1985-1986 policy only -0.53 -0.87 -1.27 -1.63 1.80 2.50 model shocks only -0.02 0.45 -0.42 model policy&shocks -0.91 -1.24 2.15 correlation with data variance decomposition of change 0.998 0.93 -0.94 0.04 0.99 0.96 regression coefficient a regression coefficient b private consumption private investment government consumption government investment exports -imports 0.00 0.73 -1.23 1.81 -0.06 -0.06 -0.42 -0.03 0.00 -3.45 -0.51 -0.58 -0.38 -0.07 -0.69 2.23 0.00 0.85 -1.78 1.32 -0.44 -0.13 -1.07 2.10 correlation with data variance decomposition of change 0.40 0.20 0.77 0.35 0.83 0.58 regression coefficient a regression coefficient b 0.00 0.87 0.00 1.49 0.00 1.24 -0.81 1.09 -0.02 -0.06 -3.40 3.20 Public Finances in the Spanish Model (Percent of GDP) data model model model variable 1985-1986 policy only shocks only policy&shocks indirect taxes and subsidies 2.38 3.32 -0.38 2.98 tariffs -0.58 -0.82 -0.04 -0.83 social security payments 0.04 -0.19 -0.03 -0.22 direct taxes and transfers -0.84 -0.66 0.93 0.26 government capital income -0.13 -0.06 0.02 -0.04 correlation with data variance decomposition of change regression coefficient a regression coefficient b 0.99 0.93 -0.70 0.08 0.92 0.86 -0.06 0.74 0.35 -1.82 -0.17 0.80 Models of NAFTA Did Not Do a Good Job! Ex-post evaluations of the performance of applied GE models are essential if policy makers are to have confidence in the results produced by this sort of model. Just as importantly, they help make applied GE analysis a scientific discipline in which there are well-defined puzzles and clear successes and failures for alternative hypotheses. Changes in Trade/GDP in Brown-Deardorff-Stern Model (Percent) variable Canadian exports Canadian imports Mexican exports Mexican imports U.S. exports U.S. imports data 1988-1999 52.9 57.7 240.6 50.5 19.1 29.9 weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b model 4.3 4.2 50.8 34.0 2.9 2.3 0.64 0.08 23.20 2.43 Changes in Canadian Exports/ GDP in the Brown-Deardorff-Stern Model (Percent) sector agriculture mining and quarrying food textiles clothing leather products footwear wood products furniture and fixtures paper products printing and publishing chemicals petroleum and products rubber products nonmetal mineral products glass products iron and steel nonferrous metals metal products nonelectrical machinery electrical machinery transportation equipment miscellaneous manufactures weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b exports to Mexico 1988–1999 model 122.5 3.1 -34.0 -0.3 89.3 2.2 268.2 -0.9 1544.3 1.3 443.0 1.4 517.0 3.7 232.6 4.7 3801.7 2.7 240.7 -4.3 6187.4 -2.0 37.1 -7.8 678.1 -8.5 647.4 -1.0 333.5 -1.8 264.4 -2.2 195.2 -15.0 38.4 -64.7 767.0 -10.0 376.8 -8.9 633.9 -26.2 305.8 -4.4 1404.5 -12.1 exports to United States 1988–1999 model 106.1 3.4 75.8 0.4 91.7 8.9 97.8 15.3 237.1 45.3 -14.4 11.3 32.8 28.3 36.5 0.1 282.6 12.5 113.7 -1.8 37.2 -1.6 109.4 -3.1 -42.5 0.5 113.4 9.5 20.5 1.2 74.5 30.4 92.1 12.9 34.7 18.5 102.2 15.2 28.9 3.3 88.6 14.5 30.7 10.7 100.0 -2.1 -0.91 0.003 -0.43 0.02 249.24 -15.48 79.20 -2.80 Changes in Mexican Exports/GDP in the Brown-Deardorff-Stern Model (Percent) sector agriculture mining and quarrying food textiles clothing leather products footwear wood products furniture and fixtures paper products printing and publishing chemicals petroleum and products rubber products nonmetal mineral products glass products iron and steel nonferrous metals metal products nonelectrical machinery electrical machinery transportation equipment miscellaneous manufactures weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b exports to Canada 1988–1999 model -20.5 -4.1 -35.5 27.3 70.4 10.8 939.7 21.6 1847.0 19.2 1470.3 36.2 153.0 38.6 4387.6 15.0 4933.2 36.2 23.9 32.9 476.3 15.0 204.6 36.0 -10.6 32.9 2366.2 -6.7 1396.1 5.7 676.8 13.3 32.5 19.4 -35.4 138.1 610.4 41.9 570.6 17.3 1349.2 137.3 2303.4 3.3 379.4 61.1 exports to United States 1988–1999 model -15.0 2.5 -22.9 26.9 9.4 7.5 832.3 11.8 829.6 18.6 618.3 11.7 111.1 4.6 145.6 -2.7 181.2 7.6 70.3 13.9 122.1 3.9 70.4 17.0 66.4 34.1 783.8 -5.3 222.3 3.7 469.8 32.3 40.9 30.8 111.2 156.5 477.2 26.8 123.6 18.5 744.9 178.0 349.0 6.2 181.5 43.2 0.19 0.01 0.71 0.04 120.32 2.07 38.13 3.87 Changes in U.S. Exports/GDP in the Brown-Deardorff-Stern Model (Percent) sector agriculture mining and quarrying food textiles clothing leather products footwear wood products furniture and fixtures paper products printing and publishing chemicals petroleum and products rubber products nonmetal mineral products glass products iron and steel nonferrous metals metal products nonelectrical machinery electrical machinery transportation equipment miscellaneous manufactures weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b exports to Canada 1988–1999 model -24.1 5.1 -23.6 1.0 62.4 12.7 177.2 44.0 145.5 56.7 29.9 7.9 48.8 45.7 76.4 6.7 83.8 35.6 -20.5 18.9 50.8 3.9 49.8 21.8 -6.9 0.8 95.6 19.1 56.5 11.9 50.5 4.4 0.6 11.6 -20.7 -6.7 66.7 18.2 36.2 9.9 154.4 14.9 36.5 -4.6 117.3 11.5 exports to Mexico 1988–1999 model 6.5 7.9 -19.8 0.5 37.7 13.0 850.5 18.6 543.0 50.3 87.7 15.5 33.1 35.4 25.7 7.0 224.1 18.6 -41.9 -3.9 507.9 -1.1 61.5 -8.4 -41.1 -7.4 165.6 12.8 55.9 0.8 112.9 42.3 144.5 -2.8 -28.7 -55.1 301.4 5.4 350.8 -2.9 167.8 -10.9 290.3 9.9 362.3 -9.4 -0.01 0.14 37.27 -0.02 0.50 0.02 190.89 3.42 Changes in Canadian Trade/GDP in Cox-Harris Model (Percent) variable total trade trade with Mexico trade with United States data 1988-2000 57.2 280.0 76.2 weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b model 10.0 52.2 20.0 0.99 0.52 38.40 1.93 Changes in Canadian Trade/GDP in the Cox-Harris Model (Percent) sector agriculture forestry fishing mining food, beverages, and tobacco rubber and plastics textiles and leather wood and paper steel and metal products transportation equipment machinery and appliances nonmetallic minerals refineries chemicals and misc. manufactures total exports 1988-2000 model -13.7 -4.1 215.5 -11.5 81.5 -5.4 21.7 -7.0 50.9 18.6 194.4 24.5 201.1 108.8 31.9 7.3 30.2 19.5 66.3 3.5 112.9 57.1 102.7 31.8 20.3 -2.7 53.3 28.1 total imports 1988-2000 model 4.6 7.2 -21.5 7.1 107.3 9.5 32.1 4.0 60.0 3.8 87.7 13.8 24.6 18.2 97.3 7.2 52.2 10.0 29.7 3.0 65.0 13.3 3.6 7.3 5.1 1.5 92.5 10.4 weighted correlation with data variance decomposition of change 0.49 0.32 0.85 0.08 41.85 0.81 22.00 3.55 regression coefficient a regression coefficient b Changes in Mexican Trade/GDP in the Sobarzo Model (Percent) sector agriculture mining petroleum food beverages tobacco textiles wearing apparel leather wood paper chemicals rubber nonmetallic mineral products iron and steel nonferrous metals metal products nonelectrical machinery electrical machinery transportation equipment other manufactures weighted correlation with data variance decomposition of change regression coefficient a regression coefficient b exports to North America 1988–2000 model -15.3 -11.1 -23.2 -17.0 -37.6 -19.5 5.2 -6.9 42.0 5.2 -42.3 2.8 534.1 1.9 2097.3 30.0 264.3 12.4 415.1 -8.5 12.8 -7.9 41.9 -4.4 479.0 12.8 37.5 -6.2 35.9 -4.9 -40.3 -9.8 469.5 -4.4 521.7 -7.4 3189.1 1.0 224.5 -5.0 975.1 -4.5 0.61 0.0004 imports from North America 1988–2000 model -28.2 3.4 -50.7 13.2 65.9 -6.8 11.8 -5.0 216.0 -1.8 3957.1 -11.6 833.2 -1.2 832.9 4.5 621.0 -0.4 168.9 11.7 68.1 -4.7 71.8 -2.7 792.0 -0.1 226.5 10.9 40.3 17.7 101.2 9.8 478.7 9.5 129.0 20.7 749.1 9.6 368.0 11.2 183.6 4.2 0.23 0.002 495.08 30.77 174.52 5.35 What Do We Learn from these Evaluations? The Spanish model seems to have been far more successful in predicting the consequences of policy changes than the three models of NAFTA, but x Kehoe, Polo, and Sancho (KPS) knew the structure of their model well enough to precisely identify the relationships between the variables in their model with those in the data; x KPS were able to use the model to carry out numerical exercises to incorporate the impact of exogenous shocks. KPS had an incentive to show their model in the best possible light. 3. Much of the growth of trade after a trade liberalization experience is growth on the extensive margin. Models need to allow for corner solutions or fixed costs. T. J. Kehoe and K. J. Ruhl, “How Important is the New Goods Margin in International Trade?” Federal Reserve Bank of Minneapolis, 2002. K. J. Ruhl, “Solving the Elasticity Puzzle in International Economics,” University of Texas at Austin, 2005. What happens to the least-traded goods: Over the business cycle? During trade liberalization? Indirect evidence on the extensive margin How Does Trade Grow? ʀ Intensive Margin: growth in goods already traded ʀ Extensive Margin: trade in goods not traded before The Extensive Margin ʀ The Extensive Margin has recently gained attention ʀ Models ɿ Melitz (2003) ɿ Alessandria and Choi (2003) ɿ Ruhl (2004) ʀ Empirically ɿ Hummels and Klenow (2002) ɿ Eaton, Kortum and Kramarz (2004) What Happens to the Extensive Margin? ʀ During trade liberalization? ɿ Large changes in the extensive margin ʀ Over the business cycle? ɿ Little change in extensive margin Evidence from Trade Agreements ʀ Events ɿ Greece’s Accession to the European Econ. Community - 1981 ɿ Portugal’s Accession to the European Community - 1986 ɿ Spain’s Accession to the European Community - 1986 ɿ U.S.