Thre Essays on Antidumping Duties and Risk Factors Affecting International Seafood Trade by Dengjun Zhang A disertation submited to the Graduate Faculty of Auburn University in partial fulfilment of the requirements for the Degree of Doctor of Philosophy Auburn, Alabama December 12, 2011 Keywords: Antidumping duties, exchange rate risk, price risk, demand system, cointegration Copyright 2011 by Dengjun Zhang Approved by Henry Kinnucan, Chair, Profesor of Agricultural Economics and Rural Sociology Henry Thompson, Profesor of Agricultural Economics and Rural Sociology Norbert Wilson, Asociate Profesor of Agricultural Economics and Rural Sociology Andrew Muhamad, United States Department of Agriculture ! ! ! ii Abstract This disertation includes three esays to addres influences of the antidumping policy, price risk, and exchange risk in the US seafood import market. In the first chapter, I atempt to evaluate the efect of trade diversion on the efectivenes of antidumping duties. Antidumping duties can have the unintended consequence of increasing imports from countries not named in the dispute. Previous research suggests ?trade diversion? significantly undermines the efectivenes of antidumping duties. However, domestic consumers could reduce their losses, as they can substitute out of the dutied good and into both the domestic good and the non- dutied good, which is, as might be expected, les expensive than the domestic good. Testing these propositions using the 2003 antidumping duty imposed by the United States on catfish imports from Vietnam, I find that the trade-diversion efect was significant in the sense that the quasi-rents of US producers enjoyed from the antidumping duty were reduced. The impact on consumer welfare was modest due to the great loss generated from the higher price of the dutied-good. An implication for public policy of the empirical results is that, in terms of the national welfare gain, it is optimal to lower the tarif rate when the amount of trade diversion is large. The second chapter examines the impact of the export price risk and exchange-risk on the import demand. An extended Rotterdam model reveals that risk factors take efect on marginal utility via ?adjusting? prices. This coincides with viewpoints in trade literature that ! ! ! iii risk-averse importers atach a risk premium as an extra mark-up to cover the cost of exchange risk (Balg and Metcalf, Bergin) and/or price risk (Wolak and Kolstad). The derived model further demonstrates that the trade efects of risk factors depend on own-price elasticities and substitutability betwen products within the same group. The modified specification makes testing restrictions on the efects of the risk variables plausible, resulting in a reduction in the parameter space. By further decomposing import price risk into export price risk and exchange-rate risk, the empirical model is applied to the US salmon import market. The results support the hypothesis that importers are sensitive to price and exchange-rate risks but reject the proposition that those two risk factors exert an identical efect on import demand. The third chapter focuses on the diference betwen the dynamic and static specification for an import demand system inclusive of the price risk factor. I build risk factors into an Almost Ideal Demand System (AIDS) where the price risk may play a role by influencing baseline imports or import responsivenes to price. The risk-augmented AIDS model tests causality betwen price risk and trade. A multivariate GARCH approach estimates the conditional variances of prices. Nonstationarities in the data and endogeneity of price and price volatility are taken into acount with Johansen?s approach and a Vector Error Correction Model (VECM). The empirical results uncover a substantial role of risk factors in the US codfish import market. China?s import share (CIF) increased 18% from 2004 to 2005, while Canada lost 9%. Holding other factors constant, high fluctuation of Canada?s price would reduce its share by 69%, and raise China?s share almost two times. China?s relatively stable price further diminishes Canada?s share by 4% and increases China?s share by 9%. ! ! ! iv Acknowledgments I would like to thank my advisor and commite chair, Dr. Henry Kinnucan, for initiating me to do research from an economic perspective. This is one of the most important principles I have learnt in Auburn and has been contributed substantialy to each chapter of my disertation. Other appreciation goes to Dr. Henry Thompson, Dr. Andrew Muhamad, and Dr. Norbert Wilson for their profesional comments. Special thanks to Dr. Hyeongwoo Kim for his econometric help. I also want to expres my appreciation to Dr. Valentina Hartarska and Dr. Diane Hite for their excelent teaching and research guidance. Finaly, I owe my parents, my wife Jenny, and my son Andy a big debt of gratitude and appreciation for their love, support, and patient in these years. In this period, with litle help from me, Andy learnt to speak the Norwegian language fluently and speak English at least as well as me. Your growing gives me the motivation to concentrate on my studies every day, because it is the only way I can shorten the time of separation so that I can be reunited with you and Mum. ! ! ! v Table of Contents Abstract ........................................................................................................................................... ii Acknowledgments .......................................................................................................................... iv List of Tables ................................................................................................................................ vii List of Figures ................................................................................................................................ ix Chapter I. Antidumping Duties and Trade Diversion: An Armington Procedure .......................... 1 1.1 Introduction ................................................................................................................. 2 1.2 Model ........................................................................................................................... 5 Tarif Pas-Through Elasticity (PTE) ............................................................... 6 Trade-Diversion Efect ..................................................................................... 9 1.3 Parameterization ........................................................................................................ 11 1.4 Simulated Tarif Pas-Through Elasticities (PTE) .................................................... 13 1.5 Welfare Analysis ....................................................................................................... 15 Distribution efects .......................................................................................... 17 Sensitivity Analysis ........................................................................................ 19 Tarif Rate and Trade Diversion ..................................................................... 21 1.6 Concluding Comments .............................................................................................. 22 Chapter II. Import Demand under Price and Exchange-Rate Uncertainties ................................. 25 2.1 Introduction ............................................................................................................... 26 ! ! ! vi 2.2 Theoretical Model ..................................................................................................... 29 2.3 Model Specifications and Data .................................................................................. 35 2.4 Measurement of Uncertainty ..................................................................................... 37 2.5 Regresion Results .................................................................................................... 38 Model selection and Hypothesis tests ............................................................. 39 Price and Expenditure Elasticities .................................................................. 41 Volatility Elasticities ....................................................................................... 43 2.6 Conclusions ............................................................................................................... 44 Chapter III. A Risk-Augmented Cointegrating Import Demand System ..................................... 46 3.1 Introduction ............................................................................................................... 47 3.2 Background ............................................................................................................... 49 3.3 The Theoretical Model .............................................................................................. 52 3.4 Empirical Model and Data ........................................................................................ 54 3.5 Price Volatility .......................................................................................................... 56 3.6 Econometric Procedure ............................................................................................. 58 3.7 Regresion Results .................................................................................................... 60 3.8 Summary and Implications ........................................................................................ 64 References ..................................................................................................................................... 66 Appendix 1: Tables and Figures for Chapter I .............................................................................. 74 Appendix 2: Tables and Figures for Chapter II ............................................................................ 82 Appendix 3: Tables and Figures for Chapter III ........................................................................... 93 ! ! ! vii List of Tables Table 1.1 Prices, Quantities, and Market Shares for Domestic and Imported Frozen Catfish Filets, United States, 1999 - 2010 ........................................................................... 75 Table 1.2 Simulated Duty Pas-through Elasticities a ............................................................... 76 Table 1.3 Distributional Efects of a 35% Catfish Antidumping Duty Under Alternative Asumptions About the Market Share of the Non-Dutied Good ............................. 77 Table 1.4 Sensitivity of Welfare Efects of the Catfish Antidumping Duty to the Armington Elasticity (!) and Import Supply Elasticities (" = " 2 = " 3 ) a ...................................... 78 Table 1.5 Sensitivity of Trade-Diversion Efects to Antidumping Duty Rates a ...................... 79 Table 2.1 US Salmon Imports and Market Shares by Sources .................................................83 Table 2.2 GM Estimates of U.S. Import Demand System for Salmon, Rotterdam Model Inclusive Import Price Risk, 1995-2008 Monthly Data (1 = Chile, 2 = Canada, 3. Norway, 4 = United Kingdom, 5 = ROW) .............................................................. 84 Table 2.3 Tests of Theoretical Restrictions ............................................................................. 85 Table 2.4 GM Estimates of U.S. Import Demand System for Salmon, Rotterdam Model 1995-2008 Monthly Data (1 = Chile, 2 = Canada, 3 = Norway, 4 = United Kingdom, 5 = ROW) ............................................................................................... 86 Table 2.5 Conditional Demand Elasticities (1 = Chile, 2 = Canada, 3 = Norway, 4 = United Kingdom, 5 = ROW) ............................................................................................... 87 Table 2.6 Conditional Demand Elasticities (1 = Chile, 2 = Canada, 3 = Norway, 4 = United Kingdom, 5 = ROW) (cont) ..................................................................................... 88 Table 3.1 US Cod Imports, Expenditure Shares, Prices by Sources .........................................94 Table 3.2 Johansen Cointegration Test Results ....................................................................... 95 Table 3.3 VECM Estimates of US Import Demand System for Codfish, AIDS Model, 1989 - ! ! ! vii 2010 Monthly Data (1 = Canada, 2 = China, 3 = Iceland, 4 = ROW) ..................... 96 Table 3.4 Derived Demand Elasticities (1 = Canada, 2 = China, 3 = Iceland, 4 = ROW) ...... 97 Table 3.A1 ARCH Test for Price (1 = Canada, 2 = China, 3 = Iceland) ...................................100 Table 3.A2 M-GARCH Estimates of Conditional Variance of Price (1 = Canada, 2 = China, 3 = Iceland) .................................................................................................................. 101 Table 3.A3 Unit Root Tests (1 = Canada, 2 = China, 3 = Iceland, 4 = ROW) ......................... 102 Table 3.A4 SUR Estimates of US Import Demand System for Cod, AIDS Model, 1989-2010 Monthly Data (1 = Canada, 2 = China, 3 = Iceland, 4 = ROW) ............................ 103 ! ! ! ix List of Figures Figure 1.1 Welfare Analysis of an Antidumping Duty ............................................................. 80 Figure 2.1 Import US Dollar Prices of Salmon by Sources .......................................................89 Figure 2.2 Conditional Variance Estimates of Import Prices (in US dollar): January 1995 - December 2008 ........................................................................................................ 90 Figure 2.3 Conditional Variance Estimate of Export Prices (in Foreign Currency): January 1995 - December 2008 ............................................................................................. 91 Figure 2.4 Conditional Variance Estimate of Bilateral Exchange Rate: January 1995 - December 2008 ........................................................................................................ 92 Figure 3.1 M-GARCH Estimated Conditional Price Volatility .................................................98 Figure 3.A1 US Import Cod Prices by Sources ............................................................................99 ! 1 Chapter I. Antidumping Duties and Trade Diversion: An Armington Procedure ! 2 1.1 Introduction An antidumping (AD) duty aims to asist domestic producers by raising price in the home market. Blonigen (2003) reports that AD cases worldwide increased from only a few in the 1970s to 2,200 in the 1990s. In the United States, calculated dumping margins betwen 1980 and 2000 rose from 15% to over 63%, and the probability of an afirmative ruling rose from 45% to over 60% (Blonigen 2006). Durling and Prusa (2006 p. 676) note that the range of products subjected to AD has been expanded, with measures frequently targeting agricultural products. Frequent enhancements of AD measures induce extensive disputes about the validity of this policy, which spurs persistent interest in the international trade literature. The most common consensus is that the targeted tarif is a priori ineficacy due to: (i) a large import demand elasticity (Kinnucan 2003, Kinnucan and Myrland 2006), (ii) the les competitive import- competing industry (Chang and Gayle 2006), and (iii) the trade-diversion effect (Prusa 2001, Galaway et al. 1999). Hencefore, the AD duty wil induce the non-targeted foreign suppliers to increase their shipments. The trade-diversion efect on AD duties is the primary focus of this paper. The standard result in the economics literature is that trade diversion undermines the efectivenes of AD duties. However, as noted by Carter and Galant-Trant (2010), most of the studies relate to industrial products, and there are good reasons to expect a diferent outcome for agricultural products. Unlike industrial products, agricultural products tend to be perishable, are produced seasonaly, and must met stringent quality and food safety standards before they can be admited for importation. These factors, coupled with weak control over production and export decisions due to the biological and atomistic nature of agricultural production, constrain the extent to which non-targeted countries can fil the gap in the domestic market left by a ! 3 particular antidumping action. Indeed, in their empirical investigation Carter and Galant (2010, p. 119) find ?a relatively smal amount of trade diversion for agricultural products, and this efect does not extend beyond the year in which the case was initiated.? Although economists find litle justification for AD duties because of the losses to consumers and downstream industries in the importing countries, the policy persists, filings have multiplied, and rulings have become more protectionist. Irwin (2005) reports that import- competing firms in the United States are increasingly turning to AD duties as an ?easy? means to gain protection. A likely explanation is that the government distributes the revenue gained from AD duties to domestic producers aleging harm. 1 Therefore, the national welfare can be used to assess efectivenes of AD duties, which should also be influenced by trade diversion. Furthermore, part of what domestic producers gain is matched by the los to domestic consumers. Trade diversion should benefit US consumers if the price of the non-dutied good is les expensive than the price of the protected good. The purpose of this research is to determine the efects of trade diversion on the efectivenes of the antidumping duty that was imposed by the United States on catfish imports from Vietnam (US Department of Commerce 2003, 2009). The ?Catfish War? generated national media atention, with articles appearing in The New York Times, The Wall Stret Journal, The Christian Science Monitor, and The Economist discussing the policy and ethical dilemmas posed by the dispute. 2 For the purposes of this paper, catfish is a useful case study in that the trade diversion isue is clearly indicated in the data (table 1). Specificaly, prior to the implementation !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 1 In the US, the Continued Dumping and Subsidy Ofset Act, comonly known as the ?Byrd Amendment? (se Chang and Gayle 2006), was enacted by the US Congres in 2000 to provide for the distribution of antidumping duty revenues to petitioners. It was repealed in 2006, with a phase-out period for duty distribution. 2 The WSJ article entitled ?Catfish Case Mudies Waters for Bush ?Fast Track?? apeared 13 July 201. For citations of the remaining articles and a god discusion of the isues involved, se Coleman (205, pp. 6-8). ! 4 of antidumping duties ranged from 37% to 64% in 2003 US producers, in esence, owned the domestic market with a market share of 0.84. One year after the antidumping duty went into efect, China entered the US market, and within four years, saw its market share increase from 0.01 to 0.17. Although by 2008 Vietnam had re-established itself as an important competitor with a market share of 0.18, it sems clear that without the entry of China: a) the quasi-rents US producers enjoyed from the antidumping duty would have been higher, and b) the ability of the duty to protect US producers? market share would have been greater. Examining these hypotheses, I find that the entry of China did indeed undermine duty efectivenes. But the degree of the trade-diversion efect is primarily limited by the domestic industry?s dominant market share. This ?home bias? in an Armington (1969) framework implies an atenuated pas- through elasticity and weak cross-price efects for the domestic product. 3 Previous research on the pas-through elasticity shows that the Armington substitution elasticity (!) plays an important role in the extent to which a tarif-induced increase in the price of an imported good wil raise the price of the protected good (Warr 2008). However, no study to our knowledge investigates the relationship betwen the market share, the pas-through elasticity, and trade-diversion efects. Thus, as a by-product of our analysis, I fil a gap in the literature relative to a key elasticity used to evaluate trade policy. A key finding is that the pas- through elasticity is inversely related to the market share of domestic producers. This ?home bias? means a smal market share of imports. I will further ilustrate that the smaler market share of imports, the les important trade-diversion efects are apt to be. We begin with the presentation of the structural model. Analytical expresions for the pas-through elasticity and the trade-diversion efect are then derived. The welfare efects of the !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 3 According to Whaley and Xin (209, p. 309), home bias is comonly defined in the literature as ?an Armington preference for domestic over comparable foreign products in a trade model where gods are heterogeneous across countries, a definition I adopt in this study. ! 