UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Two papers in international trade Fullerton, John Michael 1992-12-16

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
831-ubc_1992_fall_fullerton_john_michael.pdf [ 6.79MB ]
Metadata
JSON: 831-1.0086555.json
JSON-LD: 831-1.0086555-ld.json
RDF/XML (Pretty): 831-1.0086555-rdf.xml
RDF/JSON: 831-1.0086555-rdf.json
Turtle: 831-1.0086555-turtle.txt
N-Triples: 831-1.0086555-rdf-ntriples.txt
Original Record: 831-1.0086555-source.json
Full Text
831-1.0086555-fulltext.txt
Citation
831-1.0086555.ris

Full Text

TWO PAPERS INTERNATIONALTRADEByJOHN MICHAEL FULLERTONB.ScF, The University of Toronto,1980M.Sc.F. The University of Toronto,1984A THESIS SUBMITtED IN PARTIAL FULFILLMENTOFTHE REQUIREMENTS FOR THE DEGREEOFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIES(Commerce and Business Administration)We accept this thesis as confomuingTHE UNIVERSITY OF BRITISH COLUMBIAMay 1992© John Michael Fullerton, 1992to the required standardIn presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of cgmmgrce and Business Administration The University of British Columbia Vancouver, Canada Date Angn^-h 1Q. 1 QQ9 DE-6 (2/88) ABSTRACT This thesis presents two papers dealing with international trade policy in North America. The first is "An Empirical Analysis of Protectionist Forces in the United States and Canada" and the second is "Imports as a Cause of Injury: the Case of the 1986 Softwood Lumber Dispute". The first paper addresses the question of what motivates firms and industries to initiate "less than fair value" (LFV) complaints. The theory of rent-seeking is tested by analyzing data on the frequency of countervailing duty and antidumping complaints from 1975 to 1987 in both Canada and the U.S. A reduced form model is specified that expUcitly incorporates factors that are assumed to affect the supply and demand for protection, including federal elections, business cycles, and statutory changes to laws governing an tidumping and countervailing duty procedures. The results indicate that the frequency of LFV cases, in both the U.S. and Canada, rise during low points in the business cycle, and when relative competitiveness and profitability in manufacturing decline. The second paper takes a specific countervailing duty case, that of the Canada-U.S. softwood lum ber dispute of 1986, and addresses the question of whether the U.S. industry suffered injury from the al leged less than fair valued Canadian lumber imports. The relationship between various measures of injury for the U.S. industry and various hypothesized causal factors, including stumpage prices, is analyzed. The six measures of injury are: prices, output, market share, employment, accounting profits, and stock market profits. Structural models are specified and reduced form models tested for the 1975 to 1987 time period. The analysis indicates that stumpage prices, as proxied by B.C. stumpage levels, appear to have had littie effect on any injury sustained by the U.S. industry. Alternative specifications, using "leaked rents" to the B.C. industry as a proxy for the alleged subsidy, also showed little effect on the U.S. industry. Business cycle effects and the exchange rate are the most important determinants of performance. Counter-factual simulations using tiie estimated equations permit the hypothesizing of how the U.S. industry would have fared under alternative economic scenarios. This provides a useful framework for evaluating causes of injury, since one can control for all other factors besides those alleged to be unfair. TABLE OF CONTENTS Abstract ii List of Tables v List of Figures vAcknowledgement vii 1. Introduction 1 2. An Empirical Analysis of Protectionist Forces in the United States and Canada 4 2.1 Introduction 4 2.2 Literature Review 6 2.3 Review of the Macroeconomic Environment 9 2.4 Theory Development 12 2.4.1 U.S. Structural Model 3 2.4.2 Canadian Structural Model 15 2.5 Review U.S. and Canadian Trade Statutes 7 2.5.1 United States Trade Statutes 8 2.5.2 Canadian Trade Statotes 21 2.6 Choice of Study Time Period and Data Sources 24 2.7 Estimation and Results 5 2.7.1 Factors Affecting the Number of U.S. LFV Petitions per Year 25 2.7.2 Model Results for the Number of U.S. LFV Petitions per Year 27 2.7.3 Factors Affecting the Number of Canadian LFV Petitions per Year 30 2.7.4 Model Results for the Number of Canadian LFV Petitions per Year 32 2.8 Discussion and Conclusions 36 3. Imports as a Cause of Injury: the Case of the 1986 Softwood Lumber Dispute 42 3.1 Introduction 43.2 Literature Review 6 3.3 Measures of Injury 53 3.4 Relationship Between Measures of Injury 56 3.5 Empirical Overview 63 3.6 Theory Development 8 3.7 Model Specification 72 3.8 Data Description3.9 Estimation and Results 87 3.9.1.0 U.S. Softwood Lumber Price Index 88 3.9.1.1 Model results for the U.S. Softwood Lumber Price Index 89 3.9.2.0 U.S. Softwood Lumber Production 92 3.9.2.1 Model results for U.S. Softwood Lumber Production 93 3.9.3.0 Canada's Share of the U.S. Softwood Lumber Market 6 3.9.3.1 Model Results for Canada's Share of the U.S. Softwood Lumber Market 96 3.9.4.0 U.S. Forest Products Return on Sales 100 3.9.4.1 Model Results for U.S. Forest Products Return on Sales 103.9.5.0 Real Standard and Poor's U.S. Forest Products Stock Price Index 103 3.9.5.1 Model Results for the Real Standard and Poor's U.S. Forest Products Stock Price Index 104 3.9.6.0 U.S. Sawmill Production Hours 107 3.9.6.1 Model Results for U.S. Sawmill Production Hours 110 3.10 Economic Significance of Injury Determinants 114 3.11 Discussion and Conclusions 117 References 120 Appendix 2.1 U.S. LFV Data 128 Appendix 2.2 Canadian LFV Data 9 Appendix 3.1 Data Listing for Imports as a Cause of Injury 130 Appendix 3.2 U.S. Contribution to Earnings from Building Products Divisions: 1981 133 LIST OF TABLES Table 2.1 Listing of Independent Variables Used in LFV Analysis 24 Table 2.2 Regression Results: U.S. LFV Cases per Year for the Years 1975 to 1987 29 Table 2.3 Regression Results: Canadian LFV Cases per Year for the Years 1975 to 1987 33 Table 3.1 Review of Impacts on U.S. Lumber Industry Injury Measures 62 Table 3.2 Regression Results: U.S. Residential Construction Square Footage 73 Table 3.3 Canadian Regional Data and Sources Collected for Injmy Investigation 9 Table 3.4 U.S. Data and Sources Collected for Injury Investigation 80 Table 3.5 Regression Results: Canadian Regional Logging Wages Per Hour 81 Table 3.6 Regression Results: Canadian Regional Average Variable Logging Cost Per Cubic Meter 82 Table 3.7 Regression Results: Canadian Regional Sawmill Production Hours 83 Table 3.8 Representative U.S. Forest Products Companies 84 Table 3.9 Years of Inclusion in the Computation of the Annual Average Return on Sales 85 Table 3.10 Regression Results: U.S. Lumber Price Index 91 Table 3.11 Regression Results: U.S. Lumber Production 4 Table 3.12 Regression Results: Canadian Market Share 97 Table 3.13 Regression Results: U.S. Lumber Industry Return on Sales 101 Table 3.14 Regression Results: U.S. Forest Products Industry Stock Price 105 Table 3.15 Regression Results: U.S. Forest Products Industry Adjusted Stock Price 106 Table 3.16 Redistribution of U.S. Lumber Production: 1947 to 1987 109 Table 3.17 U.S. Sawmill Production and Employment 110 Table 3.18 Regression Results: U.S. Sawmill Industry Production Hours Ill Table 3.19 Economic Importance of Variables 115 LIST OF FIGURES Figure 2.1 Annual Change in U.S. GNP and Canadian GDP (%) 9 Figure 2.2 U.S. and Canadian Capacity UtiUzation Rate (%) 10 Figure 2.3 U.S. and Canadian Unemployment Rate (%)Figure 2.4 Number of U.S. LFV Petitions per Year 30 Figure 2.5 Number of Canadian LFV Petitions per Year 4 Figure 3.1.1 Canadian Lumber Market 57 Figure 3.1.2 Equilibrium ConditionFigure 3.1.3 U.S. Lumber MarketFigure 3.2.1 Equilibrium Condition: Rise in U.S. Income 58 Figure 3.2.2 U.S. Lumber Market: Rise in U.S. IncomeFigure 3.3.1 Equilibrium Conditions: Rise U.S. Dollar 60 Figure 3.3.2 U.S. Lumber Market: Rise U.S. DollarFigure 3.4.1 Equilibrium Conditions: Rise Canadian Stumpage Price 61 Figure 3.4.2 U.S. Lumber Market: Rise Canadian Stumpage PriceFigure 3.5 Plot of the U.S. Softwood Lumber Price Index 63 Figure 3.6 Plot of U.S. Softwood Lumber Production 4 Figure 3.7 Plot of Canada's Share of U.S. Softwood Lumber Consumption 6Figure 3.8 Plot of the Weighted Average Return on Sales for U.S. Sawmills 5 Figure 3.9 Plot of Standard and Poor's U.S. Stock Price Indexes: Forest Products and the 400 Top Industrials 66 Figure 3.10 Plot of U.S. Sawmill Production Hours 67 ACKNOWLEDGEMENT I would like to extend my special thanks to Dr. James A. Brander for his patience and ongoing support of my research. I also extend my thanks to Dr. llan Vertinsky, Dr. Tim Hazledine and Mr. John Howard, who, as members of my supervisory committee, helped me to improve the focus of my research. Furthermore, I want to thank all those who have been associated with the FEPA Research Unit at UBC who provided me with their support and inspired me to keep at it. Of course my efforts were made easier by family and friends. I extend my deepest thanks and appreciation to Darcie Booth, who provided advice and constant support, even during the most difficult of times. Also, thanks to Peter C. FuUerton, who never missed an opportunity to ask "have you finished your thesis yet?". Also, I want to thank Linda Keams and Greg Booth for providing food, lodging and friendship on my many trips to Vancouver to work on my thesis. I was very fortunate to receive financial support throughout my tenure at UBC. I would like to thank the Natural Sciences and Engineering Research Council, the Noranda/Bradfield Foundation and the UBC Centre for International Business Studies for their financial support. 