-Canada Free Trade Agreement - 1989 ɿ North American Free Trade Agreement - 1994 ʀ Data ʀ Four-digit SITC bilateral trade data (OECD) ɿ 789 codes in revision 2 ʀ Indirect Evidence Measure One 1. Rank codes from lowest value of exports to highest value of exports based on average of first 3 years 2. Form sets of codes by cumulating exports: the first 742.9 codes make up 10 percent of exports; the next 24.1 codes make up 10 percent of exports; and so on. 3. Calculate each set’s share of export value at the end of the sample period. Composition of Exports: Mexico to Canada 0.30 Fraction of 1989 Exports 0.25 0.20 0.15 736.6 28.8 10.3 5.3 1.8 1.9 1.5 0.8 1.4 0.8 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.10 0.05 0.00 Cummulative Fraction of 1989 Exports Composition of Exports: Mexico to Canada 0.30 Fraction of 1999 Exports 0.25 736.6 0.20 1.4 28.8 0.15 10.3 1.5 0.10 1.9 5.3 0.05 0.8 1.8 0.8 0.00 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Cummulative Fraction of 1989 Exports 0.9 1.0 Composition of Exports: Greece to EEC 0.40 730.9 Fraction of 1986 Exports 0.35 0.30 0.25 0.20 15.3 0.15 6.6 0.10 22.6 4.7 2.5 0.05 2.3 1.8 0.7 0.8 1.4 0.9 0.9 1.0 0.00 0.1 0.2 0.3 0.4 0.5 0.6 Cummulative Fraction of 1976 Exports Measure Two 1. Order codes as before. 2. Cumulate exports as before. 3. Follow the evolution of the first (least-traded) set’s share of total exports before, during, and after the liberalization. Exports: Mexico to Canada Fraction of Total Export Value 0.30 0.20 0.10 0.00 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Exports: Canada to Mexico 0.35 Fraction of Total Export Value 0.30 0.25 0.20 0.15 0.10 0.05 0.00 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Exports: Greece to EEC Fraction of Total Export Value 0.4 0.3 0.2 0.1 0.0 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 Trade Liberalization and the Extensive Margin Period 1989-1999 1989-1999 1989-1999 1989-1999 1989-1999 1989-1999 1978-1986 1982-1987 1982-1987 Trade Flow Mexico - U.S. U.S. – Mexico Mexico - Canada Canada - Mexico Canada - U.S. U.S. – Canada Greece to the EEC Spain to the EC Portugal to the EC Share of Export Growth 0.153 0.118 0.231 0.307 0.162 0.130 0.371 0.128 0.147 Business Cycles and the Extensive Margin ʀ Over same period, consider countries with stable policy ɿ U.S. – Japan ɿ U.S. – U.K. ɿ U.S. – Germany Composition of Exports: U.S. to Germany 0.30 Fraction of 1999 Exports 0.25 0.20 0.15 604.1 86.8 41.7 24.9 0.10 0.20 0.30 0.40 0.10 12.5 7.3 5.0 0.50 0.60 0.70 3.6 2.0 1.1 0.90 1.00 0.05 0.00 0.80 Cummulative Fraction of 1989 Exports Exports: United States to Germany Fraction of Total Export Value 0.30 0.25 0.20 0.15 0.10 U.S. Exports to Germany 0.05 0.00 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Exports: Germany and the United States Fraction of Total Export Value 0.30 0.25 0.20 0.15 German Exports to U.S. 0.10 U.S. Exports to Germany 0.05 0.00 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Business Cycles and the Extensive Margin Period Trade Flow Share of Export Growth 1989-1999 U.S. - U.K. 0.096 1989-1999 U.K. - U.S. 0.128 1989-1999 U.S. - Japan 0.130 1989-1999 Japan - U.S. 0.103 1989-1999 U.S. - Germany 0.104 1989-1999 Germany - U.S. 0.103 The Model ʀ Countries: foreign and home ʀ Continuum of goods: yi x 1 li x ai x x > 0,1@ ʀ Stand-in consumer in each country with labor Li . ʀ Preferences: 1 U ³ log >c x @ dx i 0 ʀ ad valorem tariffs: W i Determination of Exports ʀ x is exported by foreign if ah x ! w f a f x 1 W h ah x ! w f 1 W h a f x ʀ x is exported by home if wf ah x a f x 1W f ʀ x is not traded if wf ah x ! ! w f 1 W f 1W h a f x Dornbusch, Fisher, Samuelson (1977) ʀ Order goods according to the relative unit costs. ah x a f x w f 1 W h wf 1W f Foreign Exports Not Traded Home Exports x Dornbusch, Fisher, Samuelson (1977) ʀ Order goods according to the relative unit costs. ʀ Problems ɿ Trade data is collected in aggregates. ɿ Difficult to obtain data on relative unit costs. ɿ Both countries may export the same aggregate. Our Approach ʀ SITC ordering: an aggregate is an interval in > 0,1@ ʀ Take J evenly spaced points in > 0,1@ . ʀ Randomly assign log-productivities. Dj ª ah j º log « » ¬a f j ¼ D j u > D ,D @ ʀ Points not on the grid are filled in by linear interpolation. Relative Productivity Curve ʀ Steeper segments ɿ less trade growth ɿ more intra-industry trade ʀ For a given J larger D imply steeper segments. ʀ For a given D , larger J imply steeper segments. ah x a f x ah x a f x Low D High D 1 W h w f wf f 1 W x f 1 W wf 1 W h w f x Model Solution 1. Choose J and D . 2. Draw a realization of the relative productivity curve. 3. Solve the model and compute extensive margin measures. 4. Repeat for 5000 simulations. 5. Calculate means over simulations. Calibration ʀ Parameters: D , J , L f Lh , SITC endpoints ʀ Country size is measured by gross output of commodities. ʀ Codes are ordered by their SITC number. ʀ Code size is determined by its world export value. sizek MEX US EX WORLD EX ,k WORLD , k MEX US EX EX ¦ WORLD,k ¦ WORLD,k k k ʀ J and D determined by aggregate trade growth and Intra-industry trade Composition of Exports: Mexico to U.S. 1989-1999 By Sets if Categories Based on Export Size Two Standard Deviations Fraction of 1999 Exports 0.28 0.21 Data 0.14 Model 0.07 0.00 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Cummulative Fraction of 1989 Exports 0.9 1.0 Composition of Exports: Mexico to Canada 1989-1999 By Sets of Categories Based on Export Size 0.7 Two Standard Deviations Fraction of 1999 Exports 0.6 0.5 0.4 0.3 Data 0.2 Model 0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Cummulative Fraction of 1989 Exports 0.9 1.0 Model with Intensive and Extensive Margins ʀ Same Environment ʀ New Preferences 1 U ª i º U « ³ c x dx » ¬0 ¼ 1 U V 1 1 U ʀ Expenditure on Goods ɿ Old Model ci x pi x wi Li c x p x i ɿ New Model i 1V § p x · w L ¨¨ ¸¸ i © P ¹ i i i Composition of Exports: Mexico to U.S. 1989-1999 By Sets of Categories Based on Export Size Fraction of 1999 Exports 0.28 Two Standard Deviations 0.21 Data 0.14 Model 0.07 0.00 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Cummulative Fraction of 1989 Exports 0.9 1.0 Composition of Exports: Mexico to Canada 1989-1999 By Sets of Categories Based on Export Size Fraction of 1999 Exports 0.60 Two Standard Deviations 0.50 0.40 0.30 Data 0.20 Model 0.10 0.00 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Cummulative Fraction of 1989 Exports 0.9 1.0 Conclusions 1. The extensive margin is important. ɿ Average increase in export share: 67% ɿ Correct timing 2. Simple model can produce extensive margin growth. ɿ Calibration uses aggregate production data. Relative Productivity Parameters Growth in Trade’s Share of Production and Grubel-Lloyd Index J D ʀ Grubel-Lloyd Index GLUS MEX 1 ¦ ¦ kSITC US MEX EX MEX EX US MEX US ª EX EX US MEX º ¼ kSITC ¬ Calibration Values Grubel-Lloyd Index (1989) MEX-US MEX-CAN MEX-US MEX-CAN .487 .147 Growth in Trade/Production (1989-1999) 201% 299% D J .223 .208 3215 63 Relative Output (1989) .06 .66 Lf Lh .06 .66 Calibration Sensitivity ʀ Ideal SITC Measure: sizek ykh ykf f h y y ¦ k ¦ k k k ʀ Our Proxy: sizek EX wh ,k EX wf ,k h f EX EX ¦ w, k ¦ w, k k k K. J. Ruhl, “Solving the Elasticity Puzzle in International Economics,” University of Texas at Austin, 2005. The “Armington” Elasticity x Elasticity of substitution between domestic and foreign goods x Crucial elasticity in international economic models x International Real Business Cycle (IRBC) models: ż Terms of trade volatility ż Net exports and terms of trade co-movements x Applied General Equilibrium (AGE) Trade models: ż Trade response to tariff changes The Elasticity Puzzle x Time series (Business Cycles): ż Estimates are low ż Relative prices volatile ż Quantities less volatile x Panel studies (Trade agreement): ż Estimates are high ż Small change in tariffs (prices) ż Large change in quantities Time Series Estimates: Low Elasticity (1.5) Study Reinert and Roland Holst (1992) Range >0.1, 3.5@ Reinert and Shiells (1993) >0.1, 1.5@ >0.2, 4.9@ Gallaway et al. (2003) Trade Liberalization Estimates: High Elasticity (9.0) Study Clausing (2001) Range >8.9, 11.0@ Head and Reis (2001) >7.9, 11.4@ > 4.0, 13.0@ Romalis (2002) Why do the Estimates Differ? x Time series – no liberalization: ż Change in trade volume from goods already traded ż Change mostly on the intensive margin x Trade liberalization: ż Change in intensive margin plus ż New types of goods being traded ż Change on the extensive margin Modeling the Extensive Margin x Model: extensive margin from export entry costs x Empirical evidence of entry costs ż Roberts and Tybout (1997) ż Bernard and Wagner (2001) ż Bernard and Jensen (2003) ż Bernard, Jensen and Schott (2003) The Effects of Entry Costs x Business cycle shocks: ż Small extensive margin effect x Trade liberalization: ż Big extensive margin effect x Asymmetry creates different empirical elasticities Model Overview x Two countries: ^h, f `, with labor L x Infinitely lived consumers x No international borrowing/lending x Continuum of traded goods plants in each country ż Differentiated goods ż Monopolistic competitors ż Heterogeneous productivity x Export entry costs ż Differs across plants: second source of heterogeneity x Non-traded good, competitive market: A x Tariff on traded goods (iceberg): W Uncertainty x At date t , + possible events, Kt 1,..., + x Each event is associated with a vector of productivity shocks: zt ª¬ zh Kt , z f Kt º¼ x First-order Markov process with transition matrix / OKK c pr Kt 1 K c Kt K Traded Good Plants x Traded good technology: y I , N zI l x Plant heterogeneity I , N ż constant, idiosyncratic productivity: I ż export entry cost: N ż plant of type I , N x Q plants born each period with distribution F I , N x Fraction G of plants exogenously die each period Timing P hx I , N : plants of type I ,N who paid entry cost P hd I ,N : plants of type I , N who have not paid entry cost P P P hd P hx hd , P hx , P fd , P fx Shock/ Production Birth/ Death Shock/ Production Death Stay: non-exporter P hd c Switch: exporter Stay: exporter P hxc Consumers max J log C 1 J log A h h q ,ch L ,c f L s.t. 1 C ³ L, hh P ª ºU U U h h « ch L dL c f L dL » « h » h L, f P ¬L,h P ¼ ³ phh L chh L dL ³ ³ L, hf P 1 W p hf L c hf L dL phA A L 3h Non-traded Good max phA K , P A l s.t. A zh K l Normalize wh 1, implying phA K , P zh K Traded Goods: Static Profit Maximization S d phh , l ;I , N ,K , P max phh z K I l l phh , l s.t. chh phh ;K , P z K I l S x phf , l ;I , N ,K , P max phf z K I l l phf , l s.t. chf phf ;K , P z K I l Pricing rules: p I , N ,K , P h h p I , N ,K , P f h 1 UI z K Dynamic Choice: Export or Sell Domestically x Exporter’s Value Function: Vx I , N ,K , P d K , P S d I , N ,K , P S x I , N ,K , P 1 G E ¦Vx I , N ,K c, P c OKK c s.t. P c=0 K ,P x d K , P = multiplier on budget constraint Kc x Non-exporter’s Value Function: Vd I , N ,K , P max ^S d I , N ,K , P d K , P E 1 G ¦Vd I , N ,K c, P c OKK c , Kc ½ ¬ªS d I , N ,K , P N ¼º d K , P E 1 G ¦Vx I , N ,K c, P c OKK c ¾ Kc ¿ s.t. P c 0 K , P Equilibrium x Cutoff level of productivity for each value of the entry cost x For a plant of type I , N If I t IˆN K , P export and sell domestically If I IˆN K , P only sell domestically x In Equilibrium ż “Low” productivity/“high” entry cost plants sell domestic ż “High” productivity/“low” entry cost plants also export ż Similar to Melitz (2003) Determining Cutoffs x For the cutoff plant: ż entry cost = discounted, expected value of exporting x IˆN K , P is the level of productivity, I , that solves: d K , P N entry cost ª º 1 G E «¦Vx I , N ,K c, P c OKKc ¦Vd I , N ,K c, P c OKKc » Kc ¬ Kc ¼ expected value of exporting Costs and Benefits Finding the Cutoff Producer Value of Exporting: Steady State Entry Cost Non-Exporters Iˆss Firm Productivity Exporters I Choosing Parameters x Set V 1 1 U 2 and W 0.15 x Calibrate to the United States (1987) and a symmetric partner. E Parameters Annual real interest rate (4%) J Share of manufactures in GDP (18%) G Annual loss of jobs from plant deaths as percentage of employment (Davis et. al., 1996) (6%) Other Parameters x Distribution over new plants: FN I 1 I TI FI N 1 TN N N x N , I , Q , TI ,TN jointly determine: ż Average plant size (12 employees) ż Standard deviation of plant sizes (892) ż Average exporting plant size (15 employees) ż Standard deviation of exporting plant sizes (912) ż Fraction of production that is exported (9%) Plant Size Distribution: All Plants Plant Size Distribution: Exporting Plants 1.0 1.0 Data 0.8 0.8 Model 0.6 0.6 0.4 0.4 Data Model 0.2 0.2 0.0 0.0 1 10 100 Employees per Plant (log scale) 1000 10 100 Employees per Plant (log scale) 1000 Productivity Process x Two shocks, low and high: zi 1 H zi 1 H x Countries have symmetric processes with Markov Matrix /i ª O 1 O º « » O ¼ ¬1 O x H : standard deviation of the U.S. Solow Residuals (1.0%) x O : autocorrelation of the U.S. Solow Residuals (0.90) How does Trade Liberalization Differ from Business Cycles? x Trade liberalization ż Permanent changes ż Large magnitudes x Business cycles ż Persistent, but not permanent changes ż Small magnitudes Developing Intuition: Persistent vs. Permanent Shocks x1% positive productivity shock in foreign country ż Shock is persistent – autocorrelation of 0.90 x 1% decrease in tariffs ż Change in tariffs is permanent Costs and Benefits Response to 1% Productivity Shock Autocorrelation = 0.90 Value of Exporting: 1% Productivity Shock Value of Exporting: Steady State Entry Cost Iˆbc Iˆss Firm Productivity I Response to a 1% Foreign Productivity Shock Increase in imports on intensive margin = 1.89% Increase in imports on extensive margin = 0.16% Total increase in imports = 2.05% Change in consumption of home goods = -0.10% % Change Imports/Dom. Cons. % Change Price 2.17 0.99 2.19 Response to 1% Permanent Decrease in Tariffs Costs and Benefits Value of Exporting: 1% Decrease in Tariffs Value of Exporting: 1% Productivity Shock Value of Exporting: Steady State Entry Cost Iˆtar Iˆbc Iˆss Firm Productivity I Response to a 1% Tariff Reduction Increase in imports on intensive margin = 1.42% Increase in imports on extensive margin = 3.04% Total increase in imports = 4.46% Change in consumption of home goods = -0.33% % Change Imports/Dom. Cons. % Change Tariff 4.81 1.00 4.81 Quantitative Results x Two experiments x Trade liberalization ż Eliminate 15% tariff ż Compute elasticity across tariff regimes x Time series regressions ż Use model to generate simulated data ż Estimate elasticity as in the literature Trade Liberalization Elasticity Variable Entry Costs No Entry Costs (% change) (% change) Exports 87.1 30.5 Imports Dom. Cons. 93.0 32.2 Exporting Plants 37.7 0.0 Implied Elasticity 6.2 2.1 Elasticity in the Time Series x Simulate: produce price/quantity time series x Regress: log C f ,t / Ch ,t D V log ph ,t / p f ,t H t Parameter Estimate D -0.015 (6.36e-04) (standard error) (standard error) 1.39 (0.06) R- squared 0.30 V Conclusion x Gap between dynamic macro models and trade models ż Partially closes the gap ż Modeling firm behavior as motivated by the data ż Step towards better modeling of trade policy x Single model can account for the elasticity puzzle ż Time series elasticity of 1.4 ż Trade liberalization elasticity of 6.2 4. Modeling the fixed costs may explain why real exchange rate data indicate that more arbitrage across countries that have a strong bilateral trade relationship. C. M. Betts and T. J. Kehoe, “Real Exchange Rate Movements and the Relative Price of Nontraded Goods,” University of Minnesota and University of Southern California, 2003. C. M. Betts and T. J. Kehoe, “U.S. Real Exchange Rate Fluctuations and Relative Price Fluctuations,” University of Minnesota and University of Southern California, 2003. Argentina-U.S. Real Exchange Rate 1.00 0.80 rer log(RER) 0.60 0.40 0.20 0.00 rerN -0.20 -0.40 1986 1988 1990 1992 1994 year 1996 1998 2000 2002 U.S. BILATERAL REAL EXCHANGE RATES AND RELATIVE PRICES OF NONTRADED GOODS Germany-U.S. Real Exchange Rate Monthly CPI/PPI Canada-U.S. Real Exchange Rate Monthly CPI/PPI 0.50 0.50 0.40 rer 0.30 0.30 0.20 0.20 log(RER) log(RER) 0.40 0.10 0.00 -0.10 rerN 0.10 -0.10 -0.20 -0.20 -0.30 1980 -0.30 1980 1985 rerN 0.00 1990 year 1995 2000 rer 1985 1990 year 1995 2000 We investigate the empirical relation between the U.S. bilateral real exchange rate with 5 of her largest trade partners and the bilateral relative price of nontraded to traded goods. We measure the relation by simple correlation (comovements) relative standard deviation (volatility) variance decompositions (percent of fluctuations) We find the relation depends crucially on 1. the price series used to measure the relative price of nontraded goods 2. the choice of trade partner. Specifically, the relation is stronger when 1. traded goods prices are measured using production site prices — not final consumption price data 2. the trade intensity between the U.S. and a trade partner is larger — links international relative price behavior and size of trade flows Traditional theory real exchange rate theory (Cassel, Pigou, and many others) dichotomy of goods into x costlessly traded (arbitraged prices) x entirely nontraded (domestic prices) INAPPROPRIATE PLAN OF DISCUSSION x Methodology x Data x Results x What We Learn x Extended Results (paper 2) x Theoretical Model (paper 3) x Conclusions METHODOLOGY Traditional real exchange rate theory attributes all (or most) real exchange rate movements to changes in the relative price of nontraded goods. Recent empirical analyses using variance decompositions argue there is almost no role for relative price of nontraded goods in real exchange rate movements. All movements are driven by deviations from law of one price. REAL EXCHANGE RATE MEASUREMENT Bilateral real exchange rate between the United States and country i : RERi ,us NERi ,us Pus Pi NERi ,us : nominal exchange rate — country i currency units per U.S. dollar Pj : price deflator or index for the basket of goods consumed or produced in country j , j us, i . Traditional theory dichotomizes goods into x costlessly traded (arbitraged prices) x entirely nontraded (domestic prices) Pj ( PjT , PjN ) Decompose RERi ,us § PusT ·§ Pi T ¨ NERi ,us T ¸¨ Pi ¹ © Pi © PusT · Pus ¸¹ RERi ,us RERiT,us u RERiN,us where RERiT,us NERi ,us PusT Pi T is the real exchange rate of traded goods — the component that measures deviations from the law of one price and T T § · § · P P N i us RERi ,us ¨ P ( PT , P N ) ¸ ¨ P ( PT , P N ) ¸ © i i i ¹ © us us us ¹ is what we refer to as the (bilateral) relative price of nontraded (to traded) goods. In logarithms, reri ,us reriT,us reri ,Nus . In the case where Pj P , P T j RERiN,us N j T Jj j N 1J j j P P , § · § · Pi T PusT ¨ P ( PT , P N ) ¸ ¨ P ( PT , P N ) ¸ © i i i ¹ © us us us ¹ § T P i ¨ ¨ PT J i P N 1J i i © i T 1J i § Pi · ¨ PN ¸ © i ¹ · ¸ ¸ ¹ § T P us ¨ ¨ PT J us P N 1J us us © us T us N us §P ¨P © 1J us · ¸ ¹ · ¸ ¸ ¹ SUMMARY STATISTICS Analyze the relation between reri ,us and reri ,Nus : 1. corr (reri ,us , reri ,Nus ) cov(reri ,us , reri ,Nus ) var (rer i ,us 2. N i ,us std (rer ) std (reri ,us ) N i ,us 1/ 2 )var (rer ) N i ,us . 