5 catfish antidumping duty are then measured, with special atention to trade-diversion efects. The paper concludes with a summary of the key findings. 1.2 Model Consider a country that imports two products, Q 2 and Q 3 , diferentiated by source origin. The imported products compete with a home good, Q 1 . The three goods are weakly separable from al other goods in the home market, and consumers alocate income in acordance with the two- stage budgeting hypothesis. The home country imposes an antidumping duty on Q 2 defined as follows: (1) ! P 2 = where T is the duty expresed in ?iceberg? or proportionate form, 2 ~ P is price inclusive of the duty, hereafter ?demand price?, and 2 Pis price exclusive of the duty, hereafter ?supply price.? Prices for the three goods are determined under competitive conditions. Given these asumptions, an isue is the efect of the duty on the price of the domestic good when the presence of the non-dutied good, Q 3 , is taken into acount. To determine that, let the demand and supply equations for the three goods be defined as follows: (2) ? (4) ** 33 * 22 * 11 * ~ XPPPQ iiiii !""" +++= 3,2,1=i (5) ? (7) ** iii PQ != 3,2,1=i where the asterisk denotes proportionate change. Thus, for example, X* is the proportional change in X where( 12 + ! QP 3 )is total expenditure on the three goods by domestic consumers. In this model, the " ij are conditional price elasticities of demand; the # i are conditional expenditure elasticities; and the $ i are price elasticities of supply. The demand curves ! 6 are downward sloping (" ii < 0); the supply curves are non-decreasing ($ i # 0); the goods are non- inferior (# i # 0) and substitute for one another in consumption (" ij > 0 for i $ j). An increase in the duty shifts the supply curve for the dutied good to the left. To se this, write equation (1) in proportionate change form: (8) ! P 2 * =+T * . Substituting equation (8) into equation (6) yields the tax-burdened supply curve: (6%) Q 2 * ! 2 * . Holding constant the demand price, a 1% increase in the AD shifts the supply curve for the dutied good to the left from the initial market equilibrium point by an amount equal in proportionate terms to the dutied good?s supply elasticity. Equation (6) shows that an increase in the duty has opposite efects on the demand and supply prices. Thus, for example, if the supply of the dutied good is perfectly inelastic (" 2 = 0), foreign producers bear the full incidence andP 2 * =!T. In this instance, the duty is inefective, as it has no efect on prices in the domestic market. Thus, a necesary condition for the duty to be efective is that " 2 > 0, an isue to which I shal return. Equations (2) - (8) contain seven endogenous variables (P 1 * , 23 , ! * Q 1 , 23 * ) and two exogenous variables ( * X and T). Exogenous variables that afect demand and supply other than consumer expenditure and the duty are suppresed. Tarif Pas-Through Elasticity (PTE) Warr states (2008, p. 499) ?The manner in which the landed price of imports afects domestic prices is central to trade policy analysis.? The tarif pas-through elasticity (PTE) when an AD is used to raise the import price may be defined as follows: ! 7 (9) P 1 * T = ! 2 " # $ % & * where the first term in parenthesis indicates the price pas-through elasticity and the second term indicates duty incidence. In Warr?s analysis the supply of the imported good is asumed to be perfectly elastic ($ 2 = &), which implies the full incidence of the duty is borne by domestic consumers, i.e., ! P * T=1. Here I relax that asumption and add trade-diversion efects to develop a more complete expresion for the PTE. The PTE when third party efects are considered are (se appendix for derivation): 4 (10) P 1 * ! " # $ % & N=3 21 (! 3 '")+ 132 (11) ! 2 * T " # $ % & N=3 2 ( 1 ) 313 ) (12) P 3 * ! " # $ % & N= 23 (! 1 '")+ 123 where 231231321321333122122231131112332333222111 )()()())()(( !!!!!!!"!!!"!!!"!!!"!"!" ###########=$ , and equation (11) is trade incidence borne by domestic consumers. After presuming a downward-sloping demand curve and an upward-sloping supply curve for each product in the market, I can justify the common denominator (') in (10) - (12) is positive. Noted in the Armington framework, any pair of products in the group is a gross substitute (" ij > 0 for i $ j); I can further observe each numerator in (10) - (12) with a positive sign as wel. This indicates that the AD duty imposed on the subject country should increase !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 4 PTE in general means the tarif pas-through elasticity of the domestic god?s price. In order to elaborate the relative price movements, I also calculate the PTE of the dutied god?s domestic price and the non-dutied good?s price. Except stated otherwise, in the paper, PTE only refers to price transmission betwen tariff and the price of the protected good. ! 8 domestic prices of the protected good, the dutied product, and the not-dutied good. An increase in the price of the dutied product is induced by an inward supply shift, whereas a rise in the prices of the protected and non-dutied products is induced by upward demand shifts. In order to demonstrate the tarif pas-through proces, equation (10) is restated as follows: (13) P 1 * T ! " # $ % & N=3 12 (" 3 '!)+ 132 $ % & ! P * T " # N=3 where the first term is price pas-through elasticity, which is primarily determined by substitution betwen goods from diferent sources. In (10) or (13), the supply elasticity of the dutied good is a critical determinant of duty efectivenes. For example, if the supply of the dutied good is perfectly inelastic ( 0 2 =! ), as might be true in a short-run situation, the entire incidence of the duty is borne by producers in the exporting country and the protected good?s PTE = 0. This is true regardles of the size of the price pas-through elasticity (the first term in equation (13)) and whether trade-diversion efects are considered. Alternatively, if supply of the dutied good is perfectly elastic, equation (13) is reduced to: (14) P 1 * T ! " # $ % & 2 =' N3 12 (! 3 ")+ 12 33 The corresponding expresion when trade-diversion efects are ignored is: (15) P 1 * ! " # $ % & 2 =' N 12 ( Equations (14) and (15) represent the maximal efect of the duty on the price of the protected good. Equation (15) is identical in form to the PE derived by Warr (2008, p. 500, equation (6)). That this elasticity has an upper limit of one can be sen by imposing homogeneity to yield: ! 9 (16) P 1 * T ! " # $ % & 2 =' N 12 +# . In this instance, the PTE increases as the protected good becomes a closer substitute for the dutied good (larger " 12 ), as the supply of the protected good becomes les price elastic (smaler $ 1 ), and as the demand of the protected good becomes les expenditure elastic (smaler # 1 ). Trade-Diversion Efect An analytical expresion for the trade-diversion efect (TDE) can be obtained by subtracting equation (14) from equation (15) to yield: (17) ( ) ( ) 3113333111111 12311113213 ))(()( )( !!!"!"!" !!!"!! #### +## =TDE . Equation (17) is the TDE when domestic consumers bear the full incidence of the duty and thus represents the maximal efect. With this caveat in mind, the TDE is negative in sign, which means ignoring the TDE wil cause the duty?s efect on domestic price to be overstated. The degree of overstatement increases as: i) the protected good becomes a closer substitute for the non-dutied good (larger " 13 ), ii) the non-dutied good becomes a closer substitute for the dutied good (larger " 32 ), and iii) the supply of the non-dutied good becomes les price elastic (smaler $ 3 ). If $ 3 = &, the duty has no efect on the price of the non-dutied good and TDE = 0. Thus, the extent to which the presence of a third good undermines the efectivenes of an AD depends on substitution efects, but also on supply conditions in the market for the non-dutied good. If consumers respond to the duty by switching to the non-dutied good to a greater extent than to the protected good (" 32 > " 12 ), as might be expected if the non-dutied good is les expensive than the ! 10 protected good (P 3 < P 1 ), the atenuation of the PTE caused by the TDE could be empiricaly important. In the Armington model, consumer preferences are asumed to be homothetic, indicating trade paterns changes only with relative price movements. Since an AD duty increases the domestic price of each good in the group, the magnitudes of the growth rate should ilustrate redistribution of shares among suppliers, and consequently reveal the mechanisms of the trade- diversion efect. The diferences betwen PTE are given by: (18) ! P 2 * T " # $ % & N=3 ' 1 * =3 ! 2 ( 1 '" 12 )(! 3 +" 13 (' 32 ) (19) 3 * ! " # $ % & N= 1 * =3 2 3 ( 113 ) 231 ! 3 )( (20) ! P 2 * T " # $ % & N=3 ' * =3 ! 2 ( 1 '")! 332 )+" 1 ( 3 ' 12 ) For equation (18), the increase of the domestic price of the dutied good is greater than the growth of the price of the protected good, on the sufficient conditions that (1) |" 11 | > |" 12 | which is not a particularly onerous condition and (2) the non-dutied good is a closer substitute to the protected good than to the dutied good, i.e. " 31 > " 32 (as shown later, it is always true in the Armington model given the US produces dominate the market). Considering the formula of demand elasticities, equation (19) can be simplified to equal ! 2 3 ( 1 "), indicating the growth of the non-dutied good price is les than the growth of the protected good price if $ 1 < $ 3 . Although signs of (18) and (19) depend on the reasonable pre-conditions, I can tel the sign of (20) is unequivocaly positive, indicating an improved comparative advantage of the non-dutied good compared to the dutied good. In general, implementation of an AD duty wil make the ! 11 relative price movements favor both the protected good and the non-dutied good over the dutied good, resulting in trade diversion. 1.3 Parameterization To ases the importance of the trade-diversion efect for the catfish AD, I need to set parameters in the structural model. For this purpose, I first follow Warr (2008) and restrict demand elasticities to conform to the ?Armington asumption? (Armington 1969). In the Armington?s framework, imports from diferent sources and domestic production are asumed to be imperfectly substitutable. The expenditure of a particular group is alocated to diferent suppliers based on relative price movements. This implies the demand elasticities matrix is as follows: (21) ! ! ! " # $ $ $ % & ''++ +''+ ++'' = ! ! ! " # $ $ $ % & = ())()( )(())( )()(() ))) ))) ))) )1()()( )()1()( )()()1( 3321 3221 3211 333231 232221 131211 SSSS SSSS SSSS ! where " = -1 is the price elasticity of demand for the three goods combined, and ! > 1 is the Armington substitution elasticity. With the maintained hypothesis that the three goods form a weakly separable group, the conditional budget shares sum to one. One advantage of the Armington framework is that the number of parameters to derive the 9 demand elasticity is reduced to three, i.e., !, S 1 , and S 2 , noting that S 1 = 1 - S 2 - S 3 . Muhamad et al. (2010, p. 437) estimate the (long-run) own-price elasticity of demand for US catfish to be ? 1 =.4. The sample period of the data used in Muhamad et al. is from January 1993 to December 2007. Substituting this value and S 1 = 0.83 (mean value for 1999- 2007 in our sample, close to the period employed in Muhamad et al.) into the expresion for " 11 in equation (21), and solving for the substitution elasticity, yield ? !=37. Considering the decline in S 1 during the data period, our ?best-bet? estimate is 2.5. However, to gauge the ! 12 sensitivity of results to substitutability betwen products in the same group, I also conduct alternative estimates of ? !=1.5 and ? . Although the lower bound is outside the typical range (from 2 to 5) for Armington elasticity asumed in applied equilibrium models (Warr 2008), it is used to give the simulation results full play. Muhamad et al. (2010, p. 437) estimate the long-run supply elasticity for domestic catfish to be 1.05. So, I set the ?best-bet? value for ? ! 1 =. Since Muhamad et al. did not estimate supply elasticities for imported catfish, I set 2 and ? ! 3 to 2, the value used in Kinnucan?s (2003, p. 216) analysis. Also, to gauge the sensitivity of the results to import supply response, I conducted additional simulations with ? ! 2 and 3 set alternatively to 4 and infinity. As discussed before, the final determination of the AD measure was isued against the targeted catfish in 2003, and this AD case remained in efect in 2009. Acordingly, I set three baseline values from the data sample sub-periods: 2002-04, 2005-07, and 2008-10. When simulating, market shares, prices and quantities are mean values in each sub-period: In 2002-04, on average, US catfish producer dominated the domestic market with a market share of 0.83, and China owned a mere share (S 3 = 0.01). After the implementation of the AD, China saw its market share rise to 0.11 at the expense of US producers in 2005-07; however, Vietnam only lost 0.03 shares. In 2008-10, Vietnam re-established itself as an important supplier with a market share of 0.26, whereas China only gained a growth of 0.03. A comparison of ! ! ! " ! # " ! " " " # # ! # " # # 2002-20042.47 1.46 1.45 311 632 0.83 0.17 0.01 2005-20072.84 1.54 1.69 32763470.75 0.14 0.11 2008-20102.94 1.55 1.76 291127670.60 0.26 0.14 $%&'()*+,-)./0123 45678&89)*:&02)01;3 <6%=(8)>?6%( Year ! 13 simulating results with the baseline values from those sub-periods should shed light on the trade- diversion efect. Given " = -1 and the ?best-bet? value of ? !=2.50, the demand elasticities matrices (equation (21)) corresponding to the three sets of baseline market shares are: (22) N= -1.260.5.1 4 ..-2.49 ! " # $ % & (2002-04: S 1 = 0.83, S 2 = 0.17, S 3 = 0.01) (23) -1.380..16 .2.-.34 ! " # $ % & (2005-07: S 1 = 0.75, S 2 = 0.14, S 3 = 0.11) (24) N= -1.60.9.21 ..3-. ! " # $ % & (2008-10: S 1 = 0.60, S 2 = 0.26, S 3 = 0.14) The most important elements in matrix (22) - (24) are " 12 and " 13 , as these elasticities are primary determinants of tarif pas-through elasticities and trade-diversion efects (se equations (10) and (17)). Since neither elasticity is very large in relation to the own-price elasticity " 11 , I suggest that: a) the duty may not be very efective at raising the price of the domestic good, and b) trade- diversion efects may not be very important. 1.4 Simulated Tariff Pass-Through Elasticities (PTE) In order to simulate tarif pas-through elasticities (PTE), I build up scenarios using market shares from 2002-04, 2005-07, and 2008-10, respectively. Armington elasticity ( ? !) is set alternatively to 2 (?best-bet? value), 4, and 6; the supply elasticity vector is !" = (1.1, 2.0, 2.0), our ?best-bet? estimates of these parameters. ! 14 Results suggest the AD duty does not substantialy afect the price of the protected good (table 2). For the considered parameter values, the PTE of the protected good (P 1 * /T) ranges from 0.019 to 0.13. This suggests that if all imports from Vietnam were asesed at 63.88%, the highest AD margin calculated by the US Department of Commerce (2003, 2009), the price of domestic catfish would rise by at most 8.3% (0.13 ( 63.88%). This estimate assumes Armington elasticity is 5 (the upper limit) in the scenario with S 3 = 0.14. In the scenarios with S 3 = 0.01 and S 3 = 0.11, the maximum increase of the domestic catfish price is 5.6% and 4.7%, respectively. The level of trade diversion can be evaluated by a comparison of the simulated tarif pas-through elasticities. In each case, the PTE of the domestic catfish and the non-dutied catfish are much smaler than the corresponding PTE of Vietnam?s catfish. As an example, in 2002-04, for ? !=2.5 (?best-bet? value), the domestic price of Vietnam?s catfish increases by 49%; however, the prices of the protected catfish and China?s catfish only increase by 5.2% and 4.1%, respectively. Actualy in each case, the PTE of China?s catfish is even smaler than the PTE of the protected catfish, although the diferences are not substantialy significant. Changes in the relative prices imply that the market share extracted from Vietnam?s producers due to the AD policy is shared by the US and China, noting in the Armington framework, relative market shares are asociated with relative prices through Armington elasticity. 5 The upshot is that, the market share and quasi-rents US producers enjoyed from the AD policy would have been greater without the entry of China. On the other hand, for US consumers, switching from the domestic catfish to China?s catfish should increase their benefits, considering the domesticaly produced !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 5 Our simulated results show that the market share of the non-named country is increased by 0.03% in the 2002-04 scenario, 0.5% in the 2005-07 scenario, and 1.2% in the 2008-10 scenario. Taking the initial market share of the non-dutied good into acount (0.01, 0.11, and 0.14, respectively), this confirmed a positive relationship betwen the initial market share and the degre of trade diversion. ! 15 catfish is generaly more expensive than China?s catfish (table 1). Next, I evaluate the magnitudes of the welfare efects of trade diversion for al parties in the market. 1.5 Welfare Analysis The implications of trade-diversion efects from a welfare perspective are evaluated in a partial- equilibrium seting using figure 1 as a guide. Consistent with the structural model, the diagrams depict a situation where the three goods are substitutes in consumption, but independent in production. That is, the technologies used to produce the domestic and imported goods are independent, and any specialized factors used to produce the three goods (e.g., catfish fed) are perfectly elasticaly supplied to each country. The later asumption is necesary for welfare efects to have significance in the sense that they can be traced to an identifiable group (e.g., domestic producers, se Thurman 1993). The AD shifts the supply curve for the dutied good up from S to S% in panel B by an amount equal to the per-unit duty, i.e., the ad valorem duty multiplied by the price of the dutied good in the pre-duty equilibrium. After the three markets have adjusted fully to the supply shift, the market price of the dutied good increases from P 2 to P 2 ?, and the supply price of the dutied good decreases from P 2 to P S . The full welfare efect is measured in the market for the dutied good (Just, Heuth, and Schmitz 2004, pp. 322-26). Specificaly, the hatched area betwen the price lines P 2 to P S and behind the supply curve S, labeled %PS 2 in panel B, represents the welfare loss to producers of the dutied good. The hatched area betwen the price lines P 2 to P 2 ? and behind the equilibrium or ?total? demand curve D T , labeled %CS* in Panel B, represents the welfare efect of the supply shift on the remaining parties afected by the duty. Specificaly, %CS* = %CS 1 + %CS 2 + %CS 3 + %PS 1 + %PS 3 represents the welfare loss to domestic consumers ! 16 as a result of the higher prices they must pay for domestic and imported catfish following imposition of the duty. Since the duty raises the market price of al three goods, producers of the domestic and non-dutied imported goods enjoy a welfare gain equal to the shaded areas labeled %PS 1 and %PS 3 , respectively, in panels A and C. Since a paralel upward shift in a supply curve always decreases consumer surplus, %CS* < 0, which implies that consumer losses outweigh producer gains. However, the duty provides tax revenue, which, depending on duty incidence, may be sufficient to compensate consumers and yield a net welfare gain for the US. With the maintained hypothesis that supply and demand shifts are paralel, the welfare efects of the catfish duty may be approximated using the following formulas: (25) ( ) 3,2,15.01 ** =+=! iQPQPPS iiiii (US and foreign producer impacts) (26) ( ) * 2 * 222 5.01 ~ * QPQPCS +!=" (?Consumer? impact in dutied market) (27) 31 * PSPSCSCS !"!"!=! (US consumer impact) (28) ( )( ) * 2 * 2 * 222 1 ~ QPPQPTR +!=" (US treasury impact) The P i Q i represents the value of the ith product in the initial equilibrium. Numerical values for the asterisked variables in equations (25) - (28) were computed by simulating the model (equations (2) ? (8)) for a 35% increase in the duty. In January 2009, the US Department of Commerce ruled that catfish duties imposed in 2003 would remain in place (Martin 2009). The de minimis weighted-average antidumping margin in both the 2003 and 2009 USDC rulings is 36.84% (US Department of Commerce 2003, 2009). Thus, seting T * =0.35 provides a conservative estimate of welfare impact. ! 