1. INTRODUCTION This thesis presents two papers dealing with international trade policy in North America. The first is "An Empirical Analysis of Protectionist Forces in the United States and Canada" and the second is "Imports as a Cause of Injury: the Case of the 1986 Softwood Lumber Dispute". The first paper addresses the question of what motivates firms and industries to initiate countervail ing duty and antidumping complaints (or "less than fair value" (LFV) complaints). The theory of rent-seek ing would suggest that firms will tend to file complaints when it will most benefit them to do so. The hy pothesis tested in this paper is that business cycles significantly affect firms' decisions to initiate. It is hy pothesized that rent-seeking pressures will tend to be highest at the low points in the business cycle. This hypothesis is tested by analyzing data on the fi'equency of countervailing duty (CVD) and antidumping com plaints from 1975 to 1987 in both Canada and the U.S. A reduced form model is specified that explicitly incorporates factors that are assumed to affect the supply and demand for protection; for example, degree of import penetration and level of manufacturing profits. Shift variables that account for the impact of federal elections (petitions are expected to rise during federal election years) and for statutory changes to laws gov erning dumping and CVD procedures are also included. The impact of business cycle variables are explic itly accounted for. This study followed fiom an orderly specification of a structural model that relates to the theory of the firm and presents comparable structural models for the U.S. and for Canada. The results indicate that the frequency of LFV cases in both Canada and the U.S. rise during low points in the business cycle, and when relative competitiveness and profitability in manufacturing decline. While there is significant disagreement concerning the importance of various potential determinants of LFV complaints (Feigenbaum, Ortiz and Willett 1985), there is broad interest in being able to predict the timing of protectionist pressures. The second paper takes a specific countervailing duty case, that of the Canada-U.S. softwood lum ber dispute of 1986, and addresses the question of whether the U.S. industry suffered injury from the al leged less than fair valued Canadian lumber imports. Since business cycle variables are significant in whether or not a LFV case is initiated, and more cases appear to be initiated in troughs, then clearly some of the injury attributed to imports may in fact result from the basic economic conditions of the industry. The Canadian lumber industry has asserted that the injury allegedly suffered by the Americans at the time of the dispute was primarily due to external influences such as the business cycle and exchange rate changes and not a result of subsidized timber pricing poUcies as the Americans had claimed. The relevancy of this issue is clear in the light of the most recent CVD case initiated by the Department of Commerce against the Canadian softwood lumber industry in December, 1992. In 1986, a negotiated settlement was reached between Canada and the U.S. (the Memorandum of Understanding or MOU) before the ITC completed its final injury determination. In October 1992, Canada announced that it was terminating the MOU, since it felt that significant changes in its forest management pricing policies and economic conditions had taken place. The United States subsequently self-initiated another CVD inves tigation, the third against the industry in nine years. Again the allegation was that stumpage prices charged by the provinces for timber were subsidized. ^ While this analysis was carried out before this most recent CVD initiation, and does not include data from the most recent years of this dispute, the estimated relationships are likely still valid. This analysis takes various measures of injury for the U.S. industry and determines the relation ship between these measures and various hypothesized causal factors including stumpage prices. The six measures of injury are: prices, output, market share, employment, accounting profits, and stock market profits, since these are the major factors considered by the International Trade Commission (ITC) in its de termination of injury. Structural models are specified and reduced form models tested for the 1950 to 1986 time period. The analysis indicates that stumpage prices, as proxied by B.C. stumpage levels, appear to have had little effect on injury sustained by the U.S. industry. Alternative specifications, using "leaked rents" to the B.C. industry as a proxy for the alleged subsidy, also showed little effect on the U.S. industry. Business cycle effects and the exchange rate are the most important determinants of performance. Counter-factual simulations using the estimated equations permit the hypothesizing of how die U.S. industry would have fared under alternative economic scenarios. This provides a useful framework for evaluating causes of injury, since one can control for all other factors besides those alleged to be unfair. ^ The investigation also included allegations that log export restrictions constitute a subsidy by artificially lowering domestic log prices. The two papers are presented as separate, fully contained chapters in this thesis. Chapter 2 is "An Empirical Analysis of Protectionist Forces in the United States and Canada" and Chapter 3 is "Imports as a Cause of Injury: the Case of the 1986 Softwood Lumber Dispute". The Literature Cited section lists refer ences for both papers. 2. AN EMPIRICAL ANALYSIS OF PROTECTIONIST FORCES IN THE UNITED STATES AND CANADA 2.1 Introduction This paper addresses the positive question of what motivates firms and industries to initiate countervailing duty and antidumping complaints. Presumably, firms make such complaints when it is in their interest to do so. This is the basic "rent-seeking" idea: firms will seek to influence the policy process to protect or create "rents" or profits. It is, of course, not obvious what rent-seeking implies about the timing of complaints. One hypothesis is that rent-seeking pressures are more acute at low points in the business cycle. If this is so, then we should expect to observe that business cycle variables have significant explanatory power in anticipating antidumping and countervail ing duty actions. The principle objective of this paper is to investigate this hypothesis. The legislation relating to dumping and countervaiUng duties is intended to limit imports sold at "less than fair value" (LFV). Dumping and countervail cases are, therefore, sometimes referred to as LFV cases. I investigate such cases for the U.S. and Canada. Interestingly, U.S. cases trended upward much more rapidly than Canadian cases during the 1975 to 1987 period. In both countries, however, the number of cases seems to be closely related to the business cycle. The basic method of analysis in this paper is based on the empirical estimation of reduced form models of the number of LFV cases in the U.S. and Canada. The models, estimated using data for the years 1975 to 1987, fol low fi-om a specification of a structural model that explicitly incorporates the factors affecting the demand and supply for LFV protection. Basically, the reduced form models for the U.S. and Canada are based on the hypothesis that the fi-equency of LFV cases will rise when economic conditions required by dumping and countervailing duty legislation are present Thus, the models test if the number of cases rise when import penetration is high and manufacturing profits are are low. In addition, the models test the hypothesis that LFV petitions will rise during federal elections. Finally the models test the importance of statutory changes to the laws governing dumping and countervailing duty procedures. The main results of this study are that the frequency of LFV cases in the U.S. and Canada rise during low points in the business cycle and when relative competitiveness and profitability in manufacturing decline. That is, LFV cases were found to rise during general recessionary periods and fall during periods of economic growth. The close relationship between LFV cases and economic conditions that are broadly independent of imports found in this study also supports the hypothesis that rent-seeking is among the factors that motivate industries to initiate dumping and countervailing duty complaints. This is an interesting finding because it means that rent-seeking activity may be more vigorous during economic downturns and during economic upturns. This research is relevant because there is disagreement concerning the importance and magnitude of the de terminants of LFV complaints (Feigenbaum, Ortiz and Willett 1985). In addition, there is broad interest in the abil ity to predict the timing of protectionist pressures. This is because foreign firms and governments are keenly inter ested in the well being of the international trading system, and thus they want to know both when and to what extent protectionist pressures will be present given perturbations in their export markets. Furthermore, studies discussed in the literature review, such as those done by Takacs (1981), Grilli (1988), Coughlin, Terza and Khalifah (1989) and Feinberg (1989) have looked at the frequency of LFV cases in the United States. However, the research reported here is unique in that it empirically analyses the cyclical determinants of Canadian LFV cases. In addition, with the ex ception of Grilli (1988) and Coughlin et al. (1989), the selection of independent variables in studies of LFV cases has not followed from an orderly specification of a structural model that relates to the theory of the firm and has thus been quite ad hoc. The analysis in this study follows from the specification of a structural model that specifies de terminants of the demand and supply for LFV protection. Also, this study presents different, but comparable, struc tural models for the U.S. and for Canada. The time period used in this study allows (for the first time) a comparison of the U.S. Trade Act of 1974, the Trade Agreements Act of 1979 and the Trade and Tariff Act of 1984 with respect to their effect on the incidence of U.S. LFV petitions. Finally, this study is the first to empirically test the signifi cance of election years on the supply of LFV protection. The outline of the remainder of this paper is as follows. First there is a brief review of the primary litera ture dealing with the determinants of escape clause and less than fair value petitions. Following this, the paper re views the domestic economic conditions found in the U.S. and Canada during the time period studied. The paper then develops structural models and presents the reduced form models to be estimated for the U.S. and Canada. Following this, there is a presentation and discussion of trade laws in the U.S. and Canada. The sources of the vari ables used in this analysis are then presented. Finally, the paper presents the results of the analysis of U.S. and Canadian LFV cases. 