1/ 2 § var (rer ) · ¨ ¸ var ( rer ) i ,us ¹ © 3. vardec(reri ,us , reri ,Nus ) var (reri ,Nus ) var (reri ,Nus ) var (reriT,us ) (Another possibility: vardec 2 (reri ,us , reri ,Nus ) var (reri ,Nus ) cov(reri ,Nus , reriT,us ) var (reri ,us ) .) DATA Data on 5 of largest trade partners of the United States: x Canada (1) x Mexico (2) x Japan (3) x Germany (6) x Korea (7). These countries account for 53 percent of U.S. trade in 2000. Construct measures of reri ,us : Need aggregate price indices (and nominal exchange rates) 1. Deflator for Gross Output at production site for all goods and services produced by a country (GO) 2. Consumer Price Index for entire basket of consumption goods and services (CPI) 3. Personal Consumption Deflators for all personal consumption expenditures (PCD). (Another possibility: Deflator for Gross Domestic Product for all goods and services produced by a country (GDP).) Construct measures of reri ,Nus : Need traded goods price measures 1. Deflator for GO of agriculture, mining, and manufacturing 2. Producer price index for entire basket of producer goods (PPI). 3. CPI for “traded” components of consumption basket - all goods less food. 4. PCD for “traded” components of personal consumption expenditures - commodities. (Another possibility: Deflator for GDP of agriculture, mining, and manufacturing.) WHAT WE LEARN x The frequency of the data does not matter. x Detrending matters (theory should guide for the choice and the explanation for why it matters). x The size of bilateral trade relationship is crucial. The larger is the trade relationship, the more closely related are reri ,us and reri ,Nus . x The data series used to measure the relative price of nontraded goods reri ,Nus matters a lot. good conceptually sectoral GO deflators PPIs not good conceptually CPI components PCD components (sectoral GDP deflators) highly correlated CPI components-PCD components sectoral GOP deflators-PPIs (-sectoral GDP deflators) widely available PPIs less widely available difficult to obtain CPI components sectoral GO deflators PCD components (sectoral GDP deflators) SUGGESTION MODIFY THE TRADITIONAL THEORY SO THAT GOODS DIFFER BY DEGREE OF TRADABILITY TABLE 1 COMPARISON OF FREQUENCIES: CANADA-U.S. REAL EXCHANGE RATE PPI-CPI data 1980-2000 annual Levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Detrended levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) annual quarterly quarterly quarterly monthly 0.88 0.70 0.66 0.88 0.69 0.65 0.88 0.69 0.65 0.88 0.51 0.41 1 lag 4 lags 0.88 0.51 0.41 1 lag 4 lags 16 lags 0.87 0.51 0.41 1 lag (1 year) (4 years) (1 quarter) (1 year) (4 years) 0.70 0.55 0.40 0.82 0.55 0.51 0.56 0.51 0.28 0.70 0.55 0.39 0.82 0.55 0.51 monthly 3 lags (1 month) (1 quarter) 0.48 0.55 0.29 0.48 0.51 0.26 monthly monthly 12 lags 48 lags (1 year) (4 years) 0.67 0.55 0.37 0.82 0.55 0.50 log(RER) log(RER) log(RER) 0.30 0.20 0.10 0.00 -0.10 1985 1985 1985 rerN Monthly CPI/PPI rer 1990 year rerN Quarterly CPI/PPI rer 1990 year rerN Annual CPI/PPI rer 1990 1995 2000 2000 2000 1995 1995 CANADA-U.S. REAL EXCHANGE RATE -0.20 1980 0.30 0.20 0.10 0.00 -0.10 -0.20 1980 0.30 0.20 0.10 0.00 -0.10 -0.20 1980 year TABLE 2A CANADA-U.S. REAL EXCHANGE RATE Annual Data GO PPI-CPI Components Deflators of CPI 1980-1998 1980-2000 1980-2000 Levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Detrended levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 1 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 4 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Components of PCD 1980-2000 0.81 0.51 0.38 0.88 0.70 0.66 0.46 0.63 0.33 0.42 0.57 0.27 0.78 0.45 0.29 0.88 0.51 0.41 -0.43 0.17 0.02 -0.32 0.14 0.02 0.54 0.40 0.20 0.70 0.55 0.40 -0.07 0.20 0.09 -0.11 0.13 0.06 0.74 0.47 0.33 0.82 0.55 0.51 -0.19 0.16 0.12 -0.08 0.13 0.09 FIGURE 3A CANADA-U.S. REAL EXCHANGE RATE CPI / CPI components 0.40 0.40 0.20 0.20 log(RER) log(RER) GO deflators rerN 0.00 rerN 0.00 rer -0.20 -0.40 1980 rer -0.20 -0.40 1985 1990 year 1995 2000 1980 1985 1990 year 1995 2000 TABLE 2B GERMANY-U.S. REAL EXCHANGE RATE Annual Data GO PPI-CPI Components Deflators of CPI 1980-2000 1980-2000 1980-2000 Levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Detrended levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 1 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 4 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Components of PCD 1980-2000 -0.55 0.25 0.04 -0.33 0.73 0.21 -0.05 0.15 0.02 -0.24 0.25 0.05 0.18 0.20 0.04 -0.15 0.13 0.02 0.24 0.12 0.01 0.37 0.10 0.01 0.16 0.13 0.03 -0.24 0.14 0.04 0.18 0.10 0.01 -0.02 0.07 0.01 0.24 0.21 0.07 0.02 0.12 0.10 0.31 0.10 0.01 0.49 0.09 0.02 FIGURE 3B GERMANY-U.S. REAL EXCHANGE RATE GO deflators CPI / CPI components 0.40 0.40 rer rer 0.20 rerN log(RER) log(RER) 0.20 0.00 rerN 0.00 -0.20 -0.20 -0.40 1980 -0.40 1980 1985 1990 year 1995 2000 1985 1990 year 1995 2000 TABLE 2C JAPAN-U.S. REAL EXCHANGE RATE Annual Data GO PPI-CPI Components Deflators of CPI 1980-2000 1980-2000 1980-2000 Levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Detrended levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 1 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 4 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Components of PCD 1990-2000 -0.33 0.14 0.02 0.92 0.27 0.11 -0.74 0.17 0.02 -0.60 0.13 0.01 0.47 0.12 0.02 0.95 0.16 0.03 -0.27 0.07 0.00 0.35 0.07 0.00 0.30 0.12 0.02 0.87 0.17 0.05 -0.32 0.09 0.01 0.13 0.07 0.01 0.52 0.12 0.02 0.95 0.16 0.05 -0.36 0.06 0.01 0.43 0.07 0.01 FIGURE 3C JAPAN-U.S. REAL EXCHANGE RATE GO deflators CPI / CPI components 0.40 0.40 rer rer 0.20 log(RER) log(RER) 0.20 0.00 0.00 rerN rerN -0.20 -0.20 -0.40 1980 -0.40 1980 1985 1990 year 1995 2000 1985 1990 year 1995 2000 TABLE 2D KOREA-U.S. REAL EXCHANGE RATE Annual Data GO PPI-CPI Components Deflators of CPI 1980-2000 1980-2000 1980-2000 Levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Detrended levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 1 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 4 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Components of PCD 1980-2000 0.65 0.21 0.05 0.43 0.48 0.22 0.64 0.23 0.06 0.72 0.36 0.18 0.82 0.18 0.04 0.94 0.30 0.14 0.63 0.21 0.05 0.72 0.24 0.08 0.81 0.23 0.08 0.88 0.26 0.10 0.57 0.22 0.06 0.49 0.16 0.03 0.80 0.18 0.05 0.94 0.30 0.13 0.62 0.21 0.06 0.72 0.24 0.08 FIGURE 3D KOREA-U.S. REAL EXCHANGE RATE GO deflators CPI / CPI components 0.40 0.40 rer rerN 0.00 rerN 0.00 -0.20 -0.20 -0.40 1980 -0.40 1985 1990 year 1995 rer 0.20 log(RER) log(RER) 0.20 2000 1980 1985 1990 year 1995 2000 TABLE 2E MEXICO-U.S. REAL EXCHANGE RATE Annual Data GO PPI-CPI Components Deflators of CPI 1980-2000 1981-2000 1980-2000 Levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Detrended levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 1 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 4 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Components of PCD 1980-2000 0.75 0.36 0.18 0.74 0.21 0.06 0.64 0.55 0.33 0.81 0.23 0.08 0.84 0.36 0.20 0.73 0.22 0.06 0.67 0.46 0.26 0.84 0.24 0.08 0.52 0.25 0.07 0.54 0.19 0.04 0.26 0.28 0.08 0.51 0.16 0.03 0.91 0.38 0.25 0.78 0.24 0.08 0.73 0.51 0.34 0.92 0.27 0.12 FIGURE 3E MEX ICO-U.S. REAL EXCHANGE RATE GO deflators CPI / CPI components 0.40 0.40 rer rer 0.00 0.20 log(RER) log(RER) 0.20 rerN 0.00 rerN -0.20 -0.20 -0.40 1980 -0.40 1985 1990 year 1995 2000 1980 1985 1990 year 1995 2000 TABLE 3 COMPARISON OF SERIES: TRADE WEIGHTED AVERAGE Annual Data GO PPI-CPI Components Deflators of CPI Levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Detrended levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 1 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 4 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Components of PCD 0.44 0.36 0.21 0.73 0.48 0.33 0.23 0.45 0.22 0.27 0.36 0.15 0.68 0.32 0.18 0.77 0.32 0.20 0.00 0.22 0.08 0.23 0.15 0.03 0.47 0.27 0.11 0.63 0.33 0.19 0.02 0.19 0.06 0.14 0.13 0.04 0.70 0.34 0.21 0.78 0.34 0.25 0.10 0.23 0.14 0.37 0.16 0.08 TABLE 4 COMPARISON OF SERIES: CORRELATIONS OF DIFFERENT MEASURES OF rer N Annual Data Levels PPI-CPI-GO deflator CPI components-GO deflator PCD components-GO deflator PCD components-CPI components Detrended levels CPI-PPI/GO deflator CPI components/GO deflator PCD components/GO deflator PCD components/CPI components 1 year changes CPI-PPI/GO deflator CPI components/GO deflator PCD components/GO deflator PCD components/CPI components 4 year changes CPI-PPI/GO deflator CPI components/GO deflator PCD components/GO deflator PCD components/CPI components Canada Germany Japan Korea Mexico weighted average 0.97 0.52 0.54 0.996 0.92 0.91 0.99 0.88 -0.61 0.85 0.91 0.94 0.09 0.76 0.02 0.19 0.70 0.54 0.95 0.68 0.52 0.64 0.72 0.84 0.96 -0.18 -0.12 0.92 0.54 0.83 0.88 0.90 0.37 0.61 0.81 0.40 0.89 0.71 0.77 0.88 0.71 0.82 0.96 0.86 0.74 0.37 0.47 0.80 0.88 -0.19 -0.14 0.79 0.48 0.73 0.65 0.60 0.28 0.44 0.79 0.48 0.88 0.52 0.59 0.53 0.56 0.73 0.86 