17 Distribution efects Our first set of simulations focuses on the welfare implications of trade diversion for al parties (distribution efects). I chose the 2008-10 baseline values for prices and quantities. Simulations were run with the market share for the non-dutied good (S 3 ) alternatively to be 0.07, 0.14 (mean value for 2008-10), and 0.28 to miic the observed range in this parameter over the 2002-10 period. Although the later value is outside the range for S 3 reported in table 1, it is used to give trade diversion full play. Acordingly, the market share of the protected good (S 1 ) is alternatively set to be 0.67, 0.60, and 0.46, since the market share of the dutied good (S 2 ) is held constant at 0.26 (mean value for 2008-10). In each instance, the demand elasticities matrix is adjusted as required to ensure that demand elasticities are consistent with theory. Other parameters are set to be their ?best-bet? values, the domestic supply elasticity ? ! 1 =., the import supply elasticities ? !=( 2 , 3 ), and Armington elasticity ? !=2.5. The simulating results are reported in table 3. Focusing first on the case where S 3 = 0.14, results indicate the largest beneficiary of the duty is the US treasury, which gains $29.2 milion. Although domestic producers gain $8.1 milion, this comes at the expense of domestic consumers, who lose $28.3 milion. Adding together these efects yields a net gain to the domestic economy of $9.0 milion. A gain occurs because import supply is sufficiently upward- sloping to permit the US to act as a monopsonist and price discriminate against foreign suppliers via the imposition of a tarif (Enke 1944). Specificaly, for the considered parameter values, domestic consumers bear 51% of the duty, which means 49% of the US treasury gain comes from Vietnam producers. Vietnam producers lose $18.1 milion, while China producers gain $1.5 milion, for a net welfare loss to al afected parties of $7.7 milion. The upshot is that the AD did more to punish foreign producers than to reward domestic producers, a common result in the AD ! 18 literature for farm products (e.g., Asche 2001; Brester et al. 2002; Kinnucan 2003; Kinnucan and Myrland 2006). Turning to trade-diversion efects, rent disipation asociated with trade diversion increases with the market size of the dutied good, but the increase is modest. An increase in the market share of the non-dutied good from 0.07 to 0.14 causes the domestic producer surplus to decline by 12%. The corresponding decline is 33%, when the market share of the non-dutied good is increased by 4 times from 0.07 to 0.28. Thus, for the considered parameter values, trade diversion has a moderate efect on rent disipation. Trade diversion has an even les important effect on consumer surplus. When the market share of the non-dutied good increases from 0.07 to 0.28, consumers? loss is dwindled by a mere 3.2%, too smal to mater. Although domestic consumers switch to the les expensive China?s catfish when the domestic price of the protected good rises, their benefits from the switch are diluted by the huge loss from the rising price of the dutied-good due to an inward supply shift. This is consistent with the simulated tarif pas- through elasticities (PTE). As shown in table 2, in each scenario, the PTE of the protected good is much greater than the corresponding PTE of the protected good and the non-dutied good in magnitude. Linking producer surplus to consumer surplus, I calculate the value of the redistribution eficiency of the AD duty, defined as the ratio of domestic producers? gain to domestic consumers? loss (in absolute value). Litle of the lost quasi-rent rebounds to the benefit of domestic consumers, noting that the redistribution eficiency is reduced from 0.32 to 0.22 when S 3 is increased from 0.07 to 0.28. The upshot is that, for the considered parameter values, the trade-diversion efects are confined largely to domestic producer impacts, and even then they are ! 19 les important than other factors that afect duty eficacy, namely substitution efects and import supply elasticities, as shown in our following set of simulations. Sensitivity Analysis The foregoing results are based on the ?best-bet? values of Armington elasticity and the import supply elasticities. To gauge the sensitivity of results to the elasticity of substitution, I set ? ! alternatively to 1.5, 2.5 (?best-bet? value) and 5, and the import supply elasticities =( 2 , 3 ) alternatively to 2 (?best-bet? value), 4 and &. In these simulations, market shares are set to their average values for 2008-10 (S = (0.62, 0.24, 0.14)), the domestic supply elasticity to its ?best- bet? value ? ! 1 =.. The demand elasticity matrices corresponding to the diferent Armington elasticities are: (29) N= -.20.3.7 1 ..-.4 ! " # $ % & (! = 1.5) (30) = -.60.39.2 1 ..-. ! " # $ % & (! = 2.5) (31) N= -2.6.05.6 39 .1.-4. ! " # $ % & (! = 5.0) Results suggest Armington elasticity and import supply elasticities are pivotal in determining both the size of the welfare impacts and their distributional consequences (table 4). An increase in Armington elasticity ( ? !) has more substantial effects on US producers than on other parties. In the instance where = 2 , 3 ), the US producer surplus is increased 1.3 ! 20 times when ? ! is raised from 1.5 to 2.5. The corresponding price increase fals to 67% when ? ! is increased from 2 to 5. Import supply elasticities also have a substantial impact on producer surplus. Using ? !=( 2 , 3 )as the standard of comparison, if import supply is perfectly elastic, the benefits of the duty for domestic producers are overstated by a factor of 1.6 (= $5.85 milion / $3.56 milion) when ? 1.5, and by a factor of 2.2 (= $30.3 milion / $13.5 milion) when ? !=5. Thus, the erroneous treatment of import prices as exogenous causes a significant upward bias in the estimated gain to domestic producers, and this bias increases as the goods become closer substitutes. 6 Turning to distributional impacts, the national welfare gain from the duty converts to a loss for an import supply elasticity as smal as 4.0. If ? !=, the national welfare gain from a 35% duty is betwen -$10.5 milion and -$25.5 milion, with the later estimate corresponding to the upper limit of the substitution elasticity ( ? 5). When ? !, domestic consumers bear the full brunt of the duty, with losses ranging from -$39.8 milion to -$49.3 milion. Above al, if import supplies are perfectly elastic, the duty in esence represents a transfer from domestic consumers to domestic producers and the US Treasury, and the incidence of the duty is shifted entirely to the US consumers, indicating a null efect of AD on the welfare of foreign producers. For the redistributive eficiency of the duty (domestic producer gain / domestic consumer loss), it increases as import supplies become more price elastic, and as the goods become closer substitutes. For the considered parameter values, this ratio never exceds 0.61 (= $30.3 milion / $49.3 milion); for ?best-bet? parameter values, the ratio is 0.29 (= $8.11 milion / $28.3 milion). !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 6 The study by War (208) treats import price as exogenous when developing an analytical expresion for the pas-through elasticity, while the study by Muhamad et al. (2010) treats import price as exogenous when evaluating the welfare impact of the catfish duty. ! 21 Tarif Rate and Trade Diversion Since trade diversion makes the AD policy les efective in terms of producer surplus, net national welfare, and redistribution eficiency, a low tarif rate (i.e. AD margin) may be proposed to reduce trade diversion because it limits the benefits for the non-named countries to increase their exports to the US. 7 But, on the other hand, a low tarif rate would directly reduce the revenue gain for the government and reduce the extent to which the price of the protected good rises, provided that the tarif pas-through elasticity is constant. To examine the correlation betwen the tarif rate and trade diversion, I recomputed the welfare distribution with alternative AD rates. After implementing in 2003, the rate of the AD duty against Vietnam?s catfish ranges from 37% to 64%. Thus, I set the tarif rate with alternative 15%, 35%, 50%, and 70%, where 50% is equal to the inverse of the ?best-bet? import supply elasticity. The supply elasticities and the Armington elasticity are set to their ?best-bet? values ? !=(1.,2) and ? !=2 to permit isolation of the market-share efect. Prices and quantities are mean values for 2008-10. The market share of the non-dutied good is increased from 0.07 to 0.28 in three steps. Results indicate producer surplus and consumer surplus (in absolute value) are increasing functions of the tarif rate (table 5). When the tarif rate is increased, consumers? loss from the increasing price of the dutied good overwhelms the positive efects of the demand outward shift of the protected good and the non-dutied good. In the instance where S 3 = 0.14, using T* = 35% as the benchmark, if the tarif rate is increased to 50%, gains to domestic producers and the US treasury are improved by $3.6 milion and $3.1 milion, respectively; however, the loss for domestic consumers is also increased by $ 9.9 milion, resulting in a decline in net national !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! " !Konings, et al. (201) relate the lower amount of import diversion in Europe to lower duty levels, which limit the benefits of protection for the non-named countries. ! 22 welfare (%W us ). If the tarif rate is raised to 70%, the huge negative efect on domestic consumers results in a negative net national welfare. Finaly, I focus on the national welfare gain to justify trade-diversion efects on the tarif rate. In each scenario, the national welfare gain (%W us in table 5) is not a monotonic function of the tarif rate. In the instance where S 3 = 0.14, the national welfare is raised by 35% when the tarif rate is increased from 15% to 35%, but it is reduced by 13% when the tarif rate keeps rising to 50%. Moreover, the national welfare becomes negative when the tarif rate is set at 70%. The simulated results further demonstrate the tarif rate of 35% sems to be the ?optimal tariff? in each case. In the cases with T* = 35%, the national gain is $9.7 million when S 3 = 0.7, $9.0 milion when S 3 = 0.14, and $7.5 milion when S 3 = 0.28. The negative relationship betwen the maximal national welfare and the market share of the non-dutied good coincides with the previous findings that the trade-diversion effect has a stronger consequence of dwindling the domestic producer surplus than increasing consumer surplus. In the instance with T* = 70%, the national welfare loss is $3.5 million when S 2 = 0.7, and $8.4 million when S 2 = 0.28. This implies that the break-even point of the tariff rate at which %W us = 0 is negatively related to the extent of trade diversion. In other words, the greater trade diversion, the faster the national welfare converges to zero when the tarif rate is growing. The upshot is that the tarif rate tends to be lower when the extent of trade diversion is greater in order to obtain an ?optimal? gain in national welfare or avoid a nil or negative welfare efect. 1.6 Concluding Comments Despite the tarif rate as high as 64%, the antidumping duty the United States imposed in 2003 on catfish imports from Vietnam failed to prevent a precipitous decline in domestic industry ! 23 market share (from 84% in 2003 to 58% in 2010). It is tempting to ascribe AD inefectivenes to the entry of China as a new foreign competitor in 2004, as consumers substitute out of the dutied good and into both the domestic good and the non-dutied good. This switch may also reduce the loss to consumers, considering that the non-dutied good is generaly les expensive than the protected good. Study results suggest trade diversion is an epiphenomenon. Specificaly, China?s ability to capture 28% of the domestic market following AD implementation is estimated to have eroded the quasi-rents that domestic producers received by at most 33%. On the contrary, trade diversion cannot reduce consumers? loss substantialy since the pas-through elasticity (PTE) of the dutied good (related to a supply shift) is much greater than the PTE of the protected good and the non-dutied good (related to demand shifts). The simulated duty incidence suggests, for the considered parameter values, about one half of the duty appeared as a rise in the US price of the dutied product. Further, the extent to which trade diversion influences the efectivenes of an AD duty depends on the market share of the domestic good. In an Armington framework, a dominant market share for the domestic good, or ?home bias,? means that the cross-price elasticities of the protected good tend to be smal. As an example, in 2005 when US producers enjoyed a market share of 0.83, the cross-price elasticities in question consistent with an Armington elasticity of 2.5 are " 12 = 0.20 and " 13 = 0.06. 8 Even in 2010 when US producers still dominated the market but with a smaler share of 0.58, the corresponding cross-price elasticities are " 12 = 0.87 and " 13 = 0.15. This indicates: (i) the efect of an AD on the price of the protected good (tarif pas- !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 8 The complete demand elasticity matrices coresponding to the 205 and 2010 market shares given in table 1are N 205 = -1.6.20.6 3 ..-.4 ! " # $ % & and N 201 = -..87. 463 ..-. ! " # $ % & . It may be noted that the " 12 element in N 2005 matrices is close to Muhamad et al.?s (2010, p. 437) estimate of this parameter (0.18), even though in their study Armington restrictions were not imposed. ! 24 through elasticity, i.e. equation (10)) wil tend to be smal, and (ii) the atenuation of the tarif pas-through elasticity of the domestic good caused by trade diversion is tiny (equation (17)). Moreover, the trade-diversion efect on demand for the protected good is even weak, leading a modest efect of trade diversion on both price and volume of the protected good. The implications of these results would sem to be straightforward. First, in terms of net national welfare, the higher the extent of trade diversion, the smaler the tarif rate, ceteris paribus. A high tarif rate punishes domestic consumers more, since the tarif incidence is borne by the domestic consumers and the targeted foreign suppliers, given an upward sloping import supply curve. The side-efect of a large tarif rate is to exaggerate import diversion, which has a primarily negative efect on domestic producers? welfare. On the other hand, trade diversion has little consequence of reducing consumers? loss. This implies that a tarif rate beyond some limit may reduce the net national welfare. Second, the inverse relationship betwen domestic industry market share and the tarif pas-through elasticity of the protected good indicates that an industry with a large market share at duty inception can expect to be disappointed in that quasi-rents wil be modest. To the extent the protected industry?s market share continues to decline after AD implementation (as was the case for catfish), removal of the AD wil induce larger economic consequences than its insertion. This asymmetry may help to explain the tendency for ADs to remain in place long after the initial petition was filed (Bown 2007). ! 25 Chapter II. Import Demand under Price and Exchange-Rate Uncertainties ! 26 2.1 Introduction In a study of price uncertainty, Wolak and Kolstad (1991) point out that fluctuations in exchange rates might be one of the several inherent risk sources in actual import prices. In terms of exchange risk, while some studies argue for a negative efect on agricultural trade (e.g. Anderson and Garcia 1989; Cho, Sheldon, and McCorriston 2002; Wang and Barret 2007; and Kandilov 2008), others advocate a positive relationship (se Langley, Giugale, Meyers, and Halahan 2000; Awokuse and Yuan 2005). More recently, by employing panel data, Erdema, Nazlioglub, and Erdemc (2010) argue that the increased variability of exchange leads to more reduction in imports of agricultural goods than in exports in Turkey. Two implications can be generated from the previous research. First, the agricultural industry is more sensitive to risk factors in that agricultural goods are typicaly traded with flexible pricing strategies, and they are les storable than manufactured products (Wang and Barret). Second, since the trade efect of the exchange uncertainty is asociated with properties of the market, the impact of exchange risk should be evaluated in the context of disaggregate data (McKenzie 1999). But, disaggregate agricultural trade markets, in most cases, are perfectly competitive, indicating that the import price risk may reflect al information like the exchange-rate risk. If the exchange-rate risk completely pases through into the import price risk, the trade efect of the export price risk and the exchange-rate risk should be equal. One purpose of the present paper is to test the equivalence of the exchange risk efect and the export price risk efect. Compared to exchange-rate risk, price risk has received les atention in the literature on agricultural trade. When studying the exchange-rate risk, Klasen (2004) claims that price is predictable due to contract constraints. Although such a viewpoint is appropriate for industrial goods, it is not applicable for agricultural trade where product diferentiation is weak and firms ! 27 are more numerous (Carter and Gunning-Trant 2011). A search of the literature results in only two studies that test the impact of price risk on agricultural trade flows: Seo (2001) and Muhamad (2011), which are extended from Wolak and Kolstad (1991). Taking Chinese wheat imports market as an example, Seo investigates the relationship among the expected price, the systematic risk of price, and monopolistic power of the exporters. Muhamad develops a diferential demand system by incorporating import price uncertainty and finds evidence that the UK carnation importing firms are more responsive to price risk than to the expected price. In the research of either agricultural trade or non-agricultural trade, even les atention has been given to the diferences betwen influences of exchange risk and price risk in trade flows. Ignoring the diferences betwen price risk efect and exchange risk may also lead to conflicting empirical evidence on the trade efect of exchange risk. The diferential efects of price and the exchange rate on trade may contribute to non-equivalence of efects of exchange risk and price risk. Two notable exceptions of which I am aware discuss and evaluate the relationship betwen price risk and exchange risk. Cushman (1983) states that uncertain price as wel as exchange rate leads to percentage changes in real exchange rate, implying uncertainty in real exchange rate should be included in the trade models. Cushman?s specification is more applicable for aggregate trade rather than disaggregate commodities that are of interest in the present paper. Kawai and Zilcha (1986) explicitly take exchange rate and commodity-price uncertainties into acount when examining the optimum behavior of a risk-averse international trader; however, they do not develop an empirical analysis. The inconclusive research results on exchange volatility may be also partly atributed to the flaws of its microeconomic foundation since the normaly employed import demand equation in the previous literature is not derived from a utility function or is based on a too restrictive ! 28 utility function. As De Grauwe (1988) ilustrated, the positive trade efect of exchange risk is plausible given a general utility function underlying the import demand equation. Given a sufficient concave utility function facing importers, an increase in exchange risk might cause firms to raise import demand in order to avoid the worst possible outcome. The main purpose of this research is to examine whether price and exchange volatilities have detectable efects on import demand by applying a demand system upon the utility theory. This question is of interest because the optimal-decision of importers should distinguish the impacts of price risk and exchange risk owing their diferent properties and diferent availabilities of hedging instruments. In addition, the linkage betwen risk impact and price efect needs to be revealed theoreticaly in order to analyze the empirical results. Few empirical studies combine export price risk and exchange risk, although previous research has extensively explored the trade efect of the exchange risk with ambiguous results. The US salmon import market is taken as an empirical case because salmon is one of the major seafoods imported by the US, and the export earning is a crucial instrument for one of the major suppliers, Chile, to restore balance of payments. The main contribution of the present research is to distinguish exchange risk from price risk in a diferential demand system and to examine impacts of those uncertainties on import demand. Implications of the research are an alternative method to handle diferent sources of uncertainty in the demand system and a complementary explanation based on the underlying utility function. This study is structured as follows. I start with the theoretical framework followed by model specifications and data. After estimating the variances of price and exchange rate, I ! 