2.2 Literature Review Although the literature dealing with protectionism in general and with the political economy of trade dis putes is quite large, the quantitative literature explaining the incidence of escape clause and LFV petitions consists of a relatively small number of studies dating back to the early 1980s. Furthermore, this literature deals almost exclu sively with U.S. petitions. Analyses of escape clause and LFV petitions can be divided into those which focus on the structural factors (e.g., capital intensity) of petitioning industries and those which focus on the cyclical economic factors (e.g., ex change rates) affecting the number and timing of petitions. Although similar in approach, the main questions moti vating these two areas of research are different Studies of structural determinants attempt to answer the question of which firm or industry characteristics explain the incidence of escape clause and LFV petitions over time. On the other hand, studies of cyclical determinants attempt to identify the primary economic pressures that affect the deci sion to file a petition. Good examples of studies focusing on the structural determinants of LFV cases are Finger (1981). Finger, Hall and Nelson (1982) and Feinberg and Hirsch (1989). Finger (1981) was among the first to explain the industry incidence of U.S. LFV petitions. He hypothe sized that U.S. firms used LFV mechanisms to build a public case for protection. Finger constructed two indices of LFV complaints divided into 2-digit SIC^ commodity groups filed over the life of the U.S. Trade Act The first in dex, referred to as the complaints index, was the percentage of total imports covered by complaints, and the second index, referred to as the affirmative cases index, was the percentage of total imports covered by affirmative cases. Using ordinary least squares regression (OLS), Finger found that the complaints index rose with increases in import penetration, the size of the capital stock and the level of unemployment in an industry. He also found that the affir mative cases index rose with the complaints index, the degree of product differentiation and declining domestic ship ments. However, when the reduced form models were estimated using two-stage least-squares, the complaints index became insignificant; thus, the LFV decision process may be more objective than the OLS results suggested. In a paper, using a similar data set. Finger, Hall and Nelson (1982) studied the influence of political and economic variables on the likelihood of a positive finding in U.S. LFV cases filed between 1975 and 1979. The au thors chose the case decision (affirmative or negative) in the LFV pricing and injury investigations as dependant vari-1 SIC refers to the standard industrial classification system. ables. Using logit analysis, this study found that the likelihood of a positive LFV pricing decision was related to cost disadvantages for American industries and specificity of the complaint (measured by the number of products mentioned in the complaint). The likelihood of a positive LFV injury decision was found to be related to domestic political factors (e.g., regulatory changes and size of industry affected). Feinberg and Hirsch (1989) adopted an industry rent-seeking framework to formulate a reduced form model designed to explain the incidence of U.S. LFV cases. After dividing the LFV cases for the years 1980 to 1986 into 3-digit SIC industries, this paper used tobit analysis to examine the relationship between the number of LFV peti tions and rent-seeking factors such as industry concentration, unionization, capital intensity, employment, profits and import penetration. After correcting the number of cases filed by an estimate of the number of manufacturing enterprises, Feinberg and Hirsch found that capital intensive industries with many employees suffering from high unemployment and rising import shares were more likely to file LFV petitions. They concluded that their results are largely supportive of the hypothesized rent-seeking or rent-protecting model. Good examples of studies focusing on the cyclical determinants of LFV cases are Takacs (1981), Grilli (1988), Coughlin, Terza and Khalifah (1989) and Feinberg (1989). Takacs (1981) focused on explaining the pattern of U.S. escape clause petitions over time. Using the number of petitions and the number of affmnative findings from 1949 to 1979 as indices of protectionism, Takacs used OLS regression analysis to estimate the importance of cyclical economic pressures and regulatory changes to the decision to file for and to award escape clause reUef. This study found that number of petitions per year was positively related to lower cycUcal economic activity, lower inter national competitiveness and a greater perceived likelihood of receiving protection. Furthermore, it was found that the Trade Act of 1974 discouraged fewer petitions and granted relief more often than the Trade Expansion Act of 1962. GrilU (1988) also focused on the cyclical determinants of protectionism but used the total number of non-tariff measures (LFV plus escape clause or safeguard actions) in the U.S. and in the EEC as his independent vari ables. After presenting a theory of the demand for protection, tiiis smdy used OLS regression analysis to identify the relative importance of factors theoretically linked to the frequency of petitions filed from 1969 to 1986. Grilli found that the number of petitions per year were positively related to real appreciations in exchange rates, higher unem ployment, lower production levels and a trend towards a lower level of comparative advantage. Grilli concluded that if purely structural determinants of the new protectionism are difficult to identify, the cyclical macroeconomic ones are quite clear. Coughlin et al. (1989) extended previous analyses of U.S. escape clause decisions. Based on cases filed from 1948 to 1984, this study formulates a model of a firm's decision to file a petition. Unlike previous studies, Coughlin et al. assumed that the number of escape clause petitions per year was Poisson distributed. Second, the au thors divided the number of petitions by an estimate of the number of potential petitioners to transform the cases per year into a continuous variable. The authors empirical estimation of a Poisson reduced form regression model indi cated that the frequency of petitions was inversely related to economic activity (measured in terms of capacity utiliza tion or profit levels) and the merchandise ti^ade balance. Like previous work, the Trade Act of 1974 was found to have discouraged fewer petitions than the Trade Expansion Act of 1962. Finally, the authors found a post-war trend toward more protectionist pressure. Feinberg (1989) examined the targeted country pattern of U.S. LFV petitions filed quarterly from 1982 to 1987. Using pooled tobit analysis, this sttidy focuses on the response to the absolute and relative frequency of coun try specific cases to exchange rate changes. Feinberg found that real bilateral exchange rates had a significant effect on the country-targeting of LFV cases. He concluded that the observed inverse relationship between the number of petitions and the real extemal value of the U.S. dollar was consistent with technical interpretations of LFV cases and with the view that cases were promoted by rent-seeking activities of lawyers and economists representing petitioners. Furthermore, it was concluded that the bulk of die exchange rate effect was due to cases targeted at Japanese indus-tiies. This review of the primary literature shows that most of the empirical investigations of the determinants of protection are based on U.S. conditions. The research reported here is unique in that it analyses the frequency of LFV cases per year in Canada as well as in the United States. Second, witii the exception of Grilli (1988) and Coughlin et al. (1989), die selection of independent variables has not followed from an orderly specification of a stiuctural model and has thus been quite ad hoc. This study presents a separate structural model for the U.S. and Canada that specifies the demand and supply factors to be tested in reduced form models. Furthermore, the time pe riod used in tiiis study allows for the comparison of the U.S. Trade Act of 1974, the Trade Agreements Act of 1979 and the Trade and Tariff Act of 1984 with respect to their effect on the incidence of U.S. LFV petitions. Finally, this study is the first to empirically test the significance of election years on the supply of LFV protection. 