0.75 0.64 0.29 0.41 0.69 0.98 0.04 0.03 0.87 0.56 0.89 0.97 0.92 0.47 0.54 0.79 0.19 0.91 0.69 0.75 0.91 0.81 0.86 0.97 0.89 0.80 0.46 0.54 0.75 TABLE 5 COMPARISON OF COUNTRIES: GROSS OUTPUT DEFLATORS Annual Data Canada Germany Japan Korea Importance of trade to country i 2000 bilateral trade/GDP 0.58 0.05 0.04 0.14 2000 bilateral trade/trade 0.82 0.08 0.26 0.20 Rank of U.S. as partner 1 3 1 1 Importance of trade to U.S. 2000 bilateral trade/U.S. GDP 0.04 0.01 0.02 0.01 2000 bilateral trade/U.S. trade 0.21 0.04 0.11 0.03 Rank of country i as partner 1 6 3 7 Levels corr(rer,rer N ) 0.81 -0.55 -0.33 0.65 N std(rer ) / std(rer) N vardec(rer,rer ) Detrended levels corr(rer,rer N ) N std(rer ) / std(rer) N vardec(rer,rer ) 1 year changes corr(rer,rer N ) N std(rer ) / std(rer) N vardec(rer,rer ) 4 year changes corr(rer,rer N ) N std(rer ) / std(rer) N vardec(rer,rer ) Mexico 0.44 0.83 1 0.03 0.13 2 0.75 0.51 0.25 0.14 0.21 0.36 0.38 0.04 0.02 0.05 0.18 0.78 0.18 0.47 0.82 0.84 0.45 0.20 0.12 0.18 0.36 0.29 0.04 0.02 0.04 0.20 0.54 0.16 0.30 0.81 0.52 0.40 0.13 0.12 0.23 0.25 0.20 0.03 0.02 0.08 0.07 0.74 0.24 0.52 0.80 0.91 0.47 0.21 0.12 0.18 0.38 0.33 0.07 0.02 0.05 0.25 TABLE A GROSS DOMESTIC PRODUCT DEFLATORS Annual Data Levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Detrended levels corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 1 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) 4 year changes corr(rer,rer N ) std(rer N ) / std(rer) vardec(rer,rer N ) Canada 19801998 Germany 19802000 Japan 19802000 Korea 19802000 Mexico 19802000 weighted 0.80 0.90 0.69 -0.32 0.53 0.15 -0.69 0.22 0.03 0.00 0.26 0.06 0.74 0.50 0.33 0.34 0.59 0.38 0.63 0.75 0.48 0.19 0.27 0.07 -0.24 0.18 0.03 0.56 0.18 0.04 0.84 0.47 0.33 0.47 0.49 0.29 0.05 0.81 0.33 -0.14 0.18 0.05 -0.29 0.16 0.03 0.61 0.26 0.09 0.47 0.36 0.15 0.11 0.48 0.18 0.65 0.68 0.54 0.18 0.27 0.12 -0.16 0.16 0.03 0.56 0.19 0.06 0.91 0.47 0.40 0.51 0.46 0.34 average TABLE B COMPARISON OF GROSS DOMESTIC PRODUCT DEFLATORS AND GROSS OUTPUT DEFLATORS Annual Data Canada Germany Japan Korea Mexico 1980-1998 1980-2000 1980-2000 1980-2000 1980-2000 weighted average Levels corr(rer(GDP),rer(GO)) corr(rer N (GDP),rer N (GO)) std(rer(GDP)) / std(rer(GO)) std(rer N (GDP)) / std(rer N (GO)) 0.99 0.96 1.27 2.23 0.99 0.97 1.11 1.68 0.997 0.89 1.10 1.74 0.94 0.86 1.34 1.66 0.98 0.97 1.09 1.52 0.99 0.94 1.18 1.87 0.98 0.88 1.07 1.77 0.88 0.80 0.94 1.51 0.995 0.69 1.06 1.57 0.99 0.86 1.22 1.22 0.99 0.98 1.11 1.45 0.98 0.86 1.08 1.59 0.99 0.87 1.04 2.13 0.99 0.48 1.05 1.50 0.99 0.57 1.05 1.49 0.99 0.76 1.19 1.35 0.99 0.92 1.09 1.58 0.99 0.78 1.06 1.76 0.99 0.96 1.01 1.47 0.998 0.85 1.07 1.75 0.997 0.71 1.06 1.49 0.99 0.90 1.20 1.28 0.99 0.99 1.14 1.42 0.99 0.91 1.07 1.47 Detrended levels corr(rer(GDP),rer(GO)) corr(rer N (GDP),rer N (GO)) std(rer(GDP )) / std(rer(GO)) std(rer N (GDP)) / std(rer N (GO)) 1 year changes corr(rer(GDP),rer(GO)) corr(rer N (GDP),rer N (GO)) std(rer(GDP)) / std(rer(GO)) std(rer N (GDP)) / std(rer N (GO)) 4 year changes corr(rer(GDP),rer(GO)) corr(rer N (GDP),rer N (GO)) std(rer(GDP )) / std(rer(GO)) std(rer N (GDP)) / std(rer N (GO)) EXTENDED RESULTS Examine sample of 50 countries and all possible 1225 bilateral real exchange rates. Use same methodology and summary statistics and CPI-PPI measures of prices. Examine robustness of results to 1. Presence of U.S. in bilateral trade partner pairs in the sample. 2. Presence of rich-country/poor-country bilateral trade pairs in the sample. 3. Presence of high-inflation/low inflation bilateral trade pairs in the sample. Find that there is a substantive relation between reri ,us and reri ,Nus on average. The relation does not depend on these three factors (at least in the manner one might expect). Strength of the relation depends crucially on size of the trade relationship between two trade partners. Table I COUNTRIES IN THE SAMPLE Percent World Trade in 2000 Argentina Australia Austria Belgium Brazil Canada Chile Colombia Costa Rica Cyprus Denmark Egypt El Salvador Finland France Germany Greece 0.39 1.01 1.04 2.75 0.90 3.98 0.27 0.18 0.10 0.06 0.73 0.19 0.06 0.64 5.04 8.10 0.35 Hong Kong (P.R.C.) India Indonesia Ireland Israel Italy Japan Jordan Korea Luxembourg Malaysia Mexico Netherlands New Zealand Norway Pakistan Panama 3.02 0.70 0.75 1.00 0.50 3.65 6.45 0.04 2.50 0.16 1.42 2.44 3.61 0.20 0.70 0.15 0.13 Peru Philippines Saudi Arabia South Africa Singapore Spain Sri Lanka Sweden Switzerland Thailand Trinidad and Tobago Turkey Uruguay United Kingdom United States Venezuela 0.10 0.66 0.84 0.36 2.08 2.08 0.10 1.23 1.39 0.96 0.04 0.63 0.05 4.88 15.37 0.38 Table II U.S. BILATERAL REAL EXCHANGE RATES Weighted Means income level high low inflation high low 0.63 0.44 0.26 0.62 0.45 0.28 0.65 0.42 0.21 0.69 0.37 0.18 0.61 0.46 0.28 0.74 0.49 0.32 0.28 0.29 0.10 0.58 0.27 0.11 0.68 0.63 0.43 0.50 0.35 0.17 0.47 0.36 0.19 0.54 0.34 0.11 0.51 0.29 0.11 0.49 0.37 0.19 0.53 0.40 0.22 0.39 0.20 0.08 0.42 0.19 0.06 0.58 0.52 0.28 0.66 0.61 0.36 0.35 0.22 0.23 49 24 88.13 59.80 0.75 0.39 0.18 25 28.33 0.77 0.34 0.18 18 21.32 0.62 0.37 0.23 31 66.81 0.73 0.41 0.31 25 66.22 0.44 0.21 0.10 24 21.91 0.60 0.22 0.10 31 45.92 0.71 0.52 0.35 18 42.21 all levels corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 1 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 4 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) countries percent of U.S. trade trade intensity high low std(rer) high low Table III INCOME LEVELS ALL BILATERAL REAL EXCHANGE RATES Weighted Means levels corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 1 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 4 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) bilateral pairs percent of world trade all highhigh highlow lowlow 0.53 0.66 0.32 0.50 0.74 0.33 0.60 0.47 0.26 0.63 0.64 0.43 0.45 0.46 0.19 0.42 0.50 0.21 0.52 0.36 0.13 0.58 0.43 0.21 0.60 0.54 0.26 1225 71.88 0.57 0.59 0.27 300 49.80 0.65 0.40 0.19 625 19.78 0.70 0.51 0.33 300 2.30 Table IV INFLATION LEVELS ALL BILATERAL REAL EXCHANGE RATES Weighted Means levels corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 1 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 4 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) bilateral pairs percent of world trade all highhigh highlow lowlow 0.53 0.66 0.32 0.68 0.63 0.47 0.63 0.46 0.24 0.51 0.71 0.33 0.46 0.46 0.19 0.65 0.43 0.23 0.51 0.34 0.13 0.44 0.49 0.20 0.60 0.54 0.25 1225 71.88 0.76 0.53 0.39 153 0.81 0.69 0.40 0.20 576 13.11 0.58 0.57 0.26 496 57.96 Table V TRADE INTENSITY ALL BILATERAL REAL EXCHANGE RATES Weighted Means all levels corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 1 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 4 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) bilateral pairs percent of world trade trade intensity high low 0.53 0.66 0.32 0.62 0.71 0.36 0.46 0.62 0.28 0.46 0.46 0.19 0.49 0.56 0.24 0.42 0.37 0.14 0.60 0.54 0.25 1225 71.88 0.66 0.61 0.30 51 33.51 0.54 0.48 0.21 1174 38.37 Table VI REAL EXCHANGE RATE VARIABILITY ALL BILATERAL REAL EXCHANGE RATES Weighted Means std(rer) levels corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 1 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 4 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) bilateral pairs percent of world trade all high low 0.53 0.66 0.32 0.59 0.36 0.20 0.50 0.87 0.40 0.46 0.46 0.19 0.46 0.26 0.09 0.45 0.60 0.26 0.60 0.54 0.25 1225 71.88 0.60 0.29 0.13 863 29.55 0.60 0.71 0.33 362 42.33 Table VII U.S. BILATERAL REAL EXCHANGE RATES TRADING BLOC TRADE PARTNERS Weighted Means levels corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 1 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 4 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) countries percent of U.S. trade all EU nonEU NAFTA non-NAFTA nonNAFTA-nonEU 0.63 0.44 0.26 0.23 0.27 0.08 0.74 0.49 0.31 0.82 0.45 0.35 0.50 0.44 0.20 0.65 0.53 0.27 0.50 0.35 0.17 0.41 0.17 0.04 0.52 0.40 0.20 0.55 0.42 0.22 0.46 0.30 0.13 0.49 0.38 0.18 0.66 0.36 0.22 49 88.13 0.43 0.18 0.05 14 19.19 0.72 0.41 0.26 35 68.94 0.77 0.46 0.30 2 34.31 0.58 0.30 0.16 47 53.82 0.66 0.37 0.23 33 34.63 Table VIII TRADING BLOC TRADE PARTNERS ALL BILATERAL REAL EXCHANGE RATES Weighted Means levels corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 1 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) 4 year lags corr(rer, rerN) std(rerN)/std(rer) vardec(rer, rerN) bilateral pairs percent of world trade EU& NAFTA& nonEU/ nonEU/ NAFTA& nonEU/ nonNAFTA& nonNAFTA NAFTA nonNAFTA nonEU/ nonNAFTA all EU& EU EU& NAFTA 0.53 0.66 0.32 0.40 1.05 0.40 0.25 0.28 0.09 0.47 0.59 0.29 0.82 0.44 0.34 0.65 0.53 0.27 0.61 0.56 0.36 0.46 0.46 0.19 0.43 0.70 0.27 0.42 0.17 0.04 0.37 0.34 0.13 0.55 0.42 0.22 0.50 0.39 0.19 0.47 0.40 0.17 0.60 0.54 0.25 1225 71.88 0.63 0.83 0.35 91 21.44 0.43 0.21 0.06 42 6.77 0.46 0.49 0.20 462 11.27 0.77 0.45 0.30 3 10.62 0.66 0.37 0.23 99 11.64 0.46 0.26 0.22 528 10.25 5. The major determinant of large macroeconomic fluctuations in Mexico is fluctuations in total factor productivity. If trade liberalization and the 1995 financial crisis have had large macroeconomic effects, it is through their effects on TFP. R. Bergoeing, P. J. Kehoe, T. J. Kehoe, and R. Soto “A Decade Lost and Found: Mexico and Chile in the 1980s,” Review of Economic Dynamics, 5 (2002), 166-205. T. J. Kehoe and E. C. Prescott, “Great Depressions of the Twentieth Century,” Review of Economic Dynamics, 5 (2002), 118. T. J. Kehoe and K. J. Ruhl, “Sudden Stops, Sectoral Reallocations, and Productivity Drops,” University of Minnesota, 2006. Grandes Depresiones Timothy J. Kehoe University of Minnesota y Federal Reserve Bank of Minneapolis Edward C. Prescott Federal Reserve Bank of Minneapolis y Arizona State University Instituto Tecnológico Autónomo de México Abril de 2006 www.econ.umn.edu/~tkehoe Estados Unidos: PIB real por persona en edad de trabajar índice (1900=100) 800 400 200 100 1900 1910 1920 1930 1940 1950 año 1960 1970 1980 1990 2000 Canadá: PIB real por persona en edad de trabajar índice (1900=100) 15.66 800 14.66 400 200 13.66 100 12.66 1900 1910 1920 1930 1940 1950 año 1960 1970 1980 1990 2000 Italia: PIB real por persona en edad de trabajar 1600 índice (1900=100) 800 400 200 100 1900 1910 1920 1930 1940 1950 año 1960 1970 1980 1990 2000 Nueva Zelanda: PIB real por person en edad de trabajar 800 índice (1900=100) 400 200 ? 