29 describe the regresion results for the US salmon imports demand system. The paper concludes with brief remarks and implications. 2.2 Theoretical Model The import demands for a particular good is derived from Armington?s (1969) method, which asumes consumers employ a multistage budgeting procedure for alocating expenditure among competing sources. First, total expenditure is alocated over broad groups of goods based on a weakly separable utility function or a utility tree framework. Second, expenditure on this particular good is then alocated betwen the domestic and imported varieties. Finaly, expenditure on imports of this particular good is divided among various source countries. A complete demand system built upon a utility function can reveal determinants of trade patern adequately. 9 This demand system can further demonstrate the degree of market power that is contingent on the substitutability of goods from diferent sources (Galaway, Blonigen and Flym 1999 p. 217). As Seo (2001) noted, the research of Wolak and Kolstad (1991) is limited to only homogenous goods, causing an omision of the supplier?s influence on price risk owing product diferentiation. The Rotterdam odel, which is derived from an implicit utility function, has been widely used to demonstrate the agricultural trade patern (e.g. Seale, Sparks, and Buxton 1992; Washington and Kilmer 2002; Muhamad 2009, 2011). Among them, Muhamad (2011) is the first study applying the Rotterdam model to examine the trade efect of risk variables. Diferent from Muhamad (2011), in the present research, I justify theoreticaly that uncertainty factors !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 9 The conditional demand systems can be derived from either the consumer demand theory or production theory. As Washington and Kilmer (202) point out, there are no empirical differences in estimates of conditional elasticities between consumers demand theory and production theory. ! 30 take efect through marginal utilities, and the efect depends on price efects and substitutability betwen goods. Taking volatility components into acount, the extended general Rotterdam odel can be derived from the following utility maximization problem: (1) !"# !! u = u(q, v) s.t. p?q = y where u is the utility to be maximized, q is a vector of imports from diferent source countries, p is the corresponding import price vector, v is a vector of risk variables, and y is the conditional expenditure on the imported goods of interest. The first order conditions of the utility maximization problem are: (2) ! !! ! !! = y (3) -&p i + u i = 0 i = 1, 2, ?, n. where u i = !! !! ! is marginal utility, & is the Langrage multiplier. For the maximization problem with the conditional budget constraint, a negative definite bordered Hesian matrix is the necesary and sufficient condition, implying the matrix (4) U = !! ! ! ! ! !" is negative definite. Here, u ij = !! ! !! ! . From the equations (2) and (3), the efects of price (p j ) on the ith product (q i ) and expenditure (y) are therefore given by: (5) q ij =! !!! !!! ! !" ! !! (6) q iy = ! !" ! !! ! 31 where U ij and U i are cofactors of the matrix U, |U| is determinant of this matrix. Here q ij (= !! ! !! ! ) is paralel to the Slutsky equation in Phlips (1983, p. 49). I further denote ! !" ! = ! !" !! , which is a measure of Hicksian complementary or substituting efect (Tintner 1952). In the spirit of Tintner (1952), which is recently applied in Brown and Le (2010), the impact of a volatility variable (v j ) is revealed in the demand system by diferentiating the first order equations with respect to one risk variable to yield: (7) !! ! ! ! ! !" . ! ! ! ! ! !! ! = ! ! !! ! where ! ! denotes !! !! ! , ! !! ! = !! ! !! ! represents the efect of volatility on marginal utility of the ith product, and ! !! ! is the efect of the volatility j on the ith product. Next, from (7), I solve ! !! ! : (8) ! !! ! = - ! !" !! ! !! ! !! ! Considering the equation ! !" ! = ! !" !! , equation (8) can be restated as (9) ! !! ! = - ! ! ! !!! ! !" ! ! !! ! After obtaining the results of efects of the exogenous variables (p, y, and v) on the demand for the ith good, I derive the extended Rotterdam odel by using a diferential approach (Theil 1977, 1980). First, the import demand equations are solutions to (2) and (3). (10) q i = q i (y, p, v) By diferentiating both sides of (10) and relating the parameters to (5), (6) and (9), I obtain the extended Rotterdam model in the form: (11) ! ! !!"! ! =! ! !!"! + ! !" !!!"! ! ! !! ! !" !!!"! ! ! !! ! 32 where Q = ! ! ! !!"! ! ! !!! is real expenditure, w i = ! !! ! , ' i =w i A i = p i !" ! !! , ( ij =w i ! ! ! = ! !! ! ! !" !" !! , and ) ik = ! !" ! !! ! !! ! . Here A i is expenditure elasticity, ! !" ! is Hicksian price elasticity, and ! !! ! is elasticity of marginal utility of the jth product with respect to v k . The estimated coeficients can be explicitly converted into the corresponding elasticities. Indicated by ) ik = ! !" ! !! ! !! ! , the net effect of the volatility depends on its impact on the marginal utility of each product in the group, which is further weighted by the corresponding price efect. This is consistent with what De Grauwe (1988) points out: the exchange volatility afects import demand through the consumer?s marginal utility for the good, and the direction of this efect depends on the curvature of the underlying utility function. For example, in the likely case where the own-risk (v i ) only afects the ith good () ii =! !! ! !! ! ) or the own-risk efect dominates the sign of the reduced efect of the volatility, the direction of the volatility?s impact depends solely on the efect of the ith volatility on the marginal utility of the ith good (! !! ! ), as the sign of ! !! is a priori negative. For example, the trade efect of volatility is positive when ! !! ! > 0 due to sufficiently risk-averse behavior of importers, i.e. a more concave utility function. The opposite is true if ! !! ! < 0. Hence, whether volatility exerts a positive or negative efect on trade volume depends on the direction of its efect on marginal utility. This caveat needs to be borne in mind when interpreting the empirical results. In (11), coeficients of the risk variables are in reduced form. In order to further demonstrate how volatility factors afect the market, with simple manipulation, I restate equation (11) as follows: (12) ! ! !!"! ! = ! ! !!"! + ! !" !!!"! ! ! ! !!!! ! ! !! !!"! ! ! 33 where the parameters of volatility are expresed in structural forms. This expresion reflects that change in the jth ?efective? price is the actual price change minus the summation of changes in marginal utility of the relevant jth product as a result of changes in al volatility variables in the demand system. If changes in kth volatility decrease (increase) marginal utility of the jth good, demand for the ith good would be more (les) sensitive to changes in the kth price given the ith good and the jth good are substitutable (complementary) to each other. 10 In terms of uncertainty in international trade, Balg and Metcalf (2010) and Bergin (2004) posit a risk-averse firm would atach a risk premium as an extra markup to cover the costs of exchange-rate fluctuations; Wolak and Kolstad (1991) postulate that input-price risk premium is the percentage above the current expected market price a firm would pay for riskles input supply. When implementing (12) to the empirical research of agricultural trade, I decompose the import price risk into export price risk and exchange risk to obtain the following theoretical model: (13) ! ! !!"! ! = ! ! !!"! + ! !" !!!"! ! ! !! ! !" ! !! !!!"!! ! ! - ! ! !!!" !"#!! ! ! where !! ! stands for risk of the export price measured in exporting country?s currency, and !! ! represents the exchange risk. The underlying utility function with the budget constrain represented by equation (1) indicates the general restrictions on the demand system, namely: 11 (14) ! ! ! !!! = 0 (homogeneity) (15) ( ij = ( ji (symmetry) (16) ! ! ! !! !, ! !" ! !!! = 0 (adding-up) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 10 The interpretation of the risk factor exerting its role through changes in ?efective prices? is identical to that in the advertising-augmented trade model (e.g. Dufy 195) 11 Coeficients of risk factors satisfy the ading-up restrictions through the price efects. Se Brown and Lee (202) for more detail. ! 34 Besides the general constraints, other specific restrictions can be placed on the efects of the risk variables upon the econometric tests against the unrestricted model A represented by equation (13). This results in a reduction of the parameter spaces and eficiency of regresion results. Duffy (1987) asumes no cross efect of preference variables, implying the following specific restrictions: (R1) ! !" = 0 for j $ k (zero cross-price risk efect) (R1?) ! !" = 0 for j $ l (zero cross-exchange risk efect) If Duffy?s restriction (R1) cannot be rejected, a stronger restriction (Theil 1980) that own-risk efects are identical can be further tested (for elaboration, I suppose there are 5 price risk variables and 4 exchange volatility variables in the demand system): (R2) ! !! = ! !! = ! !! = ! !! = ! !! != (constant own-risk efect) (R2?) ! !! = ! !! = ! !! = ! !! = (constant own-risk efect) Given R1 and R2 cannot be rejected, one main concern in the present paper is the equivalency of own-price risk and exchange own-risk efect: (R3) ! = (equivalency of price risk and exchange risk efects) Lastly, I test the hypothesis of risk neutrality (R4) ! = 0 (price-risk neutrality) (R4?) = 0 (exchange-risk neutrality) If R4 is rejected, the conventional Rotterdam model exclusive of risk factors is preferred. Diferent from the general restrictions that, as common treatment in the literature, can be imposed in the regresion to be consistent with the demand theory, the specific restrictions (R1- R4) are imposed only if they are compatible with the data. The above specific hypotheses are summarized as follows ! 35 Models for testing Model symbol Restrictions Model A Equation (13) with homogeneity and symmetry constraints imposed Model B Zero cross-risk efect (R1 and/or R1?) Model C A constant own-risk efect (R2 and/or R2?) Model D Equivalency of price risk and exchange risk efects (R3) Model E Risk neutrality, i.e. conventional demand system (R4) 2.3 Model Specifications and Data The forgone theoretical model is proposed to highlight the efects of export price risk and exchange risk on the US import demand for salmon, which can be diferentiated by sources. Since the introduction of salmon aquaculture in the early 1980s, the international salmon trade is growing with Chile, Canada, Norway, and the UK being the main suppliers. Among importers, the US has traditionaly been the world?s major market for salmon. Currently, Atlantic salmon is the second most imported seafood product in the US only after shrimp. Chile and Canada were the leading countries exporting farmed Atlantic salmon to the US throughout the past twenty years. In 1995, America imported salmon at the value of $277 milion, of which about 42% came from Chile and 49% from Canada (se table 1). The US salmon import surged almost 4.6 times from 1995 to 2008; however, the market was stil highly concentrated with more than 85% of total imports from Chile and Canada. During the sample period, the prices of salmon from diferent sources had diferent levels of volatility despite the stable relative price ratios (se figure 1). Subsequently, the trade efect of the importers? atitudes to price risk deserves more atention. Taking Chile as an example, the great fluctuation of price and exchange ! 36 rate did not reduce its dominant position in this market (se table 1), implying the importers might not be sensitive to the volatility of Chile?s salmon price. In this study, US salmon import data on value (CIF in US dollar) and quantity (kilogram) are from the US International Trade Commision (USITC) where import salmon is represented by 52 HTSUS-10 codes acording to diferent species and forms of salmon. Among them Atlantic salmon (fresh, frozen, filet fresh, and filet frozen) represented by 11 HTSUS-10 codes consistently acounted for about 85% of total imports during the sample period. The data period is from January 1995 to December 2008, totaly 156 observation. 12 Data on exchange-rates are obtained from the US Department of Agriculture ? Economic Research Service (ERS). The demand system for US import salmon contains five equations distinguished by import sources, namely Chile, Canada, Norway, the UK, and the rest of the world (ROW) that is an aggregation of farmed Atlantic salmon imported from countries not specified. For a particular supplier, the monthly import price is obtained by dividing the total import (CIF) value in the US dollar by quantity (kilogram). For ROW, the monthly aggregate value and quantity are used to calculate the import price. Dividing the import price (in the US dollar) by the corresponding monthly average exchange rate (foreign currency per US dollar) yields the export price in foreign currency. After creating the import price and export price, the variances of prices (and exchange rates) can be estimated. Since the US dollar is used as the representative currency for ROW, only the export price risk is measured for ROW. Subsequently, upon the theoretical model represented by equation (13), the empirical model estimated is (17) ! !! !!"#! !! = * i +! ! !!"! ! +! ! !! !" !!!"! !! ! ! !" ! !! !!!"!! !! ! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 12 The sample period is ended in 208 in order to avoid structural changes in the US salmon import market induced by the 208 economic recesion and the outbreaks of infectious salmon anemia (ISA) that has been released since later 207 (see Asche, Hansen, and Tveteras 2009). ! 37 - ! ! !!!" !!!"!! !! ! +$ i,t. where i denotes suppliers (Chile = 1, Canada = 2, Norway = 3, the UK = 4, and ROW = 5), t stands for the time subscript (monthly), w i is budget share, q i represents import volume, p i is import price (in US dollar), !! ! stands for variance of export price (! ! , in foreign currency), !! ! represents variance of the real exchange-rate !! ! , and $ i is the i.i.d. error term. The intercept a i is included in each demand equation to acount for autonomous shifts in demand due to taste change or other trend-like phenomena. Following Theil (1980), changes in real expenditure are replaced with a Divisia volume index (!!!"! ! = ! !! ! !! !!"! !! ), and finite logarithmic changes are employed to replace infinitesimal changes in the theoretical model. However, the variables are 12-month diferenced in order to acount for seasonality in demand (Le 1988) and also to save the degree of freedom. Hence, 'lnx t = lnx t - lnx t-12 ) dlnx t . Similarly, ! !! is the arithmetic mean of the expenditure share of the ith good in t-12 and t. 2.4 Measurement of Uncertainty The generalized autoregresive conditional heteroskedasticity (GARCH) model is employed to measure the price and exchange-rate volatilities. The GARCH method can fully reveal the characteristics of time-series data (Kandilov 2008) and can explicitly test whether the movement in the conditional variance of price or exchange rate over time is statisticaly significant (Patichis 2003). Those advantages lend the GARCH method to wide applications in the agricultural trade literature (for example, Wang and Barret 2007; Kandilov 2008; and Erdema, Nazlioglub, and Erdemc 2010). ! 38 For a typical AR(1)-ARCH(1,1) model in which the residuals are from a AR(1) proces, the volatility is obtained by jointly estimating the following equations: 13 (18) AR(1): E i,t = * 0 + * 1 E i,t-1 + + i,t (19) GARCH(1, 1) V(E i ) t = , 0 + , 1 ! !! ! ! + , 2 V(E i ) t-1 + - i,t where, E i is the diferential logarithm price or bilateral exchange-rate variable, V(E i ) represents the conditional volatility, and + i,t and , i,t are error terms of AR(1) and GARCH (1,1) proceses, respectively. For the GARCH(1,1) proces to be well-defined, restrictions on the estimated coeficients include - 0 > 0, - 1 + - 2 < 1, and | - 1 | < 1. Price and exchange variance estimates by country are depicted in figure 2-4. For each major supplier, the import price is more volatile than the corresponding export price. The exchange-rate variances sem to track each other reasonably wel. On the other hand, the exchange rates in general fluctuate les than either export prices or import prices. This renders equivalence of the export price risk and the exchange-rate risk efects suspect. For Chile and Canada, which succesfully captured the lion?s share of the US salmon imports, their prices and exchange rates were no les volatile than those of other suppliers. 2.5 Regresion Results The demand system of the US salmon import market contains five equations distinguished by exporting sources: Chile, Canada, Norway, the UK, and ROW. In estimation, one equation (ROW) is dropped from the system to avoid singularity. The relevant coeficients can be recovered on the basis of demand constraints. The preliminary estimation using the Semingly Unrelated Regresion (SUR) indicated the evidence of first-order autocorrelation in some !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 13 As Patichis (203) points out, the order of the AR proces has litle impact on the GARCH models. ! 39 equations. Acordingly, the General Method of Moments (GM) approach was employed to produce the Newey-West estimator in order to correct autocorrelation as wel as heteroskedasticity. Model selection and Hypothesis tests For purposes of comparison, the Rotterdam model inclusive of the import price risk variables is first estimated, and the results are displayed in table 2. 14 The results in general reflect that the diferential-based demand system is compatible with the data at a satisfactory level in that al 5 estimated expenditure coeficients and 4 own-price coeficients are significant and have the correct signs in agreement with the demand theory. In the case of Chile, the major supplier, the own-price efect is negative but not significant. Most cross-price efects are significant with a positive or negative sign. The negative cross-price elasticities betwen Chilean and Canadian salmon indicate those two varieties are complementary to each other. This contradicts the fact that there were opposite trends of market shares for Chile and Canada during the sample period. For variance estimates, in the cases of Chile, Canada, and Norway, coeficients of the export price volatilities are significantly diferent from zero, ranging from -0.6 to -1.3. However, the insignificant own-price efect in Chile?s equation and the unexpected substitutability betwen Chilean salmon and Canadian salmon render the estimated import price variance efects suspect. This leads us to the estimation of the extended Rotterdam model, where the import price risk is decomposed into the export price risk and the exchange risk. Before turning to the empirical results of the extended Rotterdam model, specific restrictions of the demand system are justified by a comparison of models with diferent !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 14 The general demand restrictions and Dufy?s restriction (zero cros-risk efects) are imposed when estimating the model. ! 40 economic hypotheses. 15 Acording to the likelihood ratio (LR) test results (table 3), I fail to reject Model C against Model B where the Duffy?s restriction is imposed, indicating the support of Theil?s hypothesis, i.e. a constant exchange risk efect. The results may relate to the common properties of data generating proceses for exchange rates, as demonstrated in figure 4. Rejection of Model C? against Model B (p-value = 0.03) indicates diferent responses of marginal utility to changes in export price risk due to diferential price signals. The information related to price should be more available for imports from major suppliers. This is further evidenced by the testing results of the hypothesis that efects of price variances for major suppliers (i.e. Chile, Canada, and Norway) are identical (p-value = 0.23). Consequently, the hypothesis of a constant export price variance efect is treated as a maintained hypothesis in the present paper. Next, I test the equivalence of import price risk and exchange risk efects by testing Model D against Model C where restrictions of a constant export price risk efect and a constant exchange risk efect are imposed. The LR test results show Model D is rejected with a p-value = 0.023. Subsequently, the conventional Rotterdam model exclusive of risk factors (Model E) is tested against Model C. Model E is firmly rejected (p < 0.0001), indicating: (1) importers are not risk neutral, reflecting risk factors should explain the observed salmon trade patern in addition to traditional variables like relative prices, and (2) the empirical results from the conventional Rotterdam odel may be misleading due to mispecification. Acording to the above test results, the remaining discussion wil rely on estimates of Model D and the results are reported in table 4, where regresion results of the conventional Rotterdam model (Model E) are also listed to highlight the diferences. Estimates of Model D are overal satisfactory in that al marginal share estimates (' i ) are significantly positive and al own-price coeficients ( ii ) are negative and significant. Chile and !