2.3 Review of the Macroeconomic Environment This section briefly reviews the macroeconomic environments in Canada and the U.S. and is included to de velop a better feeling for the relationship between cyclical macroeconomic activity and the frequency of LFV peti tions in the United States and Canada Figures 2.1 to 2.3 illustrate how similar the U.S. and Canadian economies have performed over the 1975 to 1987 period. The economic impacts of the business cycle and the oil price shocks in 1974 and 1980 can be clearly seen in these three figures. Figure 2.1 shows the annual percentage change in real U.S. GNP and Canadian GDP, and like most of the world's economies, it illustrates how the U.S. and Canada were still suffering the effects of the 1974-1975 oil price shock. However, by the end of 1975, the U.S. and Canadian economies had started to improve. Figure 2.1 Annual Change in U.S. GNP and Canadian GDP (%) 8.00 irt CO I- ir 00 CO CO 03 o o) o O) • U.S. Real GNP • Can. Real GDP As illustrated in Figure 2.2, the post recession expansion period lasted until 1979. However, although eco nomic growth was relatively strong and capacity utilization high, unemployment remained high due to a growing work force. Figure 2.3 illustrates how the rate of unemployment in the U.S. and Canadian stayed high tiirough out the 1975 to 1979 period. Figure 2.2 U.S. and Canadian Capacity Utilization Rate (%) U.S. Capacity Rate ~n Canadian Êapadty Rale Figures 2.1 and 2.2 show the impact of the second oil shock on the U.S. and Canadian economies. The weak growth in 1980 was due to declines in various components of domestic demand, and as a result, capacity uti lization in manufacturing fell by about four percent. Occurring simultaneously with this weak economic growth, the number of LFV cases in the U.S. rose in 1980 to 103 from 24 the previous year. Figure 2.3 U.S. and Canadian Unemployment Rate (%) U.S. Unemployment Canadian Unemployment The devastating impacts of the 1981-1982 recession can be seen in Figures 2.1 to 2.3. High inflation and interest rates as well as other factors combined and resulted in the worst recession since the second worid war. In the U.S., for example, unemployment reached a post war high, and capacity utilization reached a post war low. In 1982, there was a record number of LFV disputes in the U.S. and Canada. In the U.S., seven domestic steel companies si multaneously filed 93 countervailing duty and antidumping petitions (ITC 1983). In Canada, the number of LFV petitions rose to 73 in 1982, topping the previous record of 60 in 1977. Finally, the economic environment began to recover in 1983. Figures 2.1 to 2.3 show the results of the post 1982 expansion period. Healthy economic growth rates and relatively high capacity utilization rates helped to push the rates of unemployment down to nearly six percent in the U.S. and to about nine percent in Canada. However, despite the expansion, protectionist sentiment remained high due to concern over government deficits and rising imports. In the U.S., there was also substantial concern over the rapidly growing trade deficit. In 1984, pro tectionist sentiment was high in the U.S. and domestic manufacturing industries that had previously not felt the pressure from foreign competition, suddenly found themselves turning to LFV procedures for relief from rising im ports (ITC 1985). The U.S. and Canadian economies performed similarly over the 1975 to 1987 period. The apparent rela tionship between the business cycle and various macroeconomic shocks, on the one hand, and the frequency of LFV cases on the other, suggests that pressure for LFV protection is counter cyclical. As shown in Figures 2.4 and 2.5, the number of LFV petitions appear to rise during economic down-turns and fall when the economy is expanding^. The next section develops a framework for assessing the strength of this relationship and addresses the question of what motivates firms to file LFV petitions. Note that the number of LFV cases in the U.S. and Canada are not proportional to the size of their economies since the number of petitions depends more on other factors such as the number of frnns and industries, and on the proportion of tradeable goods and services in each economy. 2.4 Theory Development The starting point in this analysis of U.S. and Canadian determinants of the annual number of LFV peti tions is the specification of reduced form models. This paper adopts a simple rent-seeking-public choice framework for developing a structural model of the factors that affect the supply and demand for protection. Among the factors affecting the demand for LFV protection are the size of the industry, its concentration and its labour intensity (Feinberg and Hirsch 1989, Kaempfer and Willett 1989 and Grilli 1988). Thus, the larger and the more visible an industry is, the greater its abiUty to minimize the free rider problem and thus, to organize and pe tition for protection. It is also reasonable to hypothesize that industries will demand and receive more protection when they have fallen on hard times than when they are prosperous, even though they have more resources available for lobbying in the latter case. There are at least three reasons to expect a greater frequency of LFV petitions during hard times than when times are prosperous. The first reason, mentioned previously, is that an economy can more easily absorb increasing imports and sustain foreign competitive pressure when it is expanding. Second, it is important to distinguish between average and marginal changes when considering tiie decision to file a petition. Although it may be possible that average rent-seeking benefits are high during boom times (due perhaps to previous policy decisions), it seems more likely that marginal revenue changes will be higher during eco nomic downturns. Third, the distiibution of LFV cases suggests an inverse relationship between the number of petitions and the profitability of an industry. Most LFV cases in the U.S. and Canada involve dumping complaints^. Dumping is price discrimination in the sense that foreign firms charge lower prices in tiieir export markets than in their home market. Therefore the price gap or dumping margin should rise with increasing competition among import compet ing firms. One should expect to find an inverse relationship between tiie profitability of import competing indus tries and the number of LFV complaints. The supply of LFV protection is discussed by Grilli (1988). Factors affecting supply include the education level and political sophistication of die general public, the quality of information systems available to special inter ests and regulators, the level of organization and effectiveness of lobby groups, die extent of interventionist ti^adi-Between die years 1975 and 1987, 457 of 764 LFV petitions filed in die U.S. involved dumping complaints. The comparable figures for Canada were 420 of 432 LFV petitions. lions, regulatory changes and discretionary pohtical factors. Although it is quite easy to see how each of these fac tors can affect the supply of LFV protection, many of these factors are difficult to measure. Abstracting from real ity, this study assumes that the regulatory changes affecting the LFV processes and the tendency for politicians to enhance their own self interest by appealing to strong visible special interests during elections represent the most important shifters affecting the supply of LP^ protection. 2.4.1 U.S. Structural Model The number of LFV petitions filed in the U.S. each year (USLFV) is assumed to be a function of factors af fecting the demand (USDLFV) and the supply (USSLFV) for LFV protection and thus, can be represented by the fol lowing expression. USLFV = /(USDLFV, USSLFV) (1) Abstracting for simplicity, it is assumed that the number of LFV petitions demanded each year in the U.S. can be represented by the following expression. USDLFV = /(Economic Activity, Relative Competitiveness), (2) where economic activity in the U.S. economy is assumed to be cycUcal in nature and related to the general ups and downs in the economy (i.e., to tiie business cycle). At the macroeconomic level, manufacturing profits can be used as a measure of the state of the domestic manufacturing sector (USPROFIT). The relative international competitiveness of U.S. import competing industries is assumed to be reflected by the level of import peneO:ation and, thus, is also cycUcal in nature, but not in the same sense as Uie general busi ness cycle. A rise in import penetration represents a relative decline in U.S. international competitiveness. At a broad level, the penetration of imports over time in the U.S. (USIMPEN ) can be represented as follows: USIMPEN = /(RUSEXCH, USRELCOST, TREND), (3) where RUSEXCH is the real U.S. effective exchange rate, USRELCOST is the relative unit cost of production and TREND is a time trend. Changes in the real U.S. exchange rate (i.e., deviations from purchasing power parity), represent real shifts in comparative advantage and, thus, in the relative competitiveness of U.S. foreign manufacturers. The relative cost of production has been used as an index of the relative competitiveness (Roberts 1988; Adams, McCarl and Homayounfarrokh 1986). An increase in U.S. unit production cost relative to foreign competitors represents a de-crease in the competitiveness of U.S. producers relative to foreign exporters. Over time, there can be changes in the overall technological capability of a country; the trend variable measures the residual shifts in relative U.S. compara tive advantage. Substituting the relationships discussed above for the factors triggering the demand for LFV protection. Expression 2 can be rewritten as follows: USDLFV = /(USPROHT. RUSEXCH. USRELCOST, TREND), (4) where USPROHT is the real value of U.S. manufacturing profit and RUSEXCH, USRELCOST and TREND are de fined as above. The supply of LFV protection is a function of several complex factors, many of which are difficult to mea sure. Abstracting from reaUty, this study assumes that the supply of LFV protection can be represented as follows: USSLFV = /(Injury, Regulatory Environment, Political Environment) (5) The supply of LFV protection is assumed to increase with the level of domestic material injury^. Injury in vestigations involving LFV complaints must take into account all relevant factors that have a bearing on the state of the domestic industry and find that the impact of imports (in terms of volume and price) is significant. Furthermore Feinberg (1989) maintains that in recent years, the ITC has taken a bifurcated approach to injury determinations; first a finding of absolute injury is required, and then they determine if LFV imports have caused the injury. Although imports are the result of many interacting factors, there is a tendency in the U.S. and elsewhere to associate higher absolute levels of imports with increased injury. Consequently, this study adopts the premise that the import level is a reasonably index of material injury and thus, higher imports lead to a greater supply of protection. Changes in the U.S. regulatory environment shift the supply of LFV protection. The regulatory environ ment has changed over the years. Briefly, the the Trade Act of 1974 imposed strict time limits on investigations and allowed for private petitioners to challenge LFV rulings. The Trade