100 50 1900 1910 1920 1930 1940 1950 año 1960 1970 1980 1990 2000 España: PIB real por persona en edad de trabajar 11.68 800 índice (1900=100) 400 10.68 200 9.68 100 8.68 50 7.68 1900 1910 1920 1930 1940 1950 año 1960 1970 1980 1990 2000 Chile: PIB real por persona en edad de trabajar 9.308 800 índice (1900=100) 400 8.308 7.308 200 100 6.308 50 5.308 1900 1910 1920 1930 1940 1950 año 1960 1970 1980 1990 2000 México: PIB real por persona en edad de trabajar 800 9.64 índice (1900=100) 400 8.64 200 7.64 100 6.64 50 5.64 1900 1910 1920 1930 1940 1950 año 1960 1970 1980 1990 2000 Grandes Depresiones del Siglo XX Uso de contabilidad de crecimiento y de modelos de equilibrio general aplicados para reexaminar episodios de grandes depresiones: Reino Unido (1920s y 1930s) — Cole y Ohanian Canadá (1930s) — Amaral y MacGee Francia (1930s) — Beaudry y Portier Alemania (1930s) — Fisher y Hornstein Italia (1930s) — Perri y Quadrini Argentina (1970s y 1980s) — Kydland y Zarazaga Chile y México (1980s) — Bergoeing, Kehoe, Kehoe y Soto Japón (1990s) — Hayashi y Prescott (Review of Economic Dynamics, enero 2002; versión revisada y expandida por aparecer en un volumen de la Reserva Federal de Minneapolis) Grandes depresiones: metodología Cole y Ohanian (1999), Kehoe y Prescott (2002) Función de producción agregada: Yt = At Kta L1t-a . Cuando At = A0 g (1-a )t , el producto per cápita crece a la tasa constante g - 1. El crecimiento del producto se mide con respecto a su tendencia. • El crecimiento de tendencia representa el crecimiento – a una tasa suave en el tiempo – del conocimiento utilizable para la producción. • Este conocimiento no es específico a los países. • Los países crecen a la misma tasa, g - 1, en trayectorias equilibradas de crecimiento diferentes. • Los niveles difieren entre países porque las instituciones son diferentes. • Cambios institucionales mueven a los países a trayectorias de crecimiento equilibradas diferentes. • Adoptamos g - 1 como la tasa de crecimiento del líder industrial – Estados Unidos: g = 1,02 Contabilidad de crecimiento Yt : PIB real (cuentas nacionales) X t : inversión real (cuentas nacionales) Lt : horas trabajadas (encuestas laborales) Construcción del stock de capital: Kt +1 = (1 - d ) K t + X t La productividad total de factores es el residuo: At = Yt K ta L1t-a d = 0,05 a = 0,30 Descomponiendo los cambios en el PIB por persona en edad de trabajar § Yt · § Kt · § Lt · 1 α log ¨ ¸ = log ( At ) + log ¨ ¸ + log ¨ ¸ 1−α © Nt ¹ 1 − α © Yt ¹ © Nt ¹ Las teorías de depresión tradicionales enfatizan la caída en el stock de capital o en las horas trabajadas como los factores más importantes para explicar las depresiones. Contabilidad de crecimiento en los Estados Unidos 400 350 índice (1960=100) 300 250 Yt / N t 1 1D t 200 A Lt / N t 150 100 D 1D Kt / Yt 50 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 Contabilidad de crecimiento en España 400 1 1D t 350 A índice (1960=100) 300 Yt / N t 250 200 D 1D Kt / Yt 150 100 Lt / N t 50 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 México y Chile durante los años 80s R. Bergoeing, P. J. Kehoe, T. J. Kehoe y R. Soto Crisis similares durante 1981-1983 x más severa en Chile que en México Recuperaciones diferentes x más rápida en Chile que en México ¿Por qué patrones diferentes? PIB real por persona en edad de trabajar (15-64) sin tendencia de 2 por ciento por año 130 índice (1980=100) 120 Chile 110 100 90 80 México 70 60 1980 1982 1984 1986 1988 1990 año 1992 1994 1996 1998 2000 Crisis similares Condiciones iniciales: · abultada deuda externa · apreciación del tipo de cambio real · abultado déficit comercial · problemas bancarios Shocks: · aumento en las tasas de interés externas · caída en el precio del cobre y del petróleo · falta de acceso a créditos externos Teorías para recuperaciones diferentes Teoría de Corbo y Fischer para la rápida recuperación de Chile • La fuerte depreciación del tipo de cambio real y la caída en los salarios reales generó crecimiento liderado por exportaciones. Teoría de Sachs para la lenta recuperación de México • La deuda “pendiente” deprimió la inversión. Teoría de las reformas estructurales • Las reformas estructurales que implementó Chile durante los años 70s se realizaron en México durante los años 80s o 90s. Teoría de Corbo y Fischer para la rápida recuperación de Chile La fuerte depreciación del tipo de cambio real y la caída en los salarios reales generó crecimiento liderado por exportaciones. ¿Pero qué ocurrió en México? Indice de salarios reales en manufacturas 150 140 índice (1980=100) 130 Chile 120 110 100 90 80 México 70 60 1980 1982 1984 1986 1988 1990 año 1992 1994 1996 1998 2000 Valor de las exportaciones deflactadas por IPP de EEUU 700 índice (1980=100) 600 500 México 400 300 200 Chile 100 0 1980 1982 1984 1986 1988 1990 año 1992 1994 1996 1998 2000 Teoría de Sachs para la lenta recuperación de México Abultada deuda “pendiente” en México: · La mayoría de los nuevos créditos eran necesarios para pagar viejos créditos. · Inversiones socialmente beneficiosas no se realizaban. ¿Pero qué ocurrió en Chile? Deuda externa total como porcentaje del PIB 140 porcentaje del PIB 120 100 Chile 80 60 México 40 20 1980 1982 1984 1986 1988 1990 año 1992 1994 1996 1998 2000 Inversión como porcentaje del PIB 30 porcentaje del PIB 25 México 20 15 Chile 10 5 1980 1982 1984 1986 1988 1990 año 1992 1994 1996 1998 2000 Teoría de las reformas estructurales En 1979 Chile había privatizado y reformado su sistema tributario, su sistema bancario, su ley de quiebras y sus políticas de comercio exterior. México esperó hasta más tarde. Recuperaciones diferentes: · Chile aprovechó los beneficios de las reformas. · México pagó los costos de las distorsiones. ¿Cómo podemos determinar qué reformas fueron cruciales? · Afectaron las reformas la acumulación de insumos o su eficiencia? · ¿Cuál fue el momento de las reformas? Contabilidad de crecimiento Yt : PIB real (cuentas nacionales) X t : inversión real (cuentas nacionales) Lt : horas trabajadas (encuestas laborales) Construcción del stock de capital: Kt +1 = (1 - d ) K t + X t La productividad total de factores es el residuo: At = Yt K ta L1t-a d = 0,05 a = 0,30 Descomponiendo los cambios en el PIB por persona en edad de trabajar § Yt · § Kt · § Lt · 1 α log ¨ ¸ = log ( At ) + log ¨ ¸ + log ¨ ¸ 1−α © Nt ¹ 1 − α © Yt ¹ © Nt ¹ Las teorías de depresión tradicionales enfatizan la caída en el stock de capital o en las horas trabajadas como los factores más importantes para explicar las depresiones. Modelo de equilibrio general aplicado dinámico El consumidor representativo maximiza ¦ f t 1980 E t ª«J log Ct (1J )log(hNt Lt ) º» ¬« ¼» sujeto a Ct Kt 1 Kt donde Tt wt Lt (1 W t )(rt G ) Kt Tt W t (rt G ) Kt es la transferencia de ingreso tributario Factibilidad: Ct Kt 1 (1 G ) Kt At KtD L1t D . Calibración Condiciones de primer orden: 1 C t 1 E C 1 (1 W )(r G ) t t t J wt 1J . hNt Lt Ct Utilizar datos de 1960-1980 para calibrar los parámetros: E 0.98, W 1 J Ct E Ct 1 (rt G )Ct 1 W 0.45 en México, W 0.56 en Chile ; Ct J 0.30 en México, J 0.28 en Chile . Ct wt (hNt Lt ) Experimentos numéricos Caso básico: W t 0.45 en México, W t 0.56 en Chile , 1980-2000. Reforma tributaria: W t 0.45 en México, W t 0.56 en Chile , 1980-1988; W t 0.12 en México, W t 0.12 en Chile , 1988-2000. Productividad total de factores sin tendencia de 1,4 por ciento por año 130 índice (1980=100) 120 110 Chile 100 90 80 70 60 1980 México 1982 1984 1986 1988 1990 año 1992 1994 1996 1998 2000 Lecciones del proyecto Grandes Depresiones • Los determinantes principales de las depresiones no son caídas en los insumos de capital y trabajo — enfatizadas por la teorías tradicionales de depresiones — sino caídas en la eficiencia, medida como productividad total de factores (PTF), con que estos insumos son usados. • Shocks exógenos, como caídas en los términos de intercambio y aumentos en las tasas de interés externas que afectaron a Chile y México a comienzos de los años 80s, pueden causar una reducción en la actividad económica de la magnitud típicamente asociada al ciclo económico. • Política gubernamental equivocada puede transformar esa reducción en una caída severa y prolongada en la actividad económica bajo su tendencia — una gran depresión. Experimentos numéricos