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 15 I start from Model B due to the parameterizing problem. Following Phlips (1983, p55), al models have homogeneity and symmetry imposed though both properties are rejected in preliminary tests. ! 41 Canada?s regresion results have the best explanatory powers with R 2 equaling 0.48 and 0.55, respectively. With respect to expenditure and price efects, the estimated coeficients of Model D are substantialy diferent from the counterparts of the conventional Rotterdam model (Model E). In model D, both the constant export price risk efect and the constant exchange-rate risk efect are significant and are negative in sign, implying the US salmon importers are in general risk-averse. The magnitude of the exchange risk coeficient is 2 times the magnitude of the export price risk coeficient (-0.13 vs. -0.07). This indicates that the marginal utility of the imported salmon is more responsive to the exchange risk than to export price risk. Compared to the export price risk, the exchange risk is far beyond the control of the agricultural traders. The estimated coeficients of variances of the export price risk and the exchange risk are substantialy smaler than the estimated import price risk efects in the alternative Rotterdam odel. Because in the Rotterdam model the statistical significance of parameters have les economic meaning than elasticities, the remaining discussion wil focus on conditional expenditure elasticities, Hicksian price elasticities, and risk elasticities, which are estimated at the means of budget shares (table 5). Price and Expenditure Elasticities For Chile, Canada, and Norway, the expenditure elasticities are close to one, in the range betwen 0.98 and 1.16, suggesting consumer preferences for salmon from these sources are homothetic. When consumer preferences are homothetic, the market share only responds to changes in the relative price. This further implies that price and price risk play a crucial role in this market. On the other hand, imports from runners-up in this market (i.e. the UK and ROW) are les sensitive to changes in the total import expenditure. For example, a one-percentage growth in the conditional expenditure would improve imports from UK and ROW by 0.4% and ! 42 0.2%, respectively. The uneven distribution of benefits from the rising expenditure can explain the stable market share of Chile in 1995-2008 (table 1). The estimated own-price elasticities in the range betwen -0.3 in Chile and -2.4 in Norway indicate that importers respond diferently to changes in prices of salmon from diferent sources. The import demand is les sensitive to changes in major suppliers? prices and is more sensitive to changes in other suppliers? prices. Considering aggregate data on frozen and fresh salmon were employed in the regresion, those findings are somewhat in acordance with the previous research. Xie, Kinnucan, and Myrland (2009) estimate the world demand for fresh salmon is slightly elastic at 1.03 and for frozen salmon at 0.37. The greatest own-price elasticity for Norway (-2.4) is not unexpected, because fresh salmon acounted for a great proportion of imports from Norway during the data period. In most cases, the cross-price elasticities are significant and positive, suggesting competition betwen any pairs of salmon in the group of interest. The only exception is salmon from Norway and the UK, where the negative cross-price elasticities indicate complementary relations betwen these two goods. Import demand for Chile and Canada are les sensitive to prices of salmon from other suppliers. For example, a one-percentage increase in the price of Norway?s salmon would raise imports from Chile by 0.12% and raise imports from Canada by 0.29%. On the contrary, import demand for salmon from the runners-up is more sensitive to changes in prices of salmon from the major suppliers. Taking Norway as an example, a one- percentage increase in prices of Chile and Canada?s salmon would improve imports from Norway by 1.4% and 2.2%, respectively. These findings are consistent with the properties of the market where Chile and Canada jointly had the lion?s share of the market throughout the sample period. ! 43 Volatility Elasticities The estimates of uncertainty variables are central to this research. Since I cannot reject the Theil?s restriction that either the price risk or the exchange risk takes a role in the marginal utility via ?adjusted? price with a negative constant, the strength of the response in demand for the ith salmon with respect to changes in the risk factor of the jth salmon solely depends upon substitutability betwen the these two varieties. For example, in Canada?s equation, an increase in variance of Chile?s salmon price would expand Chile?s ?adjusted? price, which consequently efects the demand for Canada?s salmon to some extend, depending on the magnitude of import demand elasticity of Canada?s salmon with respect to Chile?s price. Uncertainty from the own-currency realignments exerts a significantly negative efect on imports from Chile (-0.04), Canada (-0.08), Norway (-0.32), and the UK (-0.11). These results are compatible with the findings in Wang and Barret (2007) and Kandilov (2008). Thus, the estimate results reject De Grauwe?s hypothesis (1988) that an increase in the exchange risk would cause firms to import more in order to avoid the worst possible outcome. Compared to exchange risks, export price risks have les substantial impacts on the trade flows. The variances of the own-price have negative efects in al cases, with the derived elasticities ranging from - 0.02 in Chile and -0.18 in Norway. With regard to the export price risk elasticities, the diferent response of the import demand is likely to be explained by market shares, transportation costs, and availability of information. This is evidenced by a greater tolerance for the price volatilities of salmon from Chile and Canada, which are major suppliers in this market, and at the same time, have a shorter distance to the US than any other suppliers, including Norway and the UK. Diferent from own-risk elasticities, which have consistently negative signs, the cross- risk efects are ambiguous, depending on substitutability betwen products. Demand for salmon ! 44 from Chile is les significantly influenced by the price volatilities of Canada?s salmon and Norway?s salmon, with a modest magnitude of 0.005 and 0.009, respectively. Demand for Canada?s salmon is more sensitive to Chile?s price volatility than the response of Chile?s salmon demand to Canada?s price volatility (0.008 vs. 0.005). The same conclusion can be made on the comparison of exchange-rate volatility elasticities. The diferent responses to risk factors may be correlated with changes in the market where Chile gained an 11% increase of shares from 1995 to 2008, and Canada lost 13%. Two arguments further shed light on trade efects of the risk variables. First, demand elasticities with respect to risk factors are generaly smal in absolute values, since the impacts of risk variables are weighted by price efects, which are, in most cases, are not strong. Second, the substantial diference betwen export price risk and exchange-rate risk may explain why, in some cases, the Rotterdam model inclusive of the import price risk failed to capture significant efects of own-price variance. 2.6 Conclusions Analogous to the treatment of preference variables in the literature (Brown and Le 2002, 2010), I augmented risk factors into the Rotterdam model in which risk factors afect marginal utilities via ?adjusted prices?. This methodology is appealing since it is commonly recognized that risk- averse firms are supposed to add a proportional markup on the realized or expected prices. When the extended Rotterdam odel inclusive of risk factors is employed to evaluate the trade patern, one hypothesis deserved more atention: the equivalence of export price risk and exchange risk efects. In the literature, few endeavors have been made to examine the combined efect of export-price risk and exchange risk, especialy regarding agricultural trade where trade participators are presumed to be more risk-averse. ! 45 The empirical evidence given above appears to sustain the conjecture that export price risk and exchange risk take significant efect on the US import demand for farmed Atlantic salmon, although they difer from each other. The diference betwen efects of export price risk and exchange risk may be asociated with abilities of importers to control the uncertainty in the market. Despite being significant in most cases, the elasticities of price risk and exchange risk are tiny. This can be explained by price inelastic demand curves and weak substitutability betwen products from diferent source countries since, theoreticaly, risk factors take efect on trade flows by changing marginal utility, which is further weighted by price efects. Those findings are consistent with the features of the US farmed salmon import market, where the two largest suppliers (Chile and Canada) acounted for about 85% of the total salmon throughout the sample period, although their prices and exchange-rates had a great degree of fluctuation. In general, relative price advantages, availability of substitutes, and consumer preferences are likely to be the main factors resulting in the observed trade patern. ! 46 Chapter II. A Risk-Augmented Cointegrating Import Demand System ! 47 3.1 Introduction The main purpose of this paper is to examine the role of price risk in the alocation of import expenditures across exporting sources by taking the US codfish market as an empirical application. The agricultural industry is more sensitive to price risk in that agricultural goods are les storable than manufactured products and are traded with flexible prices (Wang and Barret 2007), and agricultural product diferentiation is weak and firms are more numerous (Carter and Gunning-Trant 2010). Trade data from the US International Trade Commite (USITC) reveals that China replaced Canada as the major supplier of cod to the US after 2004. In addition to the traditional variables like relative prices, I want to evaluate the extent to which the import price risk can explain the observed trade patern. In the research of agricultural trade, litle atention has been paid to the influence of price uncertainty in trade flows. Seo (2001) and Muhamad (2011) are two exceptions. Taking Chinese wheat imports market as an example, Seo investigates the relationship betwen the expected price, the systematic risk of price, and monopolistic power of the exporters. Muhamad develops a diferential demand system inclusive of import price uncertainty and finds the evidence that the UK carnation importing firms are, for the most part, risk-averse. In a study of price uncertainty, Wolak and Kolstad (1991) point out that fluctuations in exchange rates might be one of the inherent risk sources in actual import prices. The significant relationship betwen exchange rate risk and import demand for agricultural commodities is verified in Langley et al. (2000); Cho, Sheldon, and McCorriston (2002); and Kandilov (2008). Considering the numerous trade firms, the US codfish market can be presumed to be reasonably competitive, indicating that the import price expresed in US dollars should contain al the relevant information such as exchange volatility. ! 48 In the above trade literature, few studies are based on a complete demand system with the exception of Muhamad (2011). This lack of atention to demand interrelationships may cause biased empirical results. As Alesie and Kapteyn (1991 p. 404) state, ?One obvious omited factor in micro-studies is the interdependence of preferences.? In a wel-integrated international market like codfish, the trade impact of the price risk of a product from a particular country can be offset by the price risk of other similar goods through a substitute or complementary efects. Moreover, the system-wide approach to demand system facilitates the joint testing of theoretical restrictions due to the underlying utility function (Duffy 1987), and evaluating the preference variable like the price risk that is the main concern in the present paper. Diferent from Muhamad (2011), where a diferential approach in line with the Rotterdam demand model is applied, in the present paper I build risk factors into the Almost Ideal Demand System (AIDS) of Deaton and Muelbauer (1980). The extended AIDS model inclusive of risk factors can distinguish the ?competitive efect? and ?baseline efect? of price risks, which is important to demonstrate the risk preference of importers, as I wil se later. Another diference of the present paper compared with Muhamad (2011) is that I acount for the stationarity of data generating proces when estimating the empirical model. On the basis of a great deal of compeling reasons and evidence from previous research, the price terms may be endogenous in the typical demand system for agricultural goods. Furthermore, the price volatility can also be deemed to be jointly determined with prices and expenditure shares, considering contracts betwen importers and leading suppliers can strongly stabilize the price. Therefore, in the present paper, the empirical model is estimated by applying a cointegrating-based Vector Error Correction Model (VECM) approach, in which the ?normalization? of the demand equation provides a useful framework to alow for endogeneity of ! 49 either price or volatility. Duffy (2003) investigates the advantages of cointegrating long-run estimates compared to static equilibrium estimates from other regresion methods like the Semingly Unrelated Regresion (SUR). As Reziti and Ozanne (1999) state, the cointegration technique can further improve the estimation of the theoretical model with regard to the adjusting mechanism of the market. This view is cited and echoed by Granger (1999, p. 16): ?The clasical approach to constructing a model starts with a sound, internaly consistent, economic theory which provides a tight specification for the empirical model. This model is then estimated and interpreted. Unfortunately, this strategy towards modeling has not always proved to be a succesful one. Models produced in this way often do not fit the data in various important directions. As one pair of applied economists put it, ?a recurring problem in empirical studies of consumer and producer behavior is that the regularity properties implied by microeconomic theory have more often than not been rejected? (Rezili and Ozanne (1997)), who then go on to say ?such rejection means that empirical work loses a good deal of its theoretical credibility.? They point out that a major problem is ?the failure of static equilibrium theory to acount for dynamic aspects of static equilibrium theory to acount for dynamic aspects of consumer and producer behavior? and show how the introduction of dynamics into an equilibrium model, by use of a stricture known as an error-correcting model, leads to clear improvements.? In what follows, the US codfish import market is briefly overviewed. Next, I discuss the theoretical framework. Afterwards, the empirical models are established upon the theory, followed by the measurement of price volatility and estimation methods. The later includes estimation results and analysis. The final section consists of summary and implications. 3.2 Background The US codfish import market is selected to implement the empirical study. For the last 20 years, this market has been relatively concentrated with Canada, China, and Iceland as the major ! 50 suppliers; however, expenditure shares alocated by US importers to the main suppliers varied dramaticaly over recent decades. Canada and Iceland have historicaly been the largest exporters to the US, but since 2004 China has dominated this market. At the same time, the total (CIF) value of codfish import reduced by 45 percent in the period from 1989 to 2010. The huge variation in this market makes the codfish an insightful case to the role of price risk in addition to traditional determinants (expenditure and prices) in the observed trade patern. Further, the suppliers include a developing country (China) and a developed country (Canada) that have diferent exchange rate regimes and diferent transportation costs, leading to diferent movements of prices. Kandilov (2008) claims that trade efects of exchange volatility are remarkably larger on agricultural exports from the developing countries than from the developed countries. Codfish, or cod, is a species of deep-sea fish common to the North Pacific Ocean and North Atlantic Ocean. Acording to USITC, from 1989 to 2010, the US has imported codfish from 80 countries, among which the leading countries include Canada, China, and Iceland. On average, the annual imports (CIF) of codfish from those countries acount for about 70% of total codfish imports. In the USITC database, there are 25 HTS-10 codes for codfish responding to diferent species and forms. Among them filet frozen, filet fresh, fresh, frozen, and dried codfish represented by 8 HTS-10 codes consistently acounts for about 90 percent of total imports. For the top suppliers, Canada has dominated dried codfish during the sample period. Canada and Iceland shared the fresh cod market, and China explored and controlled the filet frozen codfish market where other top suppliers have not substantialy entered. In this paper, the major species and forms of codfish are aggregated to construct the demand system upon a two- stage budgeting proces. ! 51 Over the last twenty years, US codfish import volumes have experienced dramatic changes. In total, the US imported 474 milion US dollars value of codfish in 1989; however, the total value sharply declined to 262 milion US dollars in 2010. This sharp decline happened in the early years of the period, and the variation of total imports has become relatively weak since 1994. The low fluctuation of total imports does not suggest a stable market share for individual suppliers. As shown in table 1, imports of codfish from China grew faster after 1999, resulting in China?s dominant role in this market since 2004. Betwen 1989-1999, on average, the annual CIF value of Canadian codfish acounted for 41% of the total codfish imports and Chinese codfish only 2.6%. However, in the period of 2000-2010, the positions for these two countries were reversed in the market (17% vs. 44%). The observed trade patern can be explained by changes in importers? budget and relative prices (table 1, figure A1). The four price series have a decidedly upward movement and sem to track each other reasonably wel. However, the price of China?s cod increased a leser degree with a comparison of prices of Canada and Iceland?s cod. Figure A1 further implies an interesting fact that there is a negative relationship betwen expenditure share and variation of price. Before 2004 when Canada dominated the codfish market, the movement of its price was weak; however, Canada?s price fluctuated strongly when its expenditure share was decreasing after 2004. For China, the degree of price variations has become significantly smaler since China was changing from a runner-up to a leader in this market after 2004. Two implications can be generated from this observation. First, the variance of price may be one of the determinants to explain the trade patern. Second, expenditure share and movement of price might jointly determine each other, indicating the plausibility of applying the cointegrating technique in the empirical analysis. ! 