Agreements Act of 1979 combined dumping with countervailing duty complaints to create what are now known as Title VII cases. This act also shifted the re sponsibility for Treasury Department procedures to the Commerce Department and shortened the time allowed for making Commerce and ITC determinations. The Trade and Tariff Act of 1984 addressed the notion of upstream sub sidies and directed the ITC to aggregate the effects of imports from all countries in its injury determinations. These Material injury is defined by the U.S. Trade Agreements Act of 1979 as harm which is not inconsequential, immaterial or unimportant. regulatory changes are generally thought to have simplified and streamlined LFV procedures, and as such, to lead to more positive determinations by the ITC and Commerce. The political environment pertains to those factors that can have an impact on the discretionary decision making powers of the Executive and Legislative Branches of the U.S. government. The political economy of deci sion-making by the U.S. government is beyond the scope of this study; however, die importance of electoral politics to the supply of LFV protection can be tested in the same manner as in Hibbs (1987). In his study of the relation ship between political parties, elections and the supply of policies and outcomes, Hibbs (1987) found statistically significant relationships between election years and die supply of changes to inflation and unemployment policies. In die same vein, diis study postulates diat presidential elections may also shift the supply of LFV protection. Substituting die relationships discussed above for die factors affecting the supply of LFV protection. Expression 5 can be rewritten as foUows: USSLFV = /(RUSIMP, TA74, TAA79, TTA84, U76, U80, U84), (6) where RUSIMP is die real value of U.S. imports, TA74 is an indicator variable for die Trade Act of 1974, TAA79 is an indicator variable for the Trade Agreements Act of 1979, TTA84 is an indicator variable for the Trade and Tariff Act of 1984, and U76, U80 and U84 are indicator variables for die 1976,1980 and 1984 U.S. presidential elections respectively.^ Substituting Expressions 4 and 6 into Expression 1 leads to the following general expression for the num ber of U.S. LFV cases per year (USLFV). USLFV = /(USPROHT, RUSEXCH, USRELCOST, TRENfD, RUSIMP, TA74, TAA79, TTA84, E76, E80, E84) (7) Expression 7 is a reduced form model of die number of U.S. LFV cases per year. Various specifications of diis gen eral expression were estimated using annual observations for the years 1975 to 1987. 2.4.2 Canadian Structural Model The model of the number of U.S. LFV cases per year may not be appropriate for Canada due to the open ness of die Canadian economy and the greater importance of exports to the Canadian economy dian in die United 5 The indicator variables TA74, TAA79, TTA84 are equal to 0 up to die year before enactment, and equal to one odierwise; die indicator variables U76, U80 and U84 are equal to 1 in die year of die election and 0 odierwise. States. However, the development of die Canadian structural model starts from the same equilibrium condition as for die U.S. model. That is, the number of LFV petitions filed in Canada each year (CLFV) is assumed to be a function of die factors affecting the demand (CDLFV) for and die supply (CSLFV) of LFV protection and, dius, can be represented by die following equilibrium condition. CLFV = /(CDLFV, CSLFV) (8) The factors affecting die number of LFV petitions demanded in Canada each year are much die same as those in the U.S.; however, the degree of interaction between economic activity and relative competitiveness in Canada is much greater than in die United States. In Canada, changes in the international competitiveness via real changes in die exchange rate will have an immediate and much larger effect on manufacturing activity than in the United States. Thus, the specification of die the Canadian demand for LFV protection cannot include bodi manufac turing profit and the exchange rate. This study assumes diat Canadian real manufacturing profit (CPROFIT) can be represented by the following expression. CPROHT = /(RCEXCH, CLPROD, CDEMAND), (9) where RCEXCH is die Canadian real effective exchange rate, CLPROP is die average productivity of labour in Canadian manufacturing and CDEMAND is the level of Canadian domestic demand. Canadian import penetration is assumed to be a function of the same variables as in the U.S. and thus, can be represented by die following expression. CIMPEN = /(RCEXCH, CRELCOST, TREND), (10) where RCEXCH is defined as above, CRELCOST is die relative unit production cost of Canadian manufacturing and TREND is a time trend. Aggregating these relationships results in the following expression for the demand for LFV protection in Canada: CDLFV = /(RCEXCH, CLPROD, CDEMAND, CRELCOST, TREND), (11) where RCEXCH, CLPROD, CDEMAND, CRELCOST, TREND are defined as above. The Canadian supply of LFV protection is assumed to be a function of die same factors as in the United States. Also, like the U.S., imports are assumed to be a reasonable index of material injury; higher imports repre sent greater injury and, dius, lead to a greater supply of LFV protection. The regulatory environment has been quite stable in Canada. However, in 1984, the Special Import Measures Act was enacted and replaced the Antidumping Act. The regulatory changes in the Special Import Measures Act were not substantial, but are generally thought to have simplified the LFV procedures for petitioners while retaining the discretionary powers of the government. Thus, it is assumed that the Special Import Measures Act had a positive effect on the number of Canadian LFV peti tions per year. Finally, as in the U.S., this study postulates that federal elections may shift the supply of LFV pro tection. That is, this stody tests if Canadian federal elections in 1974,1979,1980 and 1984 correspond to shifts in the supply of Canadian LFV protection. By aggregating these relationships, the Canadian supply for LFV protection can be represented as follows: CSLFV = /(RCIMP, SIMA, C74, C79, C80, C84), (12) where RCIMP is the real value of Canadian imports, SIMA is an indicator variable for the Special Import Measures Act and C74, C79, C80, C84 are indicator variables for the federal elections held in 1974, 1979,1980 and 1984 re spectively.^ Substituting Expressions 11 and 12 into Expression 8 results in the following general expression for the number of Canadian LFV petitions per year (CLFV). CLFV = /(RCEXCH, CLPROD, CDEMAND, CRELCOST, TREND, RCIMP, SIMA, C74, C79, C80, C84) (13) Expression 13 is a reduced form model of the number of Canadian LFV cases per year. Various specifications of this general expression were estimated using annual observations for the years 1975 to 1987. The next section reviews the statutes governing LFV procedures in the U.S. and Canada. Following this, the time horizon and the data used in die estimation of Expressions 7 and 13 are presented. 2.5 Review U.S. and Canadian Trade Statutes The dependant variable used in this analyses of protectionist pressure in the U.S. and Canada is the total number of initiated countervaiUng duty and antidumping cases. These two types of actions against unfair trading practices were combined because they both involve less-tiian-fair value complaints. Briefly, this is because "the trade practices they are intended to control involve, in legal terms, die sale of products at less Uian Uieir 'fair' value" The value of the indicator variables are as follows: the SIMA variable is equal to 0 before the Act and 1 after; the election year indicators are equal to 1 in the year of the election and 0 otherwise. (Finger Hall and Nelson 1982). There have been several changes to antidumping and countervailing duty laws in die U.S. and Canada over the years. This section reviews the major changes to the statutes governing these laws. 2.5.1 United States Trade Statutes As signatories to die GATT and die GATT subsidies code^, the U.S. and Canada have similar trade statutes. Regarding the U.S., the relevant statutes governing the rules and regulations concerning antidumping and counter vailing duty cases are die Antidumping Act of 1921, die Tariff Act of 1930, die Trade Expansion Act of 1962, the Trade Act of 1974, die Trade Agreements Act of 1979 and die Trade and Tariff Act of 1984 and die Omnibus Trade Act of 1988^. Aldiough some of the rules and procedures in die U.S. governing countervailing duty and antidump ing cases have changed over die years, die essence of what constitutes dumping and an unfair foreign subsidy has not. This section first explains the notion of dumping and dien briefly reviews some of die more substantive changes to the antidumping investigation procedures. Following diis, major changes to U.S. countervailing proce dures are reviewed. Dumping occurs when an imported good's fair market value exceeds die purchase price paid by a U.S. im porter (or an exporters sales price if die exporter markets a good in the U.S. market)^. Section 201(a) of the Antidumping Act of 1921 provided diat whenever foreign goods were being, or were likely to be sold in die U.S. domestic market or elsewhere at less than fair value, the Tariff Commission^^ shall determine if a U.S. industiy was being injured, or was likely to be injured, or was being prevented from being established by die importation of dumped goods (U.S. Tariff Commission 1975). In addition, diis act stipulated diat the dumping margin and, thus, the antidumping duty, must be equal to die difference between a good's foreign market value and its U.S. purchase price or exporter's sales price (taking account of various conditions and terms of sale), which ever is die most rele-' The full name of die 1979 GATT subsidies code is "Agreement on Interpretation and Application of Articles VI, XVI, and XXni of die General Agreement on Tariffs and Trade. However, for convenience, diis agreement will be referred to as die subsidy code. 8 In 1988, die U.S. passed die Omnibus Trade Act; however, die time series of U.S. import protection cases was only available until 1987, dius die Omnibus Trade Act of 1988 is not pertinent to this study and will not be reviewed. 