para México: PIB real por persona en edad de trabajar Caso base Reforma tributaria Y/N (sin tendencia) Y/N (sin tendencia) 120 120 100 100 80 80 modelo datos datos 60 40 1980 60 modelo 1985 1990 1995 40 2000 1980 1985 1990 1995 2000 Experimentos numéricos para Chile: PIB real por persona en edad de trabajar Caso base Reforma tributaria Y/N (sin tendencia) Y/N (sin tendencia) 140 140 120 120 modelo 100 100 datos 80 80 modelo 60 1980 1985 1990 1995 datos 60 2000 1980 1985 1990 1995 2000 ¿Qué aprendemos de la contabilidad de crecimiento y de los experimentos numéricos? Casi toda la diferencia en las recuperaciones de México y Chile es resultado de diferencias en las trayectorias de la productividad. Las reformas tributarias son importantes para explicar algunos rasgos de las recuperaciones, pero no sus diferencias. Implicancias para el estudio de la teoría de las reformas estructurales: · Las únicas reformas que podrían explicar la diferencia entre ambas recuperaciones son aquéllas que se traducen principalmente en diferencias en productividad, no las que lo hacen a través de diferencias en el uso de los insumos. · El momento de las reformas es crucial para que expliquen la diferencia en el comportamiento económico. Reformas fiscales Chile: • reformas tributarias: 1975, 1984 • reforma a la seguridad social: 1980 • excedentes fiscales México: • reformas tributarias 1980, 1985, 1987, 1989 • déficit fiscales ¡Importantes, pero no para explicar las diferencias! Reformas de política comercial Chile: a fines de los años 70s • todas las restricciones cuantitativas habían sido eliminadas • aranceles uniformes en 10 por ciento • aumento en los aranceles durante la crisis — aranceles bajo 10 por ciento en 1991 México: a mediados de los años 80s • 100 por ciento de la producción doméstica protegida por licencias de importación • barreras no arancelarias y tipos de cambio duales Reformas masivas al comercio exterior en México durante 1987-1994, culminando con el NAFTA ¡El momento parece incorrecto! Privatización Chile • principales privatizaciones durante 1974-1979 México • nacionalización masiva en 1982 ° expropiación de bancos de compañías privadas ° el gobierno controlaba entre 60 y 80 por ciento del PIB • privatizaciones principales a partir de 1989 ¿El momento parece incorrecto? Banca Chile: 1982 y después · gobierno se hizo cargo de los bancos que quebraron · tasas de interés determinadas en el mercado · reducción en los requerimientos de reservas. México: 1982 y después · nacionalización de todos los bancos · gobierno fijó tasas de interés para depósitos bajas · 75 por ciento de los créditos dirigidos al gobierno o por el gobierno. Crédito privado como porcentaje del PIB 90 80 Chile porcentaje del PIB 70 60 50 40 México 30 20 10 0 1980 1982 1984 1986 1988 1990 año 1992 1994 1996 1998 2000 Leyes de quiebra Chile había reformado la administración de sus procedimientos de quiebra en 1978. En 1982 reformó su ley de quiebra asimilándola más a la imperante en Estados Unidos. México reformó sus procedimientos de quiebra de manera similar sólo en el año 2000. Quiebras comerciales en Chile 500 número por año 400 300 200 100 0 1980 1982 1984 1986 1988 1990 año 1992 1994 1996 1998 2000 Sudden Stops, Sectoral Reallocations, and Real Exchange Rates Timothy J. Kehoe University of Minnesota and Federal Reserve Bank of Minneapolis and Kim J. Ruhl University of Texas, Austin Definition: sudden stops An abrupt decline in capital inflows. x Argentina, Mexico: 1994-1995 x Indonesia, Korea, Thailand: 1996-1998 x Germany, Sweden, Spain: 1992-1993 x Argentina 2000-2002 It is not speed that kills, it is the sudden stop. — Bankers’ Adage (Dornbusch, Goldfajn, and Valdés, 1995) Two lines of research What causes sudden stops? x large literature x take effects – loss of output – as given What are the effects of sudden stops? (our approach) x fewer studies x one sector models x take sudden stop as given Effects of sudden stops Aggregate x real exchange rate depreciation x trade balance surplus x decrease in GDP x decrease in TFP Sectoral Detail x tradable good output falls less than nontradable x labor reallocated to tradable good sector x increase in pT p N Example: Mexico 1994-1995 Opens to capital flows: late 1980s x trade deficits x real exchange rate appreciation Sudden stop: 1994-1995 x trade surplus x real exchange rate depreciation x fall in GDP, TFP End of sudden stop x trade deficits x real exchange rate appreciation x recovery of GDP, TFP Our model Small open economy x multisector: tradable, nontradable x costly to adjust labor across sectors Sudden stop x tradable goods price increase, increase production x capital and labor misallocated x costs from moving labor Model accounts for: x real exchange rate x labor allocation x trade balance Misses: x GDP, TFP Outline 1. Data: Mexico 2. Explanations 3. Model 4. Calibration 5. Results Mexico: trade balance 4.0 percent of GDP (%) 2.0 0.0 -2.0 -4.0 -6.0 1988 1990 1992 1994 1996 1998 2000 2002 Mexico: interest rates (dollar denominated debt) 25.0 15.0 10.0 nominal rate (stripped) Dec 30, 1994 = 8.90% 5.0 Mexico premia Nov 30, 1994 = 4.56% 10/26/95 07/18/95 04/09/95 12/30/94 09/21/94 06/13/94 03/05/94 11/25/93 08/17/93 05/09/93 0.0 01/29/93 percent per year 20.0 Mexico-U.S. real exchange rate RERmex ,us NERmex ,us Pus Pmex x NERmex ,us : nominal exchange rate — pesos per dollar x Pj : GDP price deflator in country Decomposing real exchange rates RERmex ,us T § PusT ·§ Pmex PusT · / ¨ NERmex ,us T ¸¨ ¸ P P P mex ¹© mex us ¹ © T N RERmex u RER mex ,us ,us Deviations from the law of one price: T RERmex ,us NERmex ,us PusT T Pmex Relative price of nontradable to tradable goods: T T § · § · P P N mex us RERmex ,us ¨ ¸/¨ ¸ P P © mex ¹ © us ¹ Mexico-U.S. real exchange rate 0.3 log(RER) 0.2 rer 0.1 0.0 rer N -0.1 -0.2 1988 1990 1992 1994 1996 1998 2000 2002 Sectoral Reallocation Mexico: sectoral value added 140 130 index (1994=100) 120 tradable goods 110 100 nontradable goods 90 80 70 1988 1990 1992 1994 1996 1998 2000 2002 Mexico: traded good employment share of total employment 0.40 0.38 0.36 Data 0.34 0.32 Trend 0.30 1988 1990 1992 1994 1996 1998 2000 2002 Mexico: traded good employment, detrended 0.08 deviation from trend 0.06 0.04 0.02 0.00 -0.02 1988 1990 1992 1994 1996 1998 2000 2002 Growth accounting Yt : real GDP (national income accounts) X t : real investment (national income accounts) Lt : hours worked (labor surveys) Construct Capital Stocks: K t 1 1 G Kt X t Total factor productivity: At Yt K tD L1tD Mexico: output and TFP 110 index (1994 = 100) 105 Y/L 100 95 TFP 90 1988 1990 1992 1994 1996 1998 2000 2002 Two theories Output and TFP fall with a sudden stop because: 1. cost of imported intermediates increases 2. real frictions that waste/mismeasure output and inputs x adjustment frictions x variable capital utilization x labor hoarding 1. Cost of imported intermediates increases Mexico: terms of trade 115 index (1994 = 100) 110 105 100 95 1988 1990 1992 1994 1996 1998 2000 2002 1. Cost of imported intermediates increases A deterioration in the terms of trade makes it expensive for an economy to import intermediate goods. International trade as production technology x Exports are inputs, imports are outputs. x decline in terms of trade § negative technology shock. Can this negative “technology shock” account for the drop in TFP during the crisis? 1. Cost of imported intermediates increases A deterioration in the terms of trade makes it expensive for an economy to import intermediate goods. International trade as production technology x Exports are inputs, imports are outputs. x decline in terms of trade § negative technology shock. Can this negative “technology shock” account for the drop in TFP during the crisis? No. Standard national income accounting (SNA, NIPA) implies that terms of trade shocks have no first-order effects on real output (GDP, GNP) A model with international trade Final good f ( A , m) y Exports: x y cx Imported intermediate inputs: m, price p (terms of trade) Balanced trade pm x Real GDP (base year prices) Y c x p0 m y p0 m f (A, m) p0 m Competitive economy solves max f ( A, m) pm m f m ( A, m( p )) { p mc( pˆ ) 1 0 f mm ( A, m( pˆ )) How does real GDP change with an increase in p — a deterioration in the terms of trade (depreciation in the real exchange rate)? Y ( p ) { f ( A, m( p )) p0 m( p ) Y c( pˆ ) f m ( A, m( pˆ ))mc( pˆ ) p0 mc( pˆ ) ( pˆ p0 ) mc( pˆ ) pˆ | p0 Y c( pˆ ) | 0 Alternative accounting concepts x Diewert and Morrison (1974, 1986) x Kohli (1983, 1996) x U.S. Bureau of Economic Analysis (Command Basis GDP) x United Nations Statistics Division (Gross National Income) x GNP, GDP (SNA, NIPA) do not. Alternative accounting concepts x Diewert and Morrison (1974, 1986) x Kohli (1983, 1996) x U.S. Bureau of Economic Analysis (Command Basis GDP) x United Nations Statistics Division (Gross National Income) x GNP, GDP (SNA, NIPA) do not. Terms of trade shocks are worse than you think! 