52 3.3 The Theoretical Model Wolak and Kolstad (1991) postulate that an input-price risk premium (or cost of risk) is the percentage above the current expected market price a firm would pay for riskles input supply. In terms of elements of price risk like exchange fluctuations, Balg and Metcalf (2010) and Bergin (2004) posit a risk-averse firm would atach a risk premium as an extra markup to cover the costs of uncertainty. Therefore, a change in price risk would influence the ?efective? price of the good (?competitive efect?). On the other hand, risk-averse importers are perceived to define a baseline plan due to maintenance of cooperation and diversification. The baseline expenditure patern can be modified by the information about price risk (?baseline efect?), though it might be independent of price. For a risk-averse firm, the competitive efect of risk is expected to be negative, but the baseline efect may be positive or negative depending on the firm?s atitudes to risk. If the firm is sufficiently risk-averse, the positive baseline efect may be greater than the negative competitive efect and dominates the sign of combined efect of price risk. As De Grauwe (1988) postulates, it is likely that an increase in price risk might cause firms to import more to avoid the worst possible outcome. Starting from a traditional AIDS model, I demonstrate how to incorporate the competitive efect and the baseline efect of price risk into this demand system. Following Deaton and Muelbauer (1980), the importer?s expenditure function is first defined in the form: (1) log c(u, p) = a(p) + u b(p) where c( . ) is a cost (expenditure) function, u is the utility function, p is a vector of price, and (2) a(p) = logP = a 0 + !!"#! ! + !! ! !" ! !"#!!!"#! !!! ! 53 (3) b(p) = !! ! ! . where a 0 , a k , !, !, and ! !" ! are parameters. Applying Shephard?s lema and defining ! !" !!!! !" ! ! !" ! !, the general AIDS demand system with the expenditure share (w i ) as the dependent variable can be derived from specification (1) - (3), (4) w i = a i + ) i log (x/P) + ! !"! !"#! !! where a i can be explained as the baseline portion of imports from the ith country, and x is the total expenditure. Risk-averse importers should take the risk components (v) into acount when alocating expenditure across suppliers to maximize utility or minimize expenditure. First, the competitive efect of risk factors implies that the utility function can be modified by incorporating multiplicative scaling factors. (5) u = u(q*) where ! ! ! = q i m i and m i is a scaling factor. As common in the literature on preference variables in consumption literature (for example, Duffy 1995), m i is endogenized by asuming that it varies with the level of price volatility of the ith product in a constant elasticity formulation, i.e. m i = ! ! ! ! where - i is negative and is greater than unity in absolute value if importers are risk-averse. Since the budget constraint for importers is not efected by importers? atitudes to price risk, I can define an ?efective price? (or ?adjusted price?) as ! ! ! = p i /m i , indicating the expenditure function can be specified as (6) c(u, p, v) = c(u, p*) Acordingly, (1) can be restated as: (7) log c(u, p*) = a(p*) + u b(p*), ! 54 where p* is a vector with the components like ! ! ! . Second, the baseline efect means price risk takes efect on the baseline expenditure shares, which can be performed via adjusting the price index since the baseline imports are originaly related to this index (Duffy 1995). Consequently, decomposing price risk efect into the competitive efect and the baseline efect indicates an updated price index: (8) log P* = a 0 + !!"#! ! ! ! + ! !" !"#! ! ! ! !"#! !! + !! ! !" ! !"#! ! ! !!"#! ! ! !! Hence, replacing equation (1) and (2) with (6) and (7) results in the risk-augmented AIDS demand system in the form: (9) w i = a i + ! !"!" !"#! !" + ) i log (x/P*) + ! !" !!"#! ! !!"#! !! Rearranging terms in (9) yields the theoretical model: (10) w i = a i + ) i log (x/P*) + ! !" !"#! !! + ! !"!" !"#! ! where ! !" (=! !" !! !" !) is the combined efect of price risk. For risk-averse importers, the competitive efect of own-risk is negative since ! ! is negative and ! !! is also negative due to the law of demand. Considering |- i | >1, if ' ii . 0, the parameter of own-risk in equation (10), i.e.! !! , should be greater than the own-price efect (! !! ) in absolute value. Therefore, the baseline efect is positive as long as the estimated ! !! is bigger than & ii in absolute value. The opposite is true if ! !! is smaler than & ii in absolute value. For cross efects, the sign of ! !" (i $ j) depends on substitutability betwen goods i and j, indicating an undetermined sign of the cross-risk efect. 3.4 Empirical Model and Data The empirical model of the US codfish import demand is specified by incorporating an error term into (10) to obtain: (11) w i,t = a i + ) i log (x t /! ! ! ) + ! !" !!"#! !! ! !! +! ! !" ! !! !"#!! !! + u i,t ! 55 where i represents supplier (Canada = 1, China = 2, Iceland = 3, and ROW = 4); t stands for the time subscript (monthly); u i is an i.i.d. error term; and P*, as a common in the literature (e.g. Duffy 2003), is measured in the Stone form: (12) log! ! ! = ! !! ! !! !"#!p j,t . Al other terms and variables are as previously defined. Only price volatilities of cod from the top 3 suppliers, i.e. Canada, China, and Iceland, are taken into acount in the demand system. Except for the major suppliers, there were lots of runners-up in the US codfish market, so the fluctuation of the price of ROW was smoothed during the sample period, as confirmed in figure A1. Furthermore, as discussed in the later sector, the statistical test rejected the existence of a conditional volatility of the ROW price. To satisfy properties of the demand system, namely homogeneity, symmetry, and adding- up, the following restrictions are imposed: (13) ! !" ! !!! = 0 (homogeneity) (14) # ij = # ji (symmetry) (15) ! ! ! !! ! , ! ! ! !! = ! !" ! !!! = ! !" ! !! = 0 (adding-up) After estimating the demand system represented by (11), the estimate parameters can be used to derive the Marshalian elasticities (16) ! !" ! = ) i / ! ! (income elasticity) (17) E ii = -1 + / ii / ! ! - ) i (own-price elasticity) (18) E ij = / ij / ! ! - () i ! ! !) / ! ! (cross-price elasticity) (19) ! !" ! = 0 ij / ! ! (price-risk elasticity) where ! ! is the expenditure share of commodity i in the base year ( to be discussed in detail). ! 56 Monthly cod import data from January 1989 to December 2010 are employed to estimate the risk-augmented AIDS model represented by (11). The data are from the US International Trade Commite (USITC) and the US National Marine Fishery Sources (NMFS). Import values are in the US dollar on a cost-insurance-freight (CIF) basis and import quantities are measured in unit of kilograms, resulting in the unit of prices in US$ per kilogram (dividing value by quantity for each month). For a particular month, expenditure share of the ith country (w i ) is calculated by dividing import value for country i by total import value in this month. Further, price and expenditure time series are normalized to one (with respect to the base year 2004) in order to minimize the estimating problems due to the Stone Index approximation (Pashardes 1993, Moschini 1995). 3.5 Price Volatility Before estimating the import demand system, the moments of import price distribution must be estimated. The methods of measuring risk have evolved over time, and they can fal roughly into two categories: the unconditional volatility represented by the moving standard deviation and the conditional volatility represented by the generalized autoregresive conditional heteroskedasticity method (GARCH). While a clearly dominant method has not yet emerged in the empirical research (Bahmani-Oskooee and Hegerty 2007), the conditional volatility has become more popular after the advent of cointegrating analysis because: (i) the GARCH model acounts wel for the heavy tails of the distribution of the original variable (Kandilov 2008); (ii) this method alows for time-varying conditional variance (i.e. volatility clustering) in the original variable (Wang and Barret 2007); and (iii) the GARCH model can test whether the movement in the conditional variance of a variable over time is statisticaly significant or not. These ! 57 advantages have led GARCH-type models to be very eficient in evaluating price uncertainty (Rezitis and Stavropoulos 2011). Given the correlation betwen Canada?s price, China?s price, and Iceland?s price, the Multivariate-GARCH method (M-GARCH) rather than the univariate GARCH method is applied in the present paper. One condition for the implement of the GARCH-type model is the existence of an ARCH efect in the price dynamics. I first test the individual and joint ARCH efects of the three codfish import prices by applying the Lagrange Multiplier (LM) method. 16 The sample value is larger than the LM critical value at the 5% level of significance for each individual test and the joint test (Table A1). Thus, the M-GARCH method is employed in the present paper to obtain the proxy for price risk. The M-GARCH model is expanded from the univariate GARCH approach by acounting for the conditional correlation betwen innovations from diferent price data generating proceses. In the paper, I focus on uncertainty in the prices of cod from Canada, China, and Iceland, indicating a 3-dimensional vector GARCH (1,1) proces such that (20) V t = a 0 + A ! !! !! + B V t-1 where V t is a vector (3(1) of conditional price variance, ! !! !! is a vector (3(1) of the squared error terms from mean regresions, and others are parameters to be estimated. Given diferent asumptions of the conditional correlations, there are many types of M-GARCH specification in the literature. In the present paper, I apply the DCC-GARCH model of Robert Engle (2002) since it presumes time-varying conditional correlations, and is les restrictive. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! #$ !For an individual ARCH test, I first regres one price variable on its lags to obtain the residual. Then I regres residual on m lags, and ases joint significance of coeficients of lag variables. If the coeficients are diferent from zero based on LM test then the null of conditional homoscedasticity can be rejected. ! 58 Figure 1 presents the estimated conditional variance for each price series by applying the DCC-GARCH approach (table A2). Canada and China?s prices had high degrees of variation throughout the period. Consistent with the patern of price movements reflected in figure 1, Canada?s price became more volatile in the later period; however, China?s price had more fluctuation in the early period. For Iceland, except for several extreme observations, the price volatility had a considerably low degree of fluctuation. Extreme observations also emerged in the estimated Canadian and Chinese price variances. In order to avoid distortion impacts of extreme observations when justifying the impact of price volatility, I compute the mean values of price volatility variables for 2000-2002 when China?s share began to rise and for 2008-2010 when China finaly dominated the market. By a comparison of those two periods, Canada?s price volatility increased as high as 5.7 times, and Iceland?s price volatility increased 7.4 times. On the contrary, China?s price volatility reduced by 70%. Intuitively, price uncertainty could be one of the crucial determinants explaining the cod trade patern during the sample period. 3.6 Econometric Procedure In a multivariate time series context, the first step is to test the stationarity of each variable. Properties of the data series generating proceses dictate the methods to verify a common stochastic trend among the data series. Given the existence of cointegration among variables, the long-run relation betwen variables can be estimated by applying the Vector Error Correction Model (VECM). If any linear combination of these variables fails to result in common stochastic trends, the SUR estimates of the demand system ay be appropriate after transforming variables to be stationary acording to the properties of the data generating proceses. ! 59 Prior to the cointegration test, the stationarity properties of variables are evaluated by implementing the Augmented Dickey Fuller (ADF) test. The order of augmentation in the ADF test is determined by the Akaike Information Criterion (AIC). If the null hypothesis cannot be rejected, I can conclude that the data series is integrated of order one, I(1). The outcomes of the ADF test suggest (table A3), at the conventional percent level, the null hypothesis cannot be rejected for almost al variables, indicating most variables included in the models are integrated of order one. After evaluating the stationarity of each variable, atempts are made to test for the presence of common stochastic trends among the variables. As most variables included in the demand system are I(1), the Johansen procedure is implemented to test for cointegration in the import demand expresion. This method is based on an unrestricted vector autoregresive regresion (VAR) approach, and hence al variables incorporated in the model are asumed to be endogenous. The test results produced from the Johansen trace and eigenvalue tests are reported in table 2. Both the trace test and the eigenvalue test indicate five cointegration relationships betwen variables in the demand system, which contradicts the expected ranks of three cointegration spaces upon the demand theory. With a further inspection of the test results, the sample values of the null hypothesis r = 5 in the eigenvalue test and the null hypotheses r = 5 and r = 4 in the trace test are only marginaly larger than the critical value at the 5% level. On balance, therefore, the predictions from the theoretical framework and the moderate support from the econometric test results ensure the three cointegrating relationships betwen variables in the demand system. Since the Johansen test, which is based on an unrestricted VAR approach, just confirms the cointegrating space, economic constraints need to be imposed on VAR to obtain the ! 60 identified long-run relationship. At the same time, the short-run adjusting mechanism is controlled by an error correction proces. Thus, the VECM approach is employed to estimate the demand system represented by equation (11). Subsequently, I restate the demand system in a VCEM form: 17 (21) 'Y t = *)`Y t-1 + . 1 %Y t-1 + ? + . p %Y t-g + ?D t + U t where Y t is a 11(1 vector of al variables in the demand system, i.e. w 1,t , w 2,t , w 3,t , log(x t /! ! ! *), logp 1,t , logp 2,t , logp 3,t , logp 4,t logv(p 1,t ), logv(p 2,t ), and logv(p 3,t ); D t is a vector of dummy variables utilized to control seasonal adjustments in the trade patern; g represents the number of lags for the first diference variables. Here, * is the loading matrix (11(3), matrix ) (11(3) has the long- run coeficients, . i (11(11) is short run parameters matrix, and ? is a matrix of coeficients of dummy variables. As usual, U t is a vector of error term. Considering the existence of three cointegrating relationships in the unrestricted VAR model, I need 3(3 = 9 constraints to identify each demand equation (Iootty, Pinto Jr, and Ebeling 2009). Hence, besides three normalizing constraints, three homogeneity constraints and three symmetry constraints are imposed on VECM when estimating the long-run economic demand relationship. 3.7 Regresion Results Table 3 reports the estimated long-run coeficients ()?s) of the US codfish import demand system expresed in the VECM specification, i.e. equation (21). For the purpose of comparison, the SUR estimates of the static demand system are also regresed (Table A4). The VECM !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! #" !The Akaike criterion proposed a lag order of 3 while the Schwarz information criterion and Hannan-Quinn criterion proposed a lag order of 1, so the lag order of 1 is chosen in order to avoid unstable estimate results after considering 11 variables in the system. ! 61 estimated coeficients are more significant and larger in magnitude when compared to the SUR estimates. Moreover, there is strong evidence of autocorrelation in the residues of SUR regresion. Therefore, the remaining discussion wil rely on the regresion results from VECM. The regresion results overal are satisfactory in that most of the estimated parameters are statisticaly significant. Consistent with the trade theory, the estimated coeficients of own-price are significant with negative signs for al countries, suggesting that an improvement in competitivenes (a lower relative price) yields an increase in expenditure share. The 3 own-risk efects, in the range betwen -0.11 and -0.18, are significantly negative, implying the competitive efect dominates the sign of the risk factor. What is an interesting finding is that, in each case, the coeficient of own-price is greater than the corresponding coeficient of own-risk, indicating, as discussed in the previous sector, the baseline efect of the price risk is positive. Both positive and negative coeficients of cross-risk factors are captured in the regresion, and 9 out of the 12 cross efects are substantialy diferent from zero. Moreover, when the cross-price efect is positive (negative), the corresponding cross-risk efect is also positive (negative). Given a positive expenditure efect, the positive cross-price efect indicates the two goods are substitutes for each other. Thus, a rise in the fluctuation of one good?s price would decrease its marginal utility and consequently improve the marginal utility of its substitute, leading to a positive efect of the cross-price variance, and vice versa for complementary goods. Considering the lack of economic meaning of the estimated coeficients in the demand system, I derive expenditure elasticities (! ! ! ), price elasticities (E ij ), and volatility elasticities (! !" ! ) by employing suppliers? expenditure shares in the base year, 2004. The estimated elasticities are displayed in Table 4. ! 62 Imports from Canada, China, and Iceland are more sensitive to changes in total imports. For example a one-percentage increase in the conditional expenditure would increase imports from Canada and China by 2.0% and 4.5%, respectively. The expenditure elasticity of Iceland?s cod is negative (! ! ! = -4.5), suggesting that imports from Iceland would increase when the total expenditure decreases. Since 1995, when China entered into the US codfish market substantialy (expenditure share > 1), the total expenditure has kept a rising trend until 2008. Hence, China?s dominating market role is first due to changes in import expenditure of codfish. In general, US codfish demand is strongly sensitive to changes in price. China?s codfish has the greatest price elasticity (E 22 = -11.7), followed by Iceland (E 33 = -8.5), and Canada (E 11 = -3.9). This suggests, taking China as an example, a one-percentage decrease in China?s price would increase cod imports from China by 11.7%, holding other determinants constant. Cross- price elasticities reflect the patern commonly found in trade literature. Most of the off-diagonal elements are positive, implying gross substitutes betwen cod from diferent sources. A one- percentage increase in Canada and Iceland?s prices would raise imports from China by 1.1% and 6.0%, respectively. On the other hand, a one-percentage increase in China?s price only benefits Canada by a 2.5% increase of import volume, and a one-percentage increase in Iceland?s price would further drop Canada?s imports by 0.96%. Turning to price volatility elasticities (! !" ! ), the main concern in the present paper, al own-risk elasticities are statisticaly significant with expected negative signs. This means the competitive efect of the risk factor dominates the sign of the risk factor, though the baseline efect of the risk factor is positive. For example, a 1% increase in Canada?s price volatility is estimated to decrease the import demand for Canada?s cod by 0.51%, given the other variables are constant. For Iceland, a 1% increase in price volatility should reduce import demand for ! 63 Iceland?s cod by 0.85%. The US importers are les sensitive to the fluctuation of China?s price, considering the smalest own-risk elasticity (! !! ! = -0.32). The own-risk elasticities are much smaler than the corresponding own-price elasticities, -0.51 vs. -3.88 in the case of Canada, -0.32 vs. -11.7 in the case of China, and -0.85 vs. -8.52 in the case of Iceland. I further evaluate the extent to which price and risk may have contributed to the rising market share of China at the expense of Canada. During the sample period, the great annual growth rate of China?s share happened in 2005. China?s share increased 18% from 2004 to 2005, while Canada lost 9%. In 2005, the price of Canada?s cod was increased by 0.59%, resulting in a 2.3% decrease of volume of cod imported from Canada. For China, the price of cod was increased by 5.6%, shrinking imports from China by 66%, ceteris paribus. Unlike changes in price, Canada?s price volatility in 2005 was as large as 1.4 times of the magnitude in 2004. On the contrary, China?s price was relatively stable as the magnitude of volatility decreased by 29% in 2005. Considering the estimated own-risk elasticities, the increased price volatility would reduce imports from Canada by 69%; however, the declined volatility would increase demand for China?s cod by 9.3%, which offset the negative efect of the rising price in the same year. The import demand for codfish is also sensitive to the changes in the cross-risk factors. A one-percentage increase in volatility of China?s cod price and Iceland?s cod price would increase imports from Canada by 0.13% and 0.36%, respectively. Imports of China?s cod are only sensitive to Canada?s price volatility (! !" ! = 1.28). Codfish from Iceland only responds negatively to Canada?s price volatility, though the corresponding cross-price elasticity is not significant. Taking the annual growth rates (2004-2005) of Canada and China? price volatilities into account (136% and -29%, respectively), ceteris paribus, changes in Canada?s price volatility would increase cod imports from China 1.7 times, which completely offset the negative efect of ! 64 the rising own-price. In contrast, China?s relatively stable price would further reduce cod imports from Canada by 3.9%. Overal, the inclusion of risk factors in the demand system can adequately track the observed trade patern. In the case of China, influences of the low volatility of own-price and the high volatility of cross-prices offset the negative efect of the rising own-price and contribute substantialy to the rising trend of codfish imported from China during the sample period. By way of comparison, for Canada, in spite of a low degree of changes in the annual average price, the high fluctuation of price and the strong substitutability betwen Canada and China?s cod are the main reasons explaining the downward trend of cod imports from Canada during the sample period. 3.8 Summary and Implications In the present paper, I developed a risk-augmented Almost Ideal Demand System (AIDS) model to explore the extent to which risk factors explain the observed trade patern. In the extended model, the risk efect on import demand is decomposed into the competitive efect and the baseline efect. Taking nonstationarities in the data and endogeneity into acount, I employ a cointegrating-based Vector Error Correction Model (VECM) approach to estimate the long-run responses of imports to the changes in the one or several determinants. The resulting model is utilized to the US codfish market and the estimate results of the demand system inclusive of price volatility adequately revealed formulation of the observed trade patern. First, the estimated VECM long-run responses of importers to expenditure, price, and volatility are more significant and larger in magnitude when compared to the Semingly Unrelated Regresion (SUR) estimates, implying the static model may understate the responses ! 65 of import to changes in expenditure, price, and volatility. On the basis of the VCEM regresion results, it can be argued that in the long run, the price volatility exerts significant efects on importers? strategic decision to alocate expenditure among diferent source suppliers. For example, the low fluctuation of China?s price and the high fluctuation of the competing suppliers? prices contributed substantialy to China?s rapid rise as a major supplier of codfish imported to the US after 2004; whereas, there was a huge negative efect of the rising price on cod imports from China due to the extremely elastic demand curve. The policy implications of these results for agriculture trade would likely sem to be straightforward. Exporting countries interested in a target market should atempt to stabilize the price. However, as Wolak and Kolstad (1991) state, much of price uncertainty is induced by factors, which are out of the control of suppliers. Furthermore, the negative relationship betwen price uncertainty and expenditure share limits abilities of runners-up to reduce price uncertainty. Consequently, perhaps a more reasonable implication is related to trade policy designing which, in most cases, is esentialy based on price analysis. For example, regarding the antidumping policy, the relationship betwen the antidumping duties and price risk is investigated in Blonigen (2004), which is cited in Carter and Gunning-Trant (2011, p. 99) as follows: ?Substantial price volatility in agricultural markets often leads to higher AD margins compared with those of manufacturing, particularly when the product is highly perishable.? ! 66 References Alessie, R., and A. Kapteyn. 1991. Habit Formation, Interdependent Preferences and Demographic Effects in the Almost Ideal Demand System. The Economic Journal 101(406): 404-419. Anderson, M. and P. Garcia. 1989. Exchange Rate Uncertainty and the Demand for U.S. Soybeans. American Journal of Agricultural Economics 71(3): 721-729. Armington, P. S. 1969. A Theory of Demand for Diferentiated Products by Place of Production. International Monetary Fund Staff Papers 26: 159-178. Asche, F. 2001. Testing the Efect of an Anti-Dumping Duty: The US Salmon Market. Empirical Economics 26: 343-355. Asche, F., H. Hansen and R. Ragnar. 2009. The Salmon Disease Crisis in Chile. Marine Resource Economics 24: 405-411. Awokuse, T.O. and Y. Yuan. 2005. The Impact of Exchange Rate Volatility on U.S. Poultry Exports. Agribusines 22(2): 233-245.Blonigen, B. A. 2006. Evolving Discretionary Practices of US Antidumping Activity. Canadian Journal of Economics 37: 874-900. Bahmani-Oskooee, M., and S.W. Hegerty. 2007. Exchange Rate Volatility and Trade Flows: A Review Article. Journal of Economic Studies 34(3): 211-255. Balg, B.A. and H. Metcalf. 2010. Modeling Exchange Rate Volatility. Review of International Economics 18(1): 109-120. Bergin, P. 204. Measuring the Costs of Exchange Rate Volatility. Federal Reserve Bank of San Francisco Economic Leter 22:1-3. Blonigen, B. A. 2003. Food Fight: Antidumping Activity in Agricultural Goods. Invited Paper presented at the international conference ?Agricultural Policy Reform and the WTO: Where Are We Heading?? Capri, Italy. June 23-26. ! 67 Blonigen, Bruce. 2004. Food Fight: Antidumping in Agricultural Goods. in Agricultural Policy Reform and the WTO: Where Are We Heading? Ed. Giovanni Anania, Mary Bohman, Colin Carter, and Alex McCala (Cheltenham, UK: Edward Elgar) Bown, C. P. 2007. China?s WTO Entry: Antidumping, Safeguards, and Dispute Setlement. NBER Working Paper 13349. Cambridge, Masachusets. Brester, G. W., J. M. Marsh, and V. H. Smith. 2002. The Impacts on US and Canadian Slaughter and Feder Catle Prices of a US Import Tarif on Canadian Slaughter Catle. Canadian Journal of Agricultural Economics 50: 51-66. Brown G. and J. Le. 2002. Restrictions on the Efects of Preference Variables in the Rotterdam Model. Journal of Agricultural and Applied Economics 34,1: 17-26. Brown G. and J. Le. 2010. Preference Variable Impacts in Direct and Inverse Diferential Demand Systems. American Journal of Agricultural Economics 92(3): 898-889. Carter, C., and C. Gunning-Trant. 2010. US Trade Remedy Law and Agriculture: Trade Diversion and Investigation Efect. Canadian Journal of Economics 43(1): 97-126. Chang, Y. and P. G. Gayle. 2006. The Continued Dumping and Subsidy Ofset Act: An Economic Analysis. Southern Economic Journal 73: 530-545. Cho, G., M. Sheldon and S. McCorriston. 2002. Exchange Rate Uncertainty and Agricultural Trade. American Journal of Agricultural Economics 84(4): 932-942. Coleman, R. L. 2005. Progres in Developing the Global Seafood Trade. International Trade and Finance Asociation 15 th International Conference. Paper 3. Cushman. D.O. 1983. The Effects of Real Exchange Rate Risk on International Trade. Journal of International Economics 15: 45-63. ! 68 De Grauwe, P. 1988. Exchange Rate Variability and the Slowdown in Growth of International Trade. IMF Staf Papers 35: 63-84. Deaton, A., and J. Muellbauer. 1980. Economics and Consumer Behavior. Cambridge University Pres. Duffy, M. 1987. Advertising and the Inter-Product Distribution of Demand: A Rotterdam Model Approach. European Economic Review 31(5): 1051-1070. ???. 1995. Advertising in Demand Systems for Alcoholic Drinks and Tobaco: A Comparative Study. Journal of Policy Modeling 17(6): 557-577. ???. 2003. On the Estimation of an Advertising-Augmented, Cointegrating Demand System. Economic Modeling 20(1): 181-206. Engle. R. 2002. Dynamic Conditional Correlation: A Simple Clas of Multivariate Generalized Autoregresive Conditional Heteroskedasticity Models. Journal of Busines & Economic Statistics 20(3) (July): 339-350. Enke, S. 1944. The Monopsony Case for Tarifs. Quarterly Journal of Economics 58: 229-45. Erdema, E., S. Nazlioglub and C. Erdemc. 2010. Exchange Rate Uncertainty and Agricultural Trade: Panel Cointegration Analysis for Turkey. Agricultural Economics 41: 537-543. Galaway, M., B. Blonigen, and J., Flynn. 1999. Welfare Costs of the US Antidumping and Countervailing Duty Laws. Journal of International Economics 49: 211-244. Granger, C.W.J. 1999. Empirical Modeling in Economics: Specification and Evaluation. Cambridge University Pres. Hanson, T. and D. Sites. 2009 US Catfish Data Base. Available at http:/ww.aces.edu/dept/fisheries/aquaculture/catfish-database/catfish-2009.php (acesed February 2011). ! 69 Harvey, D. J. 2006. Aquaculture Outlook. United States Department of Agriculture, Economic Research Service. LDP-AQS-24, October. Iootty, M., H. Pinto Jr, and F. Ebeling. 2009. Automotive Fuel Consumption in Brazil: Applying Static and Dynamic Systems of Demand Equations. Energy Policy 37(12): 5326-5333. Irwin, D. A. 2005. The Rise of US Anti-dumping Activity in Historical Perspective. The World Economy 28: 651-668. Just, R. E., D. L. Hueth, and A. Schmitz.2004 The Welfare Economics of Public Policy: A Practical Approach to Project and Policy Evaluation. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing Limited. Kandilov, I.V. 2008. The Efects of Exchange Rate Volatility on Agricultural Trade. American Journal of Agricultural Economics 90(4): 1028-1043. Kawai, M. and I. Zilcha. 1986. International Trade with Forward-Futures Markets under Exchange Rate and Price Uncertainty. Journal of International Economics 20: 83-98. Kinnucan, H. W. 2003. Futility of Targeted Fish Tarifs and an Alternative. Marine Resource Economics 18: 211-224. Kinnucan, H. and ?. Myrland. 2006. The Efectivenes of Antidumping Measures: Some Evidence for Farmed Atlantic Salmon. Journal of Agricultural Economics 57: 459-477. Klasen, F. 2004. Why is It so Dificult to Find an Efect of Exchange Rate Risk on Trade? Journal of International Money and Finance 23: 817-839. Konings, J., H. Vandenbussche and L. Springael. 2001. Import Diversion under European Antidumping Policy. CEPR discussion paper no. 2785. ! 70 Langley, S.V., M. Giugale, W.H. Meyers and C. Halahan. 2000. International Financial Volatility and Agricultural Commodity Trade: A Primer. American Journal of Agricultural Economics 82: 695-700. Le, J.C. 1988. Nested Rotterdam Model: Application to Marketing Research with Special Reference to Telecommunications Demand. International Journal of Forecasting 4: 193- 206. Martin, M. F. 2009. US-Vietnam Economic and Trade Relations: Issues for the 111 th Congres. Congresional Research Service Report No. R40755. Washington, D.C. Moschini, G. 1995. Units of Measurement and the Stone Index in Demand System Estimation. American Journal of Agricultural Economics 77(February): 63?68. Muhamad, A. 2011. Source Diversification and Import Price Uncertainty. Working paper. Muhamad, A. and K.G. Jones. 2009. An Asesment of Dynamic Behavior in the U.S. Catfish Market: An Application of the Generalized Dynamic Rotterdam Model. Journal of Agricultural and Applied Economics 41(3): 745-759. Muhamad, A., S. J. Neal, T. R. Hanson, and K. G. Jones. 2010. The Impact of Catfish Imports on the US Wholesale and Farm Sectors. Agricultural and Resource Economics Review 39: 429-441. NOAH Fisheries Service. 2011. Monthly Trade Data by Product, Country/Asociation. Available at: http:/ww.st.nmfs.noaa.gov/st1/trade/monthly_data/TradeDataCountryMonth.html (acesed February 2011). Pashardes, P. 1993. ?Bias in estimating the Almost Ideal Demand System with the Stone Index Approximation.? The Economic Journal 103(419): 908-915. ! 71 Patichis, C. 2003. Conditional Exchange Rate Volatility, Unit Roots, and International Trade. The International Trade Journal 17(1): 1-17. Pesaran, M.H., Shin, Y. & Smith, R.J., 2000. Structural Analysis of Vector Error Correction Models with Exogenous I(1) Variables. Journal of Econometrics 97(2): 293-344. Phlips, L. 1983. Applied Consumption Analysis: Revised and Enlarged Edition. Amsterdam: North-Holland Publishing Company. Prusa, T. J. 2001. On the Spread and Impact of Antidumping. Canadian Journal of Economics 34: 591?611. Reziti, I., and A. Ozanne. 1999. Testing Regularity Properties in Static and Dynamic Duality Models: The Case of Greek Agriculture. European Review of Agricultural Economics 26(4): 461-477. Rezitis, A.N., and K.S. Stavropoulos. 2011. Greek Meat Supply Response and Price Volatility in a Rational Expectations Framework: A Multivariate GARCH Approach. European Review of Agricultural Economics 42: 419-435. Seale. J.L., A.L. Spark, and B.M. Buxton. 1992. A Rotterdam Application to International Trade in Fresh Apples: A Diferential Approach. Journal of Agricultural and Resource Economic Journal 17(1): 138-149. Seo, J.J.Y. 2001. Demand Diversification under Uncertainty and Market Power. Asian Economic Journal 15(4): 425-450. Theil, H. 1977. The Independent Inputs of Production. Econometrica 45(6): 1303-1327. Theil, H. 1980. The System-Wide Approach to Microeconomics. Chicago, IL: The University of Chicago Pres. ! 72 Thurman, W. N. 1993. The Welfare Significance and Nonsignificance of General Equilibrium Demand and Supply Curves. Public Finance Quarterly 21: 449-469. Tintner, G. 1952. Complementarity and Shifts in Demand. Metroeconomica 4(1): 1-4. US Department of Commerce. 2003. ?Final Determination in the Antidumping Duty Investigation of Certain Frozen Fish Filets from Vietnam.? Fact Sheet, US Department of Commerce. Washington, D.C.. Available at http:/ww.ita.doc.gov/media/FactSheet/0603/catfish_final_061703.html (acesed on 1 May 2011). US Department of Commerce. 2009. Issues and Decision Memorandum for the Final Results of the Expedited Sunset Review of the Antidumping Order on Certain Frozen Fish Filets from the Socialist Republic of Vietnam. A-552-801. Washington, D.C. 2009. Available at http:/ia.ita.doc.gov/frn/summary/vietnam/E9-2195-1.pdf (acesed 1 May 2011). US International Trade Commision. 2011. Interactive Tarif and Trade DataWeb. Available at http:/dataweb.usitc.gov/ (acesed February 2011). Wang, K.L. and C.B. Barret. 2007. A New Look at the Trade Volume Efects of Real Exchange Rate Risk. Journal of Agricultural and Resource Economic 32(2): 225-255. Warr, P. 2008. The Transmision of Import Prices to Domestic Prices: An Application to Indonesia. Applied Economics Leters 15: 499-503. Washington, A.A. and R.R. Kilmer. 2002. The Production Theory Approach to Import Demand Analysis: A Comparison of the Rotterdam Model and Diferential Production Approach. Journal of Agricultural and Applied Economics 34(3): 431-443. Whaley, J. and X. Xin. 2009. Home and Regional Biases and Border Efects in Armington Type Models. Economics Modeling 26: 309-319. ! 73 Wolak, F.A. and C.D. Kolstad. 1991. A Model of Homogeneous Input Demand under Price Uncertainty. American Economic Review 81(3): 514-538. Xie, J, H.W. Kinnucan, and ?. Myrland. 2009. Demand Elasticities for Farmed Salmon in World Trade. European Review of Agricultural Economics 36(3): 425-445. ! 74 Appendix 1: Tables and Figures for Chapter I ! 75 Table 1.1 Prices, Quantities, and Market Shares for Domestic and Imported Frozen Catfish Filets, United States, 1999-2010 a 1 = US, 2 = Vietnam, 3 = Rest of World (China mainly). Vietnam and ROW prices are measured by dividing CIF value by quantity. b Year in which antidumping duties were imposed on imports from Vietnam. Sources: Hanson and Sites (209), NOAH Fisheries (201), and USITC (201) ! ! ! " ! # " ! " " " # # ! # " # !$$$%%%%%%%%%"&'(%%%%%%%%%%!&$(%%%%%%%%%%!&""%%%%%%%%%%%!")%%%%%%%%%%%%%%%%'%%%%%%%%%%%%%%%%!%%%%%%%%%%)&$*%%%%%%%%%%)&)+%%%%%%%%)&))*% ")))%%%%%%%%%"&,"%%%%%%%%%%!&((%%%%%%%%%%!&+,%%%%%%%%%%%!")%%%%%%%%%%%%%!$%%%%%%%%%%%%%%%%!%%%%%%%%%%)&$!%%%%%%%%%%)&)$%%%%%%%%)&))*% "))!%%%%%%%%%"&(!%%%%%%%%%%!&#$%%%%%%%%%%!&#'%%%%%%%%%%%!!*%%%%%%%%%%%%%#)%%%%%%%%%%%%%%%%!%%%%%%%%%%)&,'%%%%%%%%%%)&!"%%%%%%%%)&))+% "))"%%%%%%%%%"&#$%%%%%%%%%%!&+*%%%%%%%%%%!&!'%%%%%%%%%%%!#!%%%%%%%%%%%%%+(%%%%%%%%%%%%%%%%!%%%%%%%%%%)&,"%%%%%%%%%%)&!,%%%%%%%%)&))"% !""# $ %%%%%%%%%!&'(%%%%%%%%%%(&'(%%%%%%%%%%(&)(%%%%%%%%%%%(!)%%%%%%%%%%%%%#*%%%%%%%%%%%%%%%%(%%%%%%%%%%"&*'%%%%%%%%%%"&()%%%%%%%%"&"")% "))+%%%%%%%%%"&("%%%%%%%%%%!&*#%%%%%%%%%%!&(,%%%%%%%%%%%!""%%%%%%%%%%%%%++%%%%%%%%%%%%%%%%#%%%%%%%%%%)&,"%%%%%%%%%%)&!'%%%%%%%%)&)!!% "))*%%%%%%%%%"&('%%%%%%%%%%!&+(%%%%%%%%%%!&*"%%%%%%%%%%%!"+%%%%%%%%%%%%%#(%%%%%%%%%%%%%!)%%%%%%%%%%)&,#%%%%%%%%%%)&!#%%%%%%%%%%)&)+% "))(%%%%%%%%%"&$"%%%%%%%%%%!&()%%%%%%%%%%!&'+%%%%%%%%%%%!!,%%%%%%%%%%%%%*)%%%%%%%%%%%%%##%%%%%%%%%%)&'"%%%%%%%%%%)&!(%%%%%%%%%%)&!"% "))'%%%%%%%%%"&$"%%%%%%%%%%!&*,%%%%%%%%%%!&,!%%%%%%%%%%%!)+%%%%%%%%%%%%%#(%%%%%%%%%%%%%+)%%%%%%%%%%)&')%%%%%%%%%%)&!#%%%%%%%%%%)&!'% ")),%%%%%%%%%"&$!%%%%%%%%%%!&*'%%%%%%%%%%!&'(%%%%%%%%%%%!)#%%%%%%%%%%%%%*#%%%%%%%%%%%%%++%%%%%%%%%%)&(*%%%%%%%%%%)&!,%%%%%%%%%%)&!'% "))$%%%%%%%%%"&$(%%%%%%%%%%!&**%%%%%%%%%%!&'"%%%%%%%%%%%%%$(%%%%%%%%%%%%%,(%%%%%%%%%%%%%+"%%%%%%%%%%)&*,%%%%%%%%%%)&"'%%%%%%%%%%)&!*% ")!)%%%%%%%%%"&$'%%%%%%%%%%!&*"%%%%%%%%%%!&'$%%%%%%%%%%%%%$,%%%%%%%%%%%!),%%%%%%%%%%%%%",%%%%%%%%%%)&*,%%%%%%%%%%)&#"%%%%%%%%%%)&!)% -./01%2345%6789&: ; <=;>?/?@%2A/8&%89B: C;.D1?%EF;.1 G1;. ! 76 Table 1.2 Simulated Duty Pass-through Elasticities a a The suply elasticities set to " 1 = 1.1, " 2 = 2.0, and " 3 = 2.0 b Computed using text equations (10) - (12) and (13). Price Pas-Through P 1 */P 2 * b / T* b P 1 * / T* b P 3 / T* b ! = 1.50.0380.5880.0220.017 ! = 2.50.1060.4860.0520.041 ! = 5.00.2400.3610.0870.076 ! = 1.50.0330.5860.0190.014 ! = 2.50.0920.4800.0440.035 ! = 5.00.211 0.3500.0740.064 ! = 1.50.0580.5970.0350.026 ! = 2.50.1550.5070.0790.063 ! = 5.00.3250.3990.1300.113 2002-2004 (S 3 = 0.01) 2005-2007 (S 3 = 0.11) 2008-2010 (S 3 = 0.14) Duty Pas-Through Armington Elasticity 23 !"#$%&'(&&)%*+,-,.,-/&01&-2%&3"-1,+2&45-/&6"++7!280592&:$"+-,;,-/&-0&<=>08-&)5>>$/&:$"+-,;,-,%+& "*?&-2%&@"8A%-&)2"8%&01&-2%&B0*745-,%?&C00?&D! ! E " " #$%&'(")*%%+," -+./(010(02/" 3x "4"x 5 "4"x ! 6 ." 7'012"" 7.//89:'&*;:" * 2 * 1 ~ PP b" <*(," #=10>2=12" * 2 * 2 ~ TP b" <*(,"" 7.//89:'&*;: " * 2 * 1 TP b" " ! !" 4"?@?A 1" B@??" ?@B5C" ?@!DC" ?@?CC" 5@??" ?@B5!" ?@D55" ?@?EC" C@??" ?@B55" ?@EFE" ?@?FC" L " ?@B5?" B@???" ?@B5?" " ! !" 4"?@BC " B@??" ?@B5C" ?@!DC" ?@?CC" 5@??" ?@B55" ?@D55" ?@?EC" C@??" ?@B5?" ?@EFD" ?@?F5" L " ?@BBE" B@???" ?@BBE" " ! !" 4"?@5F " B@??" ?@B5D" ?@!DC" ?@?CC" 5@??" ?@B5?" ?@D5B" ?@?E!" C@??" ?@BBE" ?@EFC" ?@?AG" L " ?@B?F" B@???" ?@B?F" . 9:2">&$2/(01"/*%%+,"2+./(010(,"0/"/2("(&"x B "4"B@B@" H I&$%*(2>"*/0=;"(2J("2K*.(0&=/"3G6L"3B?6".=>"3BE6"M0(:"? "4"5@5D".=>"{ "4"8B@" 1 "9:2"$.'N2("/:.'2"O&'"(:2">*(02>";&&>"0/":2+>"1&=/(.=(".("! 5 "4"?@5CL"0(/"$2.="P.+*2"O&'"5??F8B?@"9:2" $0>>+2"P.+*2"O&'"(:2"=&=8>*(02>";&&>L"! ! "4"?@BCL"0/"0(/"$2.="P.+*2"O&'"5??F8B?@" " " ! 77 Table 1.3 Distributional Effects of a 35% Catfish Antidumping Duty Under Alternative Asumptions About the Market Share of the Non-Dutied Good ! ! ! "# ! Tab l e 3. D i s tr i b u ti on al Effe c ts of a 35% C atfi s h A n ti d u mp i n g D u ty U n d e r A l te r n ati ve A s u mp ti on s A b ou t th e M ar k e t S h ar e of th e N on - D u ti e d G ood a Fo r de t a i l s , s e t e xt e qua t i ons ( 25) - (2 8 ) a n d a t e n d a n t d i s c u s i o n . b Ex c e p t f o r m a r k e t s h a r e s , o t h e r p a r a m e t e r s a r e s e t b e ? b e s t - be t ? va l ue s . M ar k et s h ar e o f t h e d u t i ed g o d i s h el d co n s t an t at S 2 = 0 . 2 6 , th e m e a n va l ue f or 2008 - 10. T he m i ddl e va lu e f o r th e n o n - dut i e d good, S 3 = 0 . 1 4 , i s t h e m e a n v a l u e o f t h i s p a r a m e t e r o v e r t h e 2 0 8 - 10 pe r i od. S 3 = 0.07 S 3 = 0.14 S 3 = 0.28 U .S . P roduc e rs ! PS 1 M i l $ 9.15 8.14 6.10 -33.4 U .S . Cons um e rs ! CS = ! CS 1 + ! CS 2 + ! CS 3 M i l $ -28.6 -28.3 -27.7 -3.16 U .S . T re a s ury ! TR M i l $ 29.2 29.2 29.1 -0.19 T ot a l U .S . Im pa c t ! W US = ! PS 1 + ! CS + ! TR M i l $ 9.69 8.96 7.48 -22.8 V i e t na m P roduc e rs ! PS 2 M i l $ -18.1 -18.1 -18.2 0.28 Chi na P roduc e rs ! PS 3 M i l $ 0.77 1.51 3.00 289 T ot a l Im pa c t ! W US + ! PS 2 + ! P S 2 M i l $ -7.66 -7.67 -7.69 0.36 Inc i de nc e of D ut y: U .S . Cons um e rs / T * %/ 100 0.508 0.507 0.506 -0.35 V i e t na m P roduc e rs P 2 * / T * %/ 100 -0.492 -0.493 -0.494 0.36 Re di s t ri but i on E f fi c i e nc y ! PS 1 / ! CS 1 0.32 0.29 0.22 -31.2 M a rke t S ha re of N on-D ut i e d G ood b G roup F orm ul a U ni t s P e rc e nt Cha nge 23 !" # $ % & ' ( & & ) % * + , - , . , - / & 0 1 & - 2 % & 3 " - 1 , + 2 & 45 - / & 6" + + 7 !2 80 5 9 2 & : $ " + - , ; , - / & -0 & <= >0 8 -& ) 5>>$ / & : $ " + -, ; , - , % + & "* ? & -2% & @ " 8 A% - & ) 2 " 8 % & 0 1 & - 2 % & B 0 * 7 45 - , % ? & C 0 0 ? & D ! ! E " " #$ % & ' ( " )* % % +, " -+ . / ( 0 1 0 ( 0 2/ " 3 x " 4" x 5 " 4" x ! 6 . " 7' 0 1 2 " " 7. / / 8 9: ' &* ; : " * 2 * 1 ~ P P b " <*(, " #= 1 0 > 2 = 1 2 " * 2 * 2 ~ T P b " <*(, " " 7. / / 8 9: ' &* ; : " * 2 * 1 T P b " " ! ! " 4 " ? @? A 1 " B@ ?? " ?@ B5C " ?@ !DC " ?@ ?CC " 5@ ?? " ?@ B5! " ?@ D55 " ?@ ?EC " C@ ?? " ?@ B55 " ?@ EFE " ?@ ?FC " L " ?@ B5? " B@ ??? " ?@ B5? " " ! !" 4" ? @ BC " B@ ?? " ?@ B5C " ?@ !DC " ?@ ?CC " 5@ ?? " ?@ B55 " ?@ D55 " ?@ ?EC " C@ ?? " ?@ B5? " ?@ EFD " ?@ ?F5 " L " ?@ BBE " B@ ??? " ?@ BBE " " ! !" 4" ? @ 5F " B@ ?? " ?@ B5D " ?@ !DC " ?@ ?CC " 5@ ?? " ?@ B5? " ?@ D5B " ?@ ?E! " C@ ?? " ?@ BBE " ?@ EFC " ?@ ?AG " L " ?@ B?F " B@ ??? " ?@ B?F " . 9: 2" > & $ 2 /( 0 1 " /* % % + , " 2 + . / ( 0 1 0 ( , " 0 / " /2 ( " ( & " x B " 4" B @ B @ " H I& $ % * ( 2 > " * /0 = ; " ( 2 J ( " 2 K * . ( 0 & = /" 3 G 6 L " 3 B ? 6 " . = > " 3 B E 6 " M 0 ( : " ? " 4" 5 @ 5 D ". = > " { " 4" 8 B@ " 1 " 9: 2" $ . ' N2( " /: . ' 2 " O & ' " ( : 2 " > * ( 0 2 > " ; & & > " 0 /" : 2 + > " 1 & = / ( . = ( " . ( " ! 5 " 4"? @ 5 C L " 0 ( /" $ 2 . = " P . + * 2 " O & ' " 5 ? ? F 8 B? @ " 9: 2 " $0 > > + 2 " P . + * 2 " O & ' " ( : 2 " = & = 8 > * ( 0 2 > " ; & & > L" ! ! " 4"? @ B C L" 0 /" 0 ( /" $ 2 . = " P . + * 2 " O & ' " 5 ? ? F 8 B? @ " " " ! 78 Table 1.4 Sensitivity of Welfare Effects of the Catfish Antidumping Duty to the Armington Elasticity (#) and Import Suply Elasticities (! = ! 2 = ! 3 ) a ! ! ! "# ! Tab l e 4. S e n s i ti vi ty of W e l far e Effe c ts of th e C atfi s h A n ti d u mp i n g D u ty to th e A r mi n gton El as ti c i ty ( ! ) an d I mp or t S u p l y El as ti c i ti e s ( " = " 2 = " 3 ) a a Th e d o m e s t i c s u p l y e l a s t i c i t y i s s e t t o ! 1 = 1 . a n d m a r k e t s h a r e s a r e s e t t o S 1 = 0 . 6 0 , S 2 = 0 . 2 6 an d S 3 = 0 . 1 4 , t h e m e a n v a l u e s f o r 2 0 8 - 10. b ? Be s t - be t ? e s t i m a t e s . ! = 2 ! = 4 ! = " ! = 2 b ! = 4 ! = " ! = 2 ! = 4 ! = " U .S . P roduc e rs 3.56 4.42 5.85 8.1 1 10.6 15.2 13.5 18.7 30.3 U .S . Cons um e rs -27.0 -32.3 -39.8 -28.3 -35.0 -44.3 -30.3 -39.3 -49.3 U .S . T re a s ury 32.0 28.8 23.4 29.2 24.1 13.7 25.8 17.4 -6.54 T ot a l U .S . Im pa c t 8.50 0.89 -10.5 8.94 -0.37 -15.4 8.96 -3.20 -25.5 V i e t na m P roduc e rs -15.4 -9.24 0.00 -18.1 -1 1.2 0.00 -21.1 -13.5 0.00 Chi na P roduc e rs 0.62 0.49 0.00 1.53 1.40 0.00 2.80 3.1 1 0.00 T ot a l Im pa c t -6.27 -7.86 -10.5 -7.67 -10.2 -15.4 -9.36 -13.6 -25.5 Inc i de nc e of D ut y: U .S . Cons um e rs 0.60 0.75 1.00 0.51 0.67 1.00 0.40 0.56 1.00 V i e t na m P roduc e rs -0.40 -0.25 0.00 -0.49 -0.33 0.00 -0.60 -0.44 0.00 Re di s t ri but i on E f fi c i e nc y 0.13 0.14 0.15 0.29 0.30 0.34 0.45 0.48 0.61 # = 1.5 # = 2.5 # = 5 G roup ! 79 Table 1.5 Sensitivity of Trade-Diversion Effects to Antidumping Duty Rates a ! ! ! "" ! Tab l e 5 . S e n s i ti vi ty of Tr ad e - D i ve r s i on Effe c ts to A n ti d u mp i n g D u ty R ate s a a Th e s u p l y e l a s t i c i t i e s s e t t o ! 1 = 1 . , ! 2 = ! 3 = 2. 0, an d " = 2 .5 , a ll ? b e s t - be t ? va l ue s . b Ma r k e t s h a r e o f t h e d u t i e d g o d i s h e l d c o n s t a n t a t S 2 = 0 . 2 6 , th e m e a n v a lu e f o r 2 0 8 - 10. T * = 15% T * = 35% T * = 50% T * = 70% T * = 15% T * = 35% T * = 50% T * = 70% T * = 15% T * = 35% T * = 50% T * = 70% U .S . P roduc e rs 3.89 9.15 13.2 18.6 3.46 8.14 1 1.7 16.5 2.59 6.10 8.76 12.4 U .S . Cons um e rs -13.2 -28.6 -38.6 -49.8 -13.1 -28.3 -38.2 -49.2 -12.8 -27.7 -37.3 -48.0 U .S . T re a s ury 16.3 29.2 32.3 27.7 16.3 29.2 32.3 27.6 16.2 29.1 32.2 27.5 T ot a l U .S . Im pa c t 6.96 9.69 6.83 -3.52 6.65 8.96 5.77 -5.05 6.04 7.48 3.62 -8.14 V i e t na m P roduc e rs -8.69 -18.12 -23.58 -28.71 -8.70 -18.1 -23.6 -28.7 -8.72 -18.2 -23.6 -28.8 Chi na P roduc e rs 0.33 0.77 1.1 1 1.58 0.64 1.51 2.18 3.08 1.27 3.00 4.33 6.13 T ot a l Im pa c t -1.41 -7.66 -15.6 -30.7 -1.41 -7.67 -15.7 -30.7 -1.41 -7.69 -15.7 -30.8 Inc i de nc e of D ut y: U .S . Cons um e rs 0.508 0.508 0.508 0.508 0.507 0.507 0.507 0.507 0.506 0.506 0.506 0.506 V i e t na m P roduc e rs -0.492 -0.492 -0.492 -0.492 -0.493 -0.493 -0.493 -0.493 -0.494 -0.494 -0.494 -0.494 Re di s t ri but i on E f fi c i e nc y 0.29 0.32 0.34 0.37 0.26 0.29 0.31 0.34 0.20 0.22 0.23 0.26 S 3 = 0.07 b S 3 = 0.14 b S 3 = 0.28 b G roup ! 80 Figure 1.1 Welfare Analysis of an Antidumping Duty ! ! ! "#! Figure 1: Welfare Analysis of an Antidumping Duty 28 ! 81 Apendix: Derivation of the Pass-Through Elasticity To derive the PTE, I first reduce the structural model (equations (2) ? (8)) to three equations by solving for equilibrium in each market to yield: (A1) 0 ~ )( * 313 * 212 * 1111 =!!! PPP """# (A2) 2 +) 2 T * (A3) 0)( ~ * 3333 * 232 * 131 =!+!! PPP "#"" . The equilibrium in matrix form is: (A4) (! 1 ) 1213 " 2 ()!" 2 3132 ( 3 ) # $ % & ' 1 * ! P 2 3 * # $ = 0 2 "% & ' T * Cramer?s rule is applied to equation (A4) to obtain the reduced-form elasticitiesP 1 * Tand ! P 2 * T (and 3 * ). Inserting these expresions into (A5) 1 * = PT * ! 2 " # $ % & ! 2 gives the PTE when N = 3. The corresponding expresion when N = 2 is obtained by the seting the structural elasticities in equations (A1) ? (A3) that have a 3 in the subscript to zero, and repeating the above steps. ! 82 Appendix 2: Tables and Figures for Chapter II ! 83 Table 2.1 US Salmon Imports and Market Shares by Sources Year Total Imports Value Shares (%) (mil. US $) Chile Canada Norway UK ROW 1995 277 42.1 49.0 5.5 1.9 1.5 1996 316 46.8 44.6 3.6 3.0 1.9 1997 428 43.2 50.1 2.5 2.1 2.1 1998 549 48.9 42.2 3.6 3.4 1.8 1999 689 40.5 39.0 9.6 7.0 3.9 2000 794 53.4 32.1 6.9 5.1 2.5 2001 879 51.4 39.8 4.7 3.1 1.0 2002 947 52.1 39.5 4.7 3.0 0.8 2003 1,072 58.7 28.3 5.2 6.3 1.5 2004 1,032 62.6 26.7 3.8 5.5 1.5 2005 1,184 62.1 29.2 3.3 3.8 1.6 2006 1,471 60.1 28.8 4.7 4.4 2.0 2007 1,579 58.9 26.5 5.6 6.4 2.7 2008 1,562 57.0 28.8 4.0 6.1 4.1 Average 52.7 36.0 4.8 4.4 2.1 Note: Data are obtained from USITC. ! 84 Table 2.2 GM Estimates of U.S. Import Demand System for Salmon, Rotterdam Model Inclusive Import Price Risk, 1995-2008 Monthly Data (1 = Chile, 2 = Canada, 3. Norway, 4 = United Kingdom, 5 = ROW) Variable Coef. Eq 1 Eq 2 Eq 3 Eq 4 Eq 5 Trend * i 0.0021 -0.0166 a 0.0019 0.0055 c 0.007 a (0.063) (0.063) (0.019) (0.028) (0.015) ' ln Q ' i 0.5414 a 0.421 a 0.029 a 0.0187 c -0.010 a (0.025) (0.0302) (0.0107) (0.0109) (0.08) ' ln p 1 ( i1 -0.0072 (0.0158) ' ln p 2 ( i2 -0.0313 b -0.179 a (0.0142) (0.0304) ' ln p 3 ( i3 0.0016 0.102 a -0.0549 a (0.07) (0.0168) (0.011) ' ln p 4 ( i4 0.0307 a 0.0559 a -0.028 a -0.047 a (0.010) (0.01) (0.063) (0.071) ' ln p 5 ( i5 0.0062 0.0524 a -0.0208 a -0.0116 a -0.026 a (0.05) (0.059) (0.045) (0.041) (0.035) ' ln v(p 1 ) / 1 -1.315 b (0.515) ' ln v(p 2 ) / 2 -1.0135 a (0.168) ' ln v(p 3 ) / 3 -0.616 a (0.23) ' ln v(p 4 ) / 4 -0.165 (0.154) ' ln v(p 5 ) / 5 0.180 (0.153) Note: Numbers in parentheses are asymptotic standard erors. a indicates significance at the p < 0.01 level; b indicates significance at the p < 0.05 level; c indicates significance at the p < 0.10 level. ! 85 Table 2.3 Tests of Theoretical Restrictions Model Against Number of Restrictions LR Statistic p-value Model B Model C Model B 4 10.75 0.029 Model C! Model B 3 3.96 0.265 Model D Model C 1 5.14 0.023 Model E Model C 2 55.10 <0.0001 Note: Model B restricts the cross-efect of price and exchange risks are zero (Duffy?s restriction). Model C and Model C! impose an identical non-zero efect of al own-price risk and an identical non-zero efect of al own-exchange risk (Theil?s restriction), respectively. Model D imposes the equivalence of the non-zero own-price and non-zero own-risk efects. odel E imposes the additional restriction that the identical non-zero efects of risk factor are equal to zero. For each model, symmetry and homogeneity are imposed. ! 86 Table 2.4 GM Estimates of U.S. Import Demand System for Salmon, Rotterdam Model 1995-2008 Monthly Data (1 = Chile, 2 = Canada, 3 = Norway, 4 = United Kingdom, 5 = ROW) ! "# ! Tab l e 4 . G M Es ti mate s of U .S . I mp or t D e man d S ys te m for S al mon , R otte r d am M od e l 1995 - 2008 M on th l y D ata (1 = C h i l e , 2 = C an ad a, 3. N or w ay , 4 = U n i te d K i n gd om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able 2.5 Conditional Demand Elasticities (1 = Chile, 2 = Canada, 3 = Norway, 4 = United Kingdom, 5 = ROW) Eqn Expenditure Elasticities Hicksian Price Elasticities E i y * 1i E * 2i E * 3i E * 4i E * 5i E 1 0.975 a -0.303 a 0.073 c 0.124 a 0.090 a 0.016 (0.03) (0.063) (0.04) (0.036) (0.014) (0.012) 2 1.163 a 0.107 c -0.619 a 0.291 a 0.084 a 0.139 a (0.063) (0.059) (0.07) (0.04) (0.03) (0.018) 3 0.993 a 1.350 a 2.165 a -2.41 a -0.701 a -0.393 a (0.18) (0.395) (0.31) (0.25) (0.101) (0.082) 4 0.344 b 1.085 a 0.691 a -0.776 a -0.851 a -0.160 a (0.182) (0.165) (0.245) (0.11) (0.149) (0.068) 5 0.194 a 0.412 a 2.422 a -0.920 a -0.339 a -1.574 b (0.234) (0.302) (0.31) (0.191) (0.143) (0.124) Note: Numbers in parentheses are asymptotic standard erors. a indicates significance at the p < 0.01 level; b indicates significance at the p < 0.05 level; c indicates significance at the p < 0.10 level. ! 88 Table 2.6 Conditional Demand Elasticities (1 = Chile, 2 = Canada, 3 = Norway, 4 = United Kingdom, 5 = ROW) (cont) ! "#! 5. iti ti il, 2 = Cada, 3. N United ) ! $%&! $'()&*+,-.)! $/01,+2+,+)1! ! 3+241+0&!5.+2)!$/01,+2+,+)1! ! E i y ! ! * 1i E ! * 2i E ! * 3i E ! * 4i E ! * 5i E ! ! 6! !"#$% & ! ! '!"(!( )& ! !"!$( * ! !"+,- )& ! !"!#! )& ! !"!+.! ! ! /!"!(0! ! /!"!.(0! /!"!-0! /!"!(.0! /!"!+-0! /!"!+,0! ! "! +"+.( )& ! ! !"+!$ * ! '!".+# )& ! !",#+ )& ! !"!1- )& ! !"+(# )& ! ! ! /!"!.(0! ! /!"!%#0! /!"!$0! /!"!-0! /!"!(0! /!"!+10! ! 7! !"##( )& ! ! +"(%! )& ! ,"+.% )& ! ',"-+ )& ! '!"$!+ )& ! '!"(#( )& ! ! ! /!"+10! ! /!"(#%0! /!"(+0! /!",%0! /!"+!+0! /!"!1,0! ! 8! !"(-- 2 ! ! +"!1% )& ! !".#+ )& ! '!"$$. )& ! '!"1%+ )& ! '!"+.! )& ! ! ! /!"+1,0! ! /!"+.%0! /!",-%0! /!"++0! /!"+-#0! /!"!.10! ! 9! !"+#- )& ! ! !"-+, )& ! ,"-,, )& ! '!"#,! )& ! '!"((# )& ! '+"%$- 2 ! ! ! /!",(-0! ! /!"(!,0! /!"(+0! /!"+#+0! /!"+-(0! /!"+,-0! ! :;,)).1!+&!(0.)&,?)1)1!0.)!01@=(,;,+2!1,0&*0.*!).;.1A! 0 !+&*+20,)1!1+B&+C+20&2)!0,!,?)!!D!EAE6!/)F)/G! > ! +&*+20,)1!1+B&+C+20&2)!0,!,?)!!D!EAE9!/)F)/G! 2 !+&*+20,)1!1+B&+C+20&2)!0,!,?)!!D!EA6E!/)F)/A! ! ! 89 Figure 2.1 Import US Dollar Prices of Salmon by Sources 3 4 5 6 19951996199719992000200120032004200520072008 Import Price of Chile's Salmon 4.0 4.5 5.0 5.5 6.0 19951996199719992000200120032004200520072008 Import Price of Canada's Salmon 4 5 6 7 19951996199719992000200120032004200520072008 Import Price of Norway's Salmon 3 4 5 6 7 8 9 10 19951996199719992000200120032004200520072008 Import Price of the UK's Salmon 3 4 5 6 19951996199719992000200120032004200520072008 Import Price of the ROW's Salmon ! 90 Figure 2.2 Conditional Variance Estimates of Import Prices (in US dollar): January 1995 - December 2008 0.03 0.04 0.05 0.06 19951996199719992000200120032004200520072008 Volatility of Import Price of Chile's Salmon 0.04 0.05 0.06 0.07 19951996199719992000200120032004200520072008 Volatility of Import Price of Canada's Salmon 0.08 0.10 0.12 0.14 0.16 0.18 19951996199719992000200120032004200520072008 Volatility of Import Price of Norway's Salmon 0.10 0.15 0.20 0.25 0.30 0.35 19951996199719992000200120032004200520072008 Volatility of Import Price of the UK's Salmon 0.20 0.25 0.30 0.35 0.40 0.45 19951996199719992000200120032004200520072008 Volatility of Import Price of the ROW's Salmon ! 91 Figure 2.3 Conditional Variance Estimate of Export Prices (in Foreign Curency): January 1995 - December 2008 0.04 0.06 0.08 0.10 0.12 0.14 19951996199719992000200120032004200520072008 Volatility of Export Price of Chile's Salmon 0.05 0.10 0.15 0.20 0.25 19951996199719992000200120032004200520072008 Volatility of Export Price of Canada's Salmon 0.15 0.20 0.25 19951996199719992000200120032004200520072008 Volatility of Export Price of Norway's Salmon 0.10 0.15 0.20 0.25 0.30 0.35 19951996199719992000200120032004200520072008 Volatility of Export Price of the UK's Salmon ! 92 Figure 2.4 Conditional Variance Estimate of Bilateral Exchange Rate: January 1995 - December 2008 0.04 0.06 0.08 0.10 19951996199719992000200120032004200520072008 Conditional Variance of US?Chile Exchange 0.02 0.04 0.06 0.08 0.10 19951996199719992000200120032004200520072008 Conditional Variance of US?Canada Exchange 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10 19951996199719992000200120032004200520072008 Conditional Variance of US?Norway Exchange 0.03 0.04 0.05 0.06 19951996199719992000200120032004200520072008 Conditional Variance of US?UK Exchange ! 93 Appendix 3: Tables and Figures for Chapter III ! 94 Table 3.1 US Cod Imports, Expenditure Shares, Prices by Sources YEAR Imports (mil. US $) Expenditure shares (%) Prices (US$ / Kilogram) Canada China Iceland ROW Canada China Iceland ROW 1989 474 61.2 0.2 21.0 17.5 1.4 2.8 2.1 1.3 1990 501 68.9 0.4 15.3 15.4 1.3 2.0 1.6 1.4 1991 510 59.3 0.7 18.4 21.6 1.6 1.8 2.0 1.7 1992 375 49.5 1.6 21.2 27.6 1.5 1.5 1.8 1.6 1993 291 36.2 1.9 36.4 25.4 1.5 1.6 1.7 1.5 1994 262 26.2 1.0 37.0 35.7 1.4 1.9 1.7 1.5 1995 281 24.5 3.1 33.6 38.7 1.3 1.5 1.8 1.6 1996 255 32.9 3.0 37.1 27.0 1.3 1.8 1.7 1.6 1997 318 30.0 5.4 33.6 31.1 1.4 1.7 1.8 1.6 1998 313 30.1 3.6 35.4 31.0 1.5 1.5 2.1 1.9 1999 379 30.5 7.8 36.0 25.6 1.8 1.9 2.2 2.2 2000 355 29.6 13.8 30.3 26.2 1.8 2.2 2.2 2.1 2001 290 28.5 15.7 29.4 26.4 1.7 2.1 2.1 2.0 2002 337 24.3 20.0 26.2 29.5 1.7 2.3 2.2 2.0 2003 315 24.5 23.5 23.3 28.7 1.8 2.3 2.3 2.1 2004 325 21.0 33.1 20.7 25.2 1.9 2.3 2.3 2.2 2005 327 11.8 51.1 15.6 21.5 1.9 2.5 2.7 2.4 2006 349 9.5 61.0 13.6 16.0 2.1 2.9 3.0 2.5 2007 355 8.6 71.1 8.2 12.2 2.3 3.1 3.2 2.7 2008 332 9.7 70.7 6.1 13.5 2.5 3.4 3.0 2.8 2009 250 11.8 63.4 9.5 15.3 2.4 2.7 2.6 2.2 2010 262 10.6 64.1 10.0 15.3 2.2 2.3 2.8 2.4 Sources: USITC and value data are in CIF measure. ! 95 Table 3.2 Johansen Cointegration Test Results Eigenvalue Method Maximum Eigen Method H o : H a : Eigenvalue 0.05 Critical value H o : H a : Max-Eigen Statistic 0.05 Critical value r = 0 r = 1 545.6* 358.7 r = 0 r # 1 138.3* 79.23 r = 1 r = 2 407.2* 307.2 r . 1 r # 2 93.2* 73.47 r = 2 r = 3 314.0* 260.8 r . 2 r # 3 73.6* 67.77 r = 3 r = 4 240.3* 216.6 r . 3 r # 4 62.1* 61.59 r = 4 r = 5 178.2* 178.0 r . 4 r # 5 56.5* 55.81 r = 5 r = 6 121.6 141.7 r . 5 r # 6 46.2 49.97 r = 6 r = 7 75.4 108.6 r . 6 r # 7 30.3 43.62 r = 7 r = 8 45.0 81.25 r . 7 r # 8 20.5 37.83 r = 8 r = 9 24.5 56.28 r . 8 r # 9 15.9 31.56 r = 9 r = 10 8.56 35.46 r . 9 r # 10 6.79 24.97 r = 10 r = 11 1.77 17.8 r . 10 r # 11 1.77 17.80 *: significant at the 5% level. Critical values are from Pesaran, Shin, and Smith (2000) ! 96 Table 3.3 VECM Estimates of US Import Demand System for Codfish, AIDS Model, 1989- 2010 Monthly Data (1 = Canada, 2 = China, 3 = Iceland, 4 = ROW) Variable Eq 1 Eq 2 Eq 3 Eq 4 constant 0.321*** -0.090 0.503*** 0.267*** (0.045) (0.174) (0.121) (0.033) log (x/P*) 0.212*** 1.168*** -1.151*** -0.229*** (0.068) (0.269) (0.186) (0.05) log p 1 -0.559*** 0.603*** -0.157 0.113*** (0.108) (0.086) (0.1) (0.058) log p 2 0.603*** -3.134*** 2.229*** 0.303*** (0.086) (0.342) (0.237) (0.062) log p 3 -0.157 2.229*** -1.820*** -0.252*** (0.1) (0.237) (0.188) (0.057) log p 4 0.113** 0.303*** -0.252*** -0.164*** (0.058) (0.067) (0.059) (0.048) log v(p 1 ) -0.106** 0.421*** -0.248*** -0.067*** (0.03) (0.117) (0.081) (0.021) log v(p 2 ) 0.028* -0.106* 0.062 0.016 (0.016) (0.061) (0.043) (0.012) log v(p 3 ) 0.075*** 0.149 -0.178*** -0.047*** (0.025) (0.098) (0.068) (0.018) Note: Numbers in parentheses are asymptotic standard erors. Single asterisk (*) indicates significance at the p < 0.10 level; double asterisk (**) indicates significance at the p < 0.05 level; triple asterisk (**) indicates significance at the p < 0.01 level. ! 97 Table 3.4 Derived Demand Elasticities (1 = Canada, 2 = China, 3 = Iceland, 4 = ROW) Note: Numbers in parentheses are standard erors computed via the Delta method. Note: Numbers in parentheses are asymptotic standard errors. Single asterisk (*) indicates significance at the p < 0.10 level; double asterisk (*) indicates significance at the p < 0.05 level; triple asterisk (**) indicates significance at the p < 0.01 level. ! "#! Table 4. Derived Demand Elasticities (1 = Canada, 2 = China, 3 = Iceland, 4 = ROW) Eqn Expenditure Elasticities Marshalian Price Elasticities Price Volatility Elasticities ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! 1 2.01*** -3.88*** 2.54*** -0.96** 0.29 -0.51*** 0.13* 0.36*** (0.324) (0.516) (0.412) (0.474) (0.274) (0.141) (0.077) (0.121) 2 4.54*** 1.08*** -11.67*** 6.01*** 0.03 1.28*** -0.32* 0.45 (0.814) (0.262) (1.036) (0.717) (0.202) (0.355) (0.186) (0.297) 3 -4.48** 0.41 12.42*** -8.52*** 0.17 -1.18*** 0.30* -0.85** (0.886) (0.474) (1.127) (0.897) (0.28) (0.387) (0.202) (0.324) 4 0.08 0.64*** 1.51*** -0.82*** -1.43*** -0.27*** 0.06 -0.19*** (0.2) (0.233) (0.248) (0.226) (0.192) (0.086) (0.047) (0.072) Note: Numbers in parentheses are stanar errs c te ia te lta ethod. Note: Numbers in parenthee are symptotic tndard erors.Singleasterisk (*)indicassignificanc at the p < 0.10 level; double asterisk (**)indicatessgnificance at thep < 0.5 level;trple asterisk (**) indicates significance at the p < 0.1 lvl. ! 98 Figure 3.1 M-GARCH Estimated Conditional Price Volatility 0.00 0.02 0.04 0.06 0.08 19891991199319951997199920022004200620082010 Volatility of Pcan 0.0 0.1 0.2 0.3 0.4 0.5 19891991199319951997199920022004200620082010 Volatility of Pchn 0.00 0.05 0.10 0.15 0.20 19891991199319951997199920022004200620082010 Volatility of Pice ! 99 Figure 3.A1 US Import Cod Prices by Sources 1.5 2.0 2.5 3.0 19891991199319951997199920022004200620082010 Price of Canada's Cod 1.0 1.5 2.0 2.5 3.0 3.5 4.0 19891991199319951997199920022004200620082010 Price of China's Cod 1.5 2.0 2.5 3.0 3.5 4.0 19891991199319951997199920022004200620082010 Price of Iceland's Cod 2 4 6 8 19891991199319951997199920022004200620082010 Price of ROW's Cod ! 100 Table 3.A1 ARCH Test for Price (1 = Canada, 2 = China, 3 = Iceland) Chi-Squared DF P - value ' log p 1 59 16 <0.00001 ' log p 2 56 16 <0.00001 ' log p 3 75 16 <0.00001 Joint test 312 180 <0.00001 Note: (1) the null hypothesis is no ARCH efect; (2) the lag length is 16 in the univariate test, and 5 in the multivariate test. ! 101 Table 3.A2 M-GARCH Estimates of Conditional Variance of Price (1 = Canada, 2 = China, 3 = Iceland) a i A i1 A i2 A i3 B i1 B i2 B i3 Equation 1 0.0001 0.201 0.000002 0.00003 0.7739 0.000003 0.0302 (0.0002) (0.0292) (0.0591) (0.0693) (0.0503) (0.0212) (0.0886) Equation 2 0.0001 0.00000001 0.1474 0.0001 0.000003 0.8546 1E-10 (0.0648) (0.0704) (0.0003) (0.0158) (0.0212) (0.0009) (0.1432) Equation 3 0.0002 0.0003 0.0001 0.333 0.0458 1E-10 0.637 (0.001) (0.0011) (0.0158) (0.0228) (0.0003) (0.1432) (0.1984) Note: Numbers in parentheses are standard erors computed. ! 102 Table 3.A3 Unit Root Tests (1 = Canada, 2 = China, 3 = Iceland, 4 = ROW) Variable ADF c a ADF c,t b w 1 -1.88 (5) c -2.77 (5) w 2 0.35 (8) -1.72 (8) w 3 -1.14 (8) -1.90 (8) w 4 -3.99 (2) -4.60 (2) log (x/P*) -1.9 (7) -6.83 (1) log p 1 -0.86 (8) -2.58 (8) log p 2 -2.64 (6) -3.39 (6) log p 3 -1.28 (2) -3.64 (2) log p 4 -1.39 (5) -2.68 (5) log v(p 1 ) -2.25 (2) -2.93 (1) log v(p 2 ) -1.04 (1) -3.13 (1) log v(p 3 ) -3.61 (1) -3.90 (1) Critical Value (5%) -2.87 -3.42 a ADF c test is in the form: 'Y t = 1 0 + 1 1 Y t-1 + ! ! ! !! ! !! + e t ; b ADF c,t test is in the form: 'Y t = 1 0 + 1 1 t + 1 2 Y t-1 + ! ! ! !! ! !! + e t ; c The number of laged diference terms in the test is given in bracket for each variable. The maximum order is 8. ! 103 Table 3.A4 SUR Estimates of US Import Demand System for Cod, AIDS Model, 1989-2010 Monthly Data (1 = Canada, 2 = China, 3 = Iceland, 4 = ROW) Variable Eq 1 Eq 2 Eq 3 Eq 4 constant 0.279*** 0.212*** 0.251*** 0.258*** (0.018) (0.018) (0.014) (0.033) log (x/P*) 0.321*** -0.237*** -0.041* -0.043* (0.027) (0.027) (0.02) (0.018) log p 1 -0.093* 0.046* 0.016 0.031 (0.056) (0.027) (0.045) (0.025) log p 2 0.046* -0.055* 0.017 -0.008 (0.027) (0.029) (0.021) (0.02) log p 3 0.016 0.017 0.021 -0.054 (0.045) (0.021) (0.05) (0.036) log p 4 0.031 -0.008 -0.054** 0.031 (0.025) (0.017) (0.02) (0.02) log v(p 1 ) -0.009 0.089*** -0.051*** -0.029*** (0.012) (0.012) (0.009) (0.008) log v(p 2 ) 0.027*** -0.094*** 0.037*** 0.030*** (0.007) (0.007) (0.005) (0.005) log v(p 3 ) 0.040*** -0.015 -0.010 -0.015* (0.01) (0.01) (0.008) (0.007) Note: Numbers in parentheses are asymptotic standard erors. Single asterisk (*) indicates significance at the p < 0.10 level; double asterisk (**) indicates significance at the p < 0.05 level; triple asterisk (**) indicates significance at the p < 0.01 level. The estimated coeficients of dumy variables are supresed from the report.