9 Economic dieory defines dumping as price discrimination in international trade. There is price discrimination when like goods sell for different prices in two or more national markets. The term dumping is most often used to describe price discrimination between home and foreign maricets when die lower price is charged in die foreign market For a discussion of dumping as price discrimination see Wares (1977). ^ ^ Renamed die International Trade Commission by the Trade Act of 1974. vanL However, within limits, the Secretary of the Treasury could decide both how intensively to investigate a par ticular dumping complaint and whether or not dumping had or was occurring (Wares 1977). The criteria for arriving at affirmative injury findings in dumping cases has changed somewhat over the years. In the early 1960s, the Trade Expansion Act of 1962 raised the standard for obtaining protection. However, the Trade Act of 1974 (the first major revision of the Trade Expansion Act of 1962) once again made die test for finding injury less stringent (U.S. Tariff Commission, 1975). Antidumping procedures were once again substan tially modified by die Trade Agreements Act of 1979. The GATT multilateral trade negotiations in general, and die GATT antidumping code in particular, provided the primary motivation for this act. The Trade Agreements Act of 1979 repealed the Antidumping Act of 1921 and amended the Tariff Act of 1930 by including antidumping and coun tervailing duty investigations under Tide VII. This is why dumping and countervailing duty complaints are now re ferred to as Tide VII cases. This act also shifted responsibility for administi-ation of antidumping procedures from the Treasury to die Department of Commerce. The Trade and Tariff Act of 1984 made several modifications to die antidumping laws; however, these changes were mainly technical in nature, and were meant to clarify and simplify investigation proceedings. The new provisions included in this act pertained more to countervailing duty investigations than to antidumping investiga tions. The major revisions to U.S. countervailing duty procedures are discussed below. U.S. countervailing duty legislation was originally enacted in 1897 and was later embodied in section 303 of the Tariff Act of 1930. This section provided for countervaiUng tariffs without regard to a determination of injury only when subsidized, dutiable goods were being imported. In odier words, if a good was imported duty free, then countervailing duties could only be imposed if there were affirmative determinations regarding die simultaneous pres ence of a subsidy and injury. The absence of an injury test for dutiable goods predates die GATT, and die U.S. relied on its grandfadiering rights to protect itself from violating die GATT requirement (contained in Article VI) to prove material injury. The Trade Act of 1974 represents the first time that countervailing duty law dealt specifically with nondutiable imports. The Trade Act of 1974, also for the first time, aUowed private petitioners to mount court challenges to ac tions taken by die Department of die Treasury. Furthermore, diis act imposed time Umits on each of die steps^^ in For example, under die Trade Act of 1974, die Department of die Treasury was required to issue a preliminary determination as to the presence of an unfair subsidy within six months of the receipt of a complaint. An the countervailing duty decision-making process. Hufbaur and Erb (1984) claim that "the right of court review, rein forced by sti-ict time limits, led to a surge of petitions". Furthermore, section 331 of diis act required that a public notice be issued upon the receipt of each complaint alleging that goods benefiting from bounties or grants in the country of origin were being exported to the U.S. Finally, this act also gave authority to the Secretary of the Treasury (until January 3, 1979) to waive imposition of a countervailing duty if, in his judgement, adequate steps had been taken to remove the subsidy or substantially reduce its injurious affects. Hufbaur and Erb (1984) state that this waiver authority was used quite Uberally. The Trade Agreements Act of 1979 modified countervailing duty law in a number of ways, and was also a major step toward implementing GATT multilateral trade agreements. First, in order to make GATT conforming changes in domestic ti^ade law, section 701 was added to the Tariff Act of 1930. This section provides for die re quirement of a positive material injury determination before a countervailing duty can be applied to imports from countries who are signatories of the GATT subsidies code. Section 303 was still applicable to countries that were not signatories. In addition, diis law expanded die audiority to suspend investigations when or if early action taken by a foreign government eliminated the unfair subsidy or its injurious effect. Finally, the act shortened the time granted to die Department of Commerce and die ITC for making dieir preliminary and fmal determinations. Section VI of die Trade and Tariff Act of 1984 also modified countervailing duty procedures. This statute revised die authority to reach a settiement agreement by introducing a verification requirement for setdements that involved the use of off-set taxes. In addition, diis statute required that die public interest be considered in setdements that involved the use of quantitative restrictions. The notion of upstream subsidies was also addressed in this act. Subject to certain conditions, countervailable subsidies were extended to include certain payments to producers of in puts that were used in a subsequent production process. This act stipulated that diese upsti-eam subsidies were to be counted as part of die total benefit to be countervailed if the upsti-eam subsidy resulted in an input factor price that was lower dian would be paid in an arm's-lengdi ti-ansaction. Anodier major revision inti-oduced by diis statute af fected injury investigations. Prior to 1984, the ITC was not given explicit direction regarding die question of whedier the injury from a subsidized import should be considered from each country separately or from all countries additional six mondi period was also provided to determine whether the the imposition of a final countervailing duty was warranted. combined. This statute required the ITC to consider die cumulative effect from all countries in its determination of injury. Two pubUcations were used as sources of information concerning U.S. LFV cases. The tides of die two publications, both issued by die ITC, are as follows: Operation of the Trade Agreements Program, Annual Report (various years) and United States International Trade Commission Annual Report (various years). The time series for the number of initiated antidumping and countervailing duty cases were consdiicted by recording the sequential ITC case numbers and dates diat die injury investigation request was received by die ITC. I assumed diat diis investiga tion request date was a reasonable proxy for die date die petition was initiated. Regarding countervailing duty peti tions, botii 303 cases involving injury determinations and 701 cases were recorded^^. 2.5.2 Canadian Trade Statotes In simple terms, dumping in Canada refers to "die sale for export at prices ('export price") lower dian those charged to domestic buyers ('normal value'), taking into account the conditions and terms of sale" (Slayton 1979). The Canadian statotes governing antidumping and countervailing duty investigations are die Antidumping Act, die Customs Tariff Act and die Special Import Measures Act The Antidumping Act of 1969 fulfilled Canada's obligations as a signatory to the GATT. This act gave in vestigative audiority to die Departments of National Revenue (Customs and Excise Branch) and Finance (die Anti dumping Tribunal). The Deputy Minister of National Revenue for Customs and Excise was audiorized to initiate in vestigations into dumping, either on his own or as a result of a dumping complaint received from a Canadian pro ducer. In addition, and on his behalf. Customs and Excise officials conducted preliminary and final investigations aimed at determining the margin of dumping. This act also provided for die creation of the Antidumping Tribunal which was authorized to "conduct inquiries as a quasi-judicial court of record to determine die impact of dumped im ports on the production in Canada of like goods" (Canadian Anti-dumping Tribunal 1983). Thus, acting like a court, die Tribunal heard evidence to determine if dumped goods had caused, were causing or were likely to cause material injury to Canadian import competing producers, or if die dumped goods had materially retarded or were materially re tarding or were likely to materially retard die establishment of a Canadian import competing producer. 12 303 cases refer to section 303 of die U.S. Tariff Act of 1930.701 cases refer to section 701 of of die U.S. Tariff Act of 1930 as amended by die Trade Agreements Act of 1979. In July, 1979 (and in response to die conclusion of the Tokyo Round of multilateral trade negotiations), Canada announced that it intended to adhere to the GATT conventions on subsidies and countervailing duties, and the revised antidumping code (released in June, 1967). To conform to these conventions the Canadian government re solved to amend or repeal all statutes diat pertained to imposition of antidumping and countervailing duties. The re sult of this review process was the Special Import Measures Act (SIMA) which received Royal Assent on June 28, 1984. The purpose of SIMA is to protect Canadian industry from unfair import competition (U.S. International Trade Tribunal 1987). SIMA repealed the Antidumping Act in order to sti-eamline and modernize Canada's ti^ade laws. With respect to antidumping investigations, SIMA retains the same division of responsibilities between the departments of Revenue and Finance, but provided for the creation of a new injury determining council which was named die Canadian Import Tribunal^^. With respect to countervailing duty laws and procedures, Canada did not have any until March, 1977 when the Governor in Council proclaimed and published die Countervailing Duty Regulations^^ under section 7 of the Customs Tariff Act The first Canadian countervailing duty investigation occurred in 1982. Preliminary and final investigations into the existence of unfair and injurious foreign subsidies were done by the Department of National Revenue, Customs and Excise. Aldiough a material injury investigation by die Antidumping tribunal was required by section 16.1 of die Antidumping Act^^, the Tribunal's findings were only advisory. A complete listing of all antidumping and countervailing duty cases was obtained from Revenue Canada, Customs and Excise. However, diese data had to be adjusted to ensure comparability widi U.S. data. Canada num bers its LFV cases differendy dian in the U.S. For example, in 1982, both Canada and the U.S. investigated dump ing complaints concerning imported steel products originating from several different countries. In the U.S., die ITC considered each of the countries named in these complaints as a separate investigation. Consequendy, in the U.S., if 13 In effect, SIMA simply provided for the renaming of die Antidumping Tribunal to die Canadian Import Tribunal. On July 18, 1988, die Canadian Import Tribunal was renamed die Canadian International Trade Tribunal as a result of die passing of Bill C-110 "An Act to Establish Canadian International Trade Tribunal and to Amend or Repeal Other Acts in Consequence Thereof. 1 Order in Council PC 1977-838 of March 24,1977. The die Countervailing Duty Regulations were superseded by SIMA in 1979. 1 ^ This section stipulated diat die Antidumping Tribunal had to "inquire into and report to the Governor in Council on any odier matter or thing in relation to the importation of goods into Canada that may cause or threaten injury to die production of like goods in Canada" (Antidumping Tribunal p. 9,1983) a petition is filed that accuses companies located in five different countries of dumping, then the ITC carries out five different investigations and numbers dien sequentially. In Canada, the Canadian Import Tribunal considered each pe tition as a separate case. Consequendy, in Canada, if a petition is filed that accuses companies located in five differ ent countries of dumping, dien the Canadian Import Tribunal (now die Canadian International Trade Tribunal) carries out only one investigation. These numbering conventions are die same for countervailing duty investigations. In order to insure comparability, the Canadian time series for antidumping and countervailing duty cases were adjusted to conform with die ITC's case numbering convention. This adjustment comprised a simple renumber ing or expansion to reflect the number of countries cited in each Canadian investigation. 2.6 Choice of Study Time Period and Data Sources The time period for die U.S. and Canadian data bases is from 1975 to 1987. The timing of trade law revi sions was die primary reason for choosing 1975 as the beginning of the time horizon for die U.S. study. As men tioned above, die Trade Act of 1974 brought about several changes in U.S. trade policy. For example, diis act al lowed private petitioners to challenge the determinations of the U.S. Treasury, imposed time limits on each step of the dumping and unfair subsidy determination processes, and gave the Secretary of die Treasury new authority to waive countervailing duty determinations. These changes led to the expectation of a bulge in the number of LFV pe titions in general, and in countervailing duty petitions in particular. This is because this act appeared to favour po-Table 2.1 Listing of Independent Variables Used in LFV Analysis Variable Name Units U.S. Gross National Product Implicit Price Deflator Canadian Gross Domestic Product Implicit Price Deflator U.S. Real Effective Exchange Rate Index Canadian Real Effective Exchange Rate Index Japanese Real Effective Exchange Rate Index German Real Effective Exchange Rate Index U.K. Real Effective Exchange Rate Index U.S. Manufacturing Profits: Durable Canadian Real Corporate Profits Before Taxes: Manufacturing U.S. Real Value of Imports: fob Canadian Real Value of Imports: fob Canadian Index of Industrial Production Canadian Manufacturing Employment Canadian Growth in Real Private Consumption 1980=100 1980=100 1980=100 1980=100 1980=100 1980=100 1980=100 BiUions of 1980 $U.S. Millions 1980 $Can. Billions of 1980 $U.S. Millions 1980 $Can. 1980=100 1985=100 tential petitioners by reducing the discretionary powers of the law's administrators. To avoid die potential problem of a change in the size of die potential pool of LFV petitioners, the study period starts at the time when the Trade Act of 1974 came into effect (i.e., January, 1975). Even diough Canadian antidumping laws have changed very littie fi-om dieir enactinent in 1969 of the Antidumping Act, the 1975 to 1987 time period was also used for die Canadian analysis. The rationale for diis deci sion was two fold. First, using the same time period would improve comparabiUty of the U.S. and Canadian results and second, Canadian real effective exchange rate data was not available before 1975. Table 2.1 lists the variables that were collected for this study. Except were noted below, all data was ob tained from the publication entitied International Financial Statistics (IMF 1988). Where necessary, conversion of nominal values to real terms was done using the appropriate country GNP/GDP implicit price deflator. U.S. annual corporate profits for manufactured durable products was collected from die Economic Report of die President Transmitted to Congress in February 1991. For Canada, before tax corporate profits from manufactur ers was obtained from Statistics Canada Catalogue No. 13-201. The index of Canadian annual manufactiiring em ployment was obtained fit)m die OECD publication entitied Main Economic Indicators: Historical Statistics 1969-1988. Finally die growdi of real Canadian private consumption was obtained from OECD Economic Outiook: 1990. Appendices 2.1 and 2.2 list die U.S. and Canadian data used in diis study. 2.7 Estimation and Results Several versions of Expressions 7 and 13 were estimated to test die validity of die two sti-uctural model specifications and to assess the relative importance of U.S. and Canadian macroeconomic conditions to die number of LFV petitions filed each year. All estimation was done using Version 6.2 of SHAZAM (White et al. 1990). Unless specified odierwise, significant means statistically significant at die 95 percent level of confidence. 2.7.1 Factors Affecting die Number of U.S. LFV Petitions per Year Widi only 12 years of data, die specification of the U.S. structural model was kept simple. In die absti^act, the number of U.S. LFV petitions per year was assumed to be a function of corporate manufacturing profits, die U.S. real effective exchange rate, U.S. relative manufacturing costs, a linear trend, real U.S. imports, regulatory changes to U.S. trade statutes and presidential elections. For die development of Expression 7,1 assumed diat lower profits would increase the demand for LFV pro tection. HowevCT, corporate manufacturing profits represent a broad array of industry sectors, many of which do not compete widi imports. Consequendy, I assumed diat durable manufacturing sectors were the most sensitive to for eign imports. This assumption is reasonable given die analysis in Finger (1981) which showed diat, from 1975 to 1979, virtually all U.S. LFV petitions involved manufactured goods and diat over half of the imports named in peti tions were related to die auto industry. Finger (1981) also found diat many of die complaints against developed countries centred on steel, steel products and electronic machinery. Real corporate profits of durable product manufacturers have fluctuated dramatically over die period from 1975 to 1987. For example, from 1975 to 1978, profits ti-ended upwards at 20.8 percent per year^^. However, dur ing the four years between 1978 and 1982, profits fell at a rate of 53.5 percent per year. Finally, from 1982 to 1987, profits from manufacturers of durable products rose at a rate of 29.0 percent per year. There was a minus 72.2 percent linear correlation between die number of U.S. LFV petitions per year and the level of real U.S. profits for manufacuirers of durable products. The U.S. real effective exchange rate and U.S. relative real manufacturing costs are assumed to be important factors in die demand for LF^ protection. A higher effective U.S. exchange rate or higher relative production costs should increase the number of LFV complaints. In order to reduce die number of variables and because of die diffi culty of developing consistent country specific estimates of production costs, the ratio of U.S. to Japanese real effec tive exchange rates based on normalized unit labour costs was used as a competitive index of factors affecting U.S. import penetration and, thus, the number of petitions filed per year. Thus, the ratio of U.S. to Japanese real effec tive exchange rates based on normalized unit labour costs was used instead of RUSEXCH and USRELCOST. This ratio reflects changes in relative competitiveness between die U.S. and Japan due to real changes in exchange rates and wages. Japan was chosen for this relative competitive index because Feinberg (1989) found diat, during the pe riod fi^om 1982 to 1987, the relationship between U.S. LFV complaints and exchange rates was particularly signifi cant for complaints concerning Japanese companies. Of the 203 cases analyzed by Feinberg (1989), only 17 percent involved imports from Japan; however, diis represented the largest single share of complaints. The larger the ratio of U.S. to Japanese real effective exchange rates, die greater the import penetiation and, thus, the greater die number of LFV complaints per year. The time series of die ratio of U.S. to Japanese real effec tive exchange rates can be divided into three periods. From 1975 to 1978, die ratio fell at an annual rate of 14.6 per cent. From 1978 to 1985, die ratio rose at a rate of 9.2 percent per year. Finally, from 1985 to 1987, die ratio fell sharply at an annual rate of 26.2 percent. U.S. relative competitiveness appears to be cyclical, at least with respect to Japan, and is closely related to die frequency of LFV complaints. There was a 57.4 percent correlation between the number of U.S. LFV petitions per year and die ratio of U.S. to Japanese real effective exchange rates. Unless stated otherwise, all annual rates of growth are calculated as the compound rates of linear trend lines. The real value of U.S. imports can be viewed as an index of potential material injury and, thus, higher im ports should lead to a greater number of LFV complaints via a shift in die supply of LFV protection. The real f.o.b. value of U.S. imports ti-ended steadily upward at a rate of 4.7 percent dirough out die study period. However, the post 1975 expansion, die recessionary trough in 1982 and the boom during the latter half of the 1980s can be clearly seen in die time series. There was a 33.2 percent correlation between die number of U.S. LFV petitions per year and die real value of U.S. imports. Expression 3 suggests that a trend variable might be needed to account for relative technological changes that affect die level of import penetration. However, because of a 90.2 percent correlation widi die real value of im ports, a time ti-end variable was not considered in diis analysis. Finally, indicator variables for U.S. legislative changes and presidential election years were incorporated into the analysis to test for supply shifts in LFV protec tion. 2.7.2 Model Results for die Number of U.S. LFV Petitions per Year Table 2.2 presents the results of the reduced form model of die number of U.S. LFV petitions per year. All four versions of Expression 7 were estimated using the natural logarithms of die explanatory variables. This trans formation was based on the results of a Box-Cox regression analysis. The results for the first model in Table 2.2 are based on estimating a diird order autocorrelation model. The choice of a diird order autocorrelation model was based on the results of die standard least squares model and on the significance of the lagged autocorrelation rho estimates^^. The coefficients for relative competitiveness (USJCI), durable profits (USRDPFT) and real imports (USRIMP) are all highly significant and have die expected sign. The coefficients for die regulatory and electoral indicator variables were not statistically significant but may have the cor rect signs. That is, it was expected diat during election years diere would be more pressure on die Department of Commerce and ITC to grant LFV protection. Furthermore, it was not unexpected diat, relative to die Trade Act of 1974, die Trade Agreements Act of 1979 and die Trade and Tariff Act of 1984 would be associated widi fewer LFV complaints. The t-statistics from the diird order autocorrelation model for die estimated rho values, lagged one, two and diree periods, were -3.94, -1.71 and -1.16 respectively. The second and diird models in Table 2.2 test the significance of regulatory and electoral indicator variables separately. The results for diese two specifications are based on first and diird order autocorrelation models respec tively. The second model indicates diat elections do not significandy influence die administration of U.S. LFV pro cedures; also, die sign does not conform with dieoretical expectations^^. Like die first model, die third model in Table 2.2 indicates that regulatory changes over die study period have not had a significant impact on LFV petitions, even though diey tend to be associated widi fewer complaints^^. The results for die fourth model in Table 2.2 are based on estimating a first order autocorrelation model^O. As widi die otiier du-ee models, die coefficients for USJCI, USRDPFT and USRIMP are all highly significant and have the expected signs. This fourth model explains 93.9 percent of die variance in the number of U.S. LFV com plaints, yet contains only three variables. Furthermore, die standard OLS assumptions appear to be reasonable and model specification concerns such as variable selection and functional form appear to be minor. The analysis of var ious Box-Cox transformations indicate that die choice of a linear functional form is reasonable. Even though most of die variance in die number of U.S. LFV complaints is explained by die fourth model, we cannot be sure diat this represents die true model. Nonetheless, the results of the fourth specification in Table 2.2 are robust, plausible and significant. Figure 2.4 shows the relationship between the actual number of U.S. LFV cases and the predicted num ber of cases using the fourth model in Table 2.2. In summary, die results indicate diat die number of U.S. LFV petitions is negatively correlated widi die business cycle and rises when firms are experiencing increasing competition from imports. Furthermore, it appears that the variability in die number of LFV cases is due to shifts in bodi supply and demand factors. On die demand side, die frequency of complaints rises widi declines in profitability and relative competitiveness. On the supply side, die frequency of complaints rises widi a real rise in imports. Combining these factors leads to the plausible ar gument diat decUnes in profitabiUty and competitiveness, that are coincident with a rise in imports, will give rise to an increase in the number of LFV petitions. That is, the results support die hypodiesis that U.S. import competing industiies seek to redistribute die benefits of ttade via LFV protection during economic downturns, i.e., when profits and relative competitiveness are low. Add to this die fact diat most LFV complaints originate from a relatively 1 ^ The t-statistics from die first order autocorrelation model for die estimated rho value was -144.9. 1 ^ The t-statistics from die diird order autocorrelation model for die estimated rho values, lagged one, two and diree periods, were -3.61, -1.62 and -1.15 respectively. 20 The t-statistics from the first order autocorrelation model for die estimated rho value was -97.74. Table 2.2 Regression Results: U.S. LFV Cases per Year for the Years 1975 to 1987 Models Variable 1 2 3 4 usja 71.463 73.917 59.153 72.223 (3.67), 0.466 (3.48), 0.335 (3.05). 0.424 (4.37). 0.237 USRDPFT -37.426 -29.512 -30.976 -29.880 (-2.46), 0.641 (-6.10). 0.320 (-4.60). 0.548 (-7.788), 0.239 USRIMP 142.47 99.718 115.27 100.24 (3.29). 0.786 (4.49). 0.012 (4.32). 0.786 (7.17), 0.012 USRIFr 8.5091 -1.7859 * * (0.42). 0.359 (-0.14), 0.153 * * TAA79 -19.480 * -1.20 (-0.59), 0.843 (-0.07). 0.833 * TTA84 -21.885 * -5.90 * (-0.87). 0.837 * (-0.31). 0.836 * Constant -585.44 -389.72 -466.15 -585.44 (-3.30), 0.000 (-4.58), 0.000 (-3.39). 0.000 (-3.30), 0.000 R2 0.9787 0.939 0.973 0.939 Durbin Watson 2.17 1.93 2.11 1.97 Het^ ok ok ok ok Degrees of Freedom 6 7 7 8 1 Six tests for heteroscedasticity were carried out. Three asterisks indicate rejection of the null hypothesis of homoscedasticity at the 0.01 significance level for at least one of the tests. White's heteroscedasticity-consistent estimates of the variance-covariance matrix were used to calculate t-statistics in these cases. 2 Values in brackets are t-statistics. The second value is the Aux for the corresponding variable 3 Aux R2 is the R^ for the regression of the i th independent variable on the remaining in dependent variables. small number of large, concentrated, capital intensive and unionized import competing industries, and the rent-seek ing model seems even more plausible. Finally, business cycle variables have significant explanatory power in anticipating U.S. antidumping and countervailing duty actions. Aldiough inconclusive, the close relationship between LFV cases and business cycle variables (diat are broadly independent of imports), supports die hypodiesis diat rent-seeking is among die factors that motivate industries to initiate dumping and countervailing duty complaints. Also, assuming an underlying rent-seeking motivation to LFV cases, it appears diat die relative attractiveness and die marginal benefit from rent-seek ing activity is higher when general competitive pressures are high as compared to when diey are low. Figure 2.4 Number of U.S. LFV Petitions USLFV=-585.44+72.223*USJCI -29.880*USRDPFT Actual D Predicted 2.7.3 Factors Affecting die Number of Canadian LFV Petitions per Year The frequency of Canadian LFV complaints per year was assumed to be a function of die Canadian real ef fective exchange rate, Canadian relative manufacturing costs, Canadian labour productivity, Canadian consumption, a linear ttend, the real value of Canadian imports, regulatory changes to Canadian trade stamtes and federal elections. Higher real effective exchange rates and relative production costs were assumed to increase the number of Canadian LFV cases via their effects on profits and import penetration. Like in the U.S. analysis, the ratio of real effective exchange rates was used to reduce the number of variables. However, because more of Canada's LFV com plaints involve imports from Europe, the ratio of the Canadian to the West German real effective exchange rates based on relative normalized unit labour costs was used as a competitive index. Furthermore, plots of the exchange rate ratio and the number of LFV cases indicated a one year lag smicture. Consequendy, die ratio of die Canadian to die West German real effective exchange rates, lagged one year, was used in die analysis of Canadian LFV cases. The larger die ratio of die Canadian to die West German real effective exchange rates, die greater the import penetration and die lower the manufacturing profit and, thus, the greater the number of LFV complaints per year. The time series of die ratio of Canadian to West German real effective exchange rates can be divided into du-ee peri ods. From 1976 to 1979, die ratio fell at an annual rate of 10.5 percent From 1979 to 1984, die ratio rose at a rate of 5.4 percent per year. Finally, from 1984 to 1987, the ratio fell at an annual rate of 8.9 percent. Canadian relative competitiveness appears to be cyclical, at least witii respect to West Germany, and is positively related to die fre quency of LFV complaints. There was a 43.4 percent correlation between the number of Canadian LFV petitions per year and die lagged ratio of Canadian to