2. Real frictions waste outputs: our model Small open economy (Mexico) Produces nontradable goods, y N , and tradable goods, yD x use intermediates plus capital and labor Composite tradable yTt f yDt , mt Frictions: x sector specific capital x costly to move labor across sectors Quantitative model Consumers \ § · U U U ª §c · º § · ¸ c f t¨ Tt Nt max ¦ t 0 E ¨ «H ¨ ¸ 1 H ¨ ¸ » 1 ¸ /\ n © nt ¹ »¼ ¨ «¬ © t ¹ ¸ © ¹ s.t. pTt cTt pNt cNt qt iDt iNt bt 1 wt A t 1 rt bt rDt k Dt rNt k Nt Tt cTt t 0 , cNt t 0 , at t A b0 , k D 0 .k N 0 given Here A t is working-age population, nt 0.5A t 0.5 popt is adult-equivalent population, Production functions Domestically produced traded good D D 1D D yDt = min ª¬ zTDt / aTD , z NDt / aND , AD k Dt A Dt º¼ 4 D A Dt 1 , A Dt A Dt 1 where 4 D A Dt 1 , A Dt § A Dt A Dt 1 · = TD ¨ ¸ A Dt 1 © ¹ 2 Nontraded good D N 1D N yNt = min ª¬ zTNt / aTN , z NNt / aNN , AN k Nt A Nt º¼ 4 N A Nt 1 , A Nt A Nt 1 where 4 N A Nt 1 , A Nt § A Nt A Nt 1 · = TN ¨ ¸ A Nt 1 © ¹ 2 Composite traded good (Armington aggregator) 1 ] yTt = M P xDt 1 P mt] ] Investment good iDt + iNt J 1J GzTIt z NIt k Dt 1 ) (iDt / k Dt )k Dt 1 G k Dt k Nt 1 ) (iNt / k Nt )k Nt 1 G k Nt ) (i / k ) ª G 1K i / k K 1 K G / K º ¬ ¼ Market clearing Domestically produced traded good xDt + xFt = yDt Nontraded good cNt z NIt z NDt z NNt y Nt Composite traded good cTt zTIt zTDt zTNt Labor market A Dt A Nt At yTt Balance of payments pDt xFt 1 rt bt mt bt 1 Foreign demand xFt Dt 1 W Ft pDt Transfer of tariff revenue Tt W Dt mt 1 1] Sudden stop! bt bt 1 , t 1995,1996 x agents do not foresee sudden stop x agents do foresee length of sudden stop x domestic interest rate is endogenously determined x interest payments on foreign debt made at r * Calibration Rest of world is U.S. x 72% of imports to Mexico from U.S. (1988-2002) x 60% of foreign direct investment from U.S. (1988-2002) x r* 5% E Elasticities x subs. tradable and nontradable cons.: 0.5 (Kravis, et al.) x intertemporal elasticity: 0.5 x subs. domestic tradable and imports: 2.0 Labor Adjustment T D T N x labor shift from sudden stop: 6.5% Calibration continued Set G 0.06 Normalize prices in 1989 to 1 Mexico input-output table, 1989 x share parameters: aTD , aND , aTN , aNN , D D ,D N , H , P x scale parameters: AD , AN , M , D Input – Output Table, 1989 9.02 36.26 24.47 11.13 14.98 50.59 86.85 Nontradable Total intermediate consumption Employee compensation 9.76 36.00 21.29 19.42 28.44 43.71 29.18 65.44 65.00 52.49 76.96 0.00 11.90 23.03 0.00 0.00 14.98 0.00 64.40 114.99 0.00 93.58 180.43 65.00 Return to capital Value added Imports 13.58 34.87 14.98 21.42 65.13 0.00 35.00 100.00 14.98 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 35.00 100.00 14.98 Tariffs Total Gross Output 0.00 86.85 0.00 93.58 0.00 180.42 0.00 76.96 0.00 23.03 0.00 14.98 0.00 114.99 0.00 295.41 Tradable aTN zTN ,1989 y N ,1989 Use Int.Tradables GO Nontradable w1989 { 1 A D ,1989 9.76 93.58 21.29, A N ,1989 Total final demand Exports 27.24 Commodity Total int. demand Tradable Nontradable Investment Total Demand Final Demand Consumption Input 0.10 43.71 No sudden stop Model begins in 1988 x k D ,1988 k N ,1988 k1988 x A D ,1988 A N ,1988 A1988 x allow labor capital to costlessly adjust (anticipated) Mexico opens to capital in 1990 Exogenous x population growth x initial capital stocks No exogenous TFP growth! No sudden stop dynamics No sudden stop: capital stock No sudden stop: trade balance 400 10.0 350 share of GDP (%) index (1988 = 100) 5.0 300 250 200 0.0 -5.0 150 100 1988 1993 1998 2003 2008 2013 2018 2023 2028 -10.0 1988 1993 1998 2003 2008 2013 2018 2023 2028 Baseline Model Model begins in 1988 x k D ,1988 k N ,1988 k1988 x A D ,1988 A N ,1988 A1988 x allow labor capital to costlessly adjust (anticipated) Mexico opens to capital in 1990 Sudden stop hits in 1995 Mexico: trade balance 8.00 percent of gdp (%) 4.00 0.00 Data -4.00 -8.00 -12.00 1988 Model 1990 1992 1994 1996 1998 2000 2002 Mexico: real exchange rates 0.5 0.3 log(RER) Data 0.0 -0.3 Model -0.5 1988 1990 1992 1994 1996 1998 2000 2002 Mexico: nontraded/traded good prices 0.6 log(RERN) 0.4 0.2 Model 0.0 Data -0.2 -0.4 1988 1990 1992 1994 1996 1998 2000 2002 Mexico: sectoral value added 140 140 130 130 120 120 tradable goods index (1994=100) index (1994=100) Mexico: sectoral value added 110 100 nontradable goods 100 nontradable goods 90 90 80 80 70 1988 1990 1992 1994 1996 model 1998 2000 2002 tradable goods 110 70 1988 1990 1992 1994 1996 data 1998 2000 2002 Mexico: traded good employment, detrended 0.08 deviation from trend 0.06 Model 0.04 0.02 Data 0.00 -0.02 -0.04 1988 1990 1992 1994 1996 1998 2000 2002 Real GDP Yt p p Dt0 yDt pTt0 zTDt pNt0 z NDt Nt0 y Nt pTt0 zTNt pNt0 z NNt Real Investment pTt0 zTIt pNt0 z NIt It Real Capital Stock K t 1 (1 G ) K t I t Total Factor Productivity TFPt Yt K tD L1tD W Dt mt Mexico: GDP per working age person Mexico: Total Factor Productivity 115 110 110 index (1994 = 100) index (1994=100) 105 105 Model 100 95 Data Model 100 Data 95 90 85 1988 1990 1992 1994 1996 1998 2000 2002 90 1988 1990 1992 1994 1996 1998 2000 2002 Extension: country risk premia Adjusted return on Mexican debt higher than U.S. x time-varying country specific risk premia Take premia as exogenous * * r V mex ,t x rmex ,t Mexico: interest rates 75 percent per year baseline model 50 risk premia model 25 data 0 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Mexico: trade balance Mexico: traded good employment, detrended 0.08 8.00 0.06 deviation from trend share of gdp (%) 4.00 0.00 data -4.00 risk premia model -8.00 risk premia model 0.04 0.02 data 0.00 -0.02 baseline model baseline model -12.00 1988 1990 1992 1994 1996 1998 2000 -0.04 1988 2002 Mexico: real exchange rates 1994 1996 1998 2000 2002 2000 2002 0.50 0.25 data log(RERN) log(RER) 1992 Mexico: nontraded/traded good prices 0.50 0.25 1990 0.00 -0.25 data 0.00 risk premia model baseline model -0.25 risk premia model baseline model -0.50 1988 1990 1992 1994 1996 1998 2000 2002 -0.50 1988 1990 1992 1994 1996 1998 Mexico: sectoral value added Mexico: sectoral value added 140 140 tradable goods 130 130 risk premia model 120 110 baseline model 100 90 tradable goods 110 100 nontradable goods 90 nontradable goods 80 70 1988 index (1994=100) index (1994=100) 120 80 1990 1992 1994 1996 model 1998 2000 2002 70 1988 1990 1992 1994 1996 data 1998 2000 2002 Mexico: GDP per working age person Mexico: Total Factor Productivity 115 110 110 risk premia model index (1994 = 100) index (1994=100) 105 105 100 95 baseline model 100 risk premia model 95 data data 90 baseline model 85 1988 1990 1992 1994 1996 1998 2000 2002 90 1988 1990 1992 1994 1996 1998 2000 2002 Extension: foresight Agents know of sudden stop in 1994 x Agents do foresee length of sudden stop x Domestic interest rate is endogenously determined Take premia as exogenous * * r V mex ,t x rmex ,t Mexico: trade balance Mexico: traded good employment, detrended 0.08 8.00 surprise model 0.06 deviation from trend percent (%) 4.00 0.00 data -4.00 surprise model 0.04 data 0.02 0.00 -0.02 no surprise model -8.00 -0.04 -12.00 1988 1990 1992 1994 1996 1998 2000 -0.06 1988 2002 Mexico: real exchange rates 0.50 0.25 0.25 no surprise model log(RERN) log(RER) 0.00 -0.25 -0.50 1988 1992 1994 1996 1992 1994 1996 1998 2000 2002 1998 no surprise model surprise model 0.00 data -0.25 surprise model 1990 1990 Mexico: nontraded/traded good prices 0.50 data no surprise model 2000 2002 -0.50 1988 1990 1992 1994 1996 1998 2000 2002 Mexico: sectoral value added Mexico: sectoral value added 140 140 tradable goods 130 no surprise model 120 120 index (1994=100) index (1994 = 100) 130 110 100 surprise model 90 nontradable goods 100 nontradable goods 90 80 70 1988 tradable goods 110 80 1990 1992 1994 1996 1998 2000 2002 70 1988 1990 1992 1994 1996 1998 2000 2002 Mexico: GDP per working age person Mexico: Total Factor Productivity 110 110 surprise model 105 index (1994 = 100) index (1994 = 100) 105 100 no surprise model 95 data no surprise model 100 surprise model 95 90 85 1988 data 1990 1992 1994 1996 1998 2000 2002 90 1988 1990 1992 1994 1996 1998 2000 2002 Accounting for GDP x Take TFP as exogenous D D 1D D yDt = min ª¬ zTDt / aTD , z NDt / aND , ADt k Dt A Dt º¼ D N 1D N y Nt = min ª¬ zTNt / aTN , z NNt / aNN , ANt k Nt A Nt º¼ x All else same Mexico: total factor productivity 110 index (1994 = 100) 105 model 100 data 95 90 1988 1990 1992 1994 1996 1998 2000 2002 Mexico: GDP per working age person 115 index (1994=100) 110 105 100 Data 95 Model 90 85 1988 1990 1992 1994 1996 1998 2000 2002 Mexico: sectoral value added 140 140 130 130 tradable goods 120 index (1994=100) 120 index (1994=100) Mexico: sectoral value added 110 100 nontradable goods 90 100 nontradable goods 90 80 70 1988 tradable goods 110 80 1990 1992 1994 1996 model 1998 2000 2002 70 1988 1990 1992 1994 1996 data 1998 2000 2002 Mexico: trade balance Mexico: traded good employment, detrended 0.08 8.00 0.06 deviation from trend percent of gdp (%) 4.00 0.00 Data -4.00 -8.00 -12.00 1988 Model 1990 1992 1994 Model 0.04 0.02 Data 0.00 -0.02 1996 1998 2000 -0.04 1988 2002 1990 Mexico: real exchange rates 1992 1994 1996 1998 2000 2002 2000 2002 Mexico: nontraded/traded good prices 0.5 0.6 0.4 0.3 log(RERN) log(RER) Data 0.0 0.2 Model 0.0 Data -0.3 -0.2 Model -0.5 1988 1990 1992 1994 1996 1998 2000 2002 -0.4 1988 1990 1992 1994 1996 1998 Further Work 1. Leisure choice 2. Nonconvex adjustment Conclusions 1. Sudden stops affect sectors differently x increase in pT p N x tradable good output falls less than nontradable x labor reallocated to tradable good sector 2. International prices cannot affect GDP/TFP 3. Multisector model accounts for x real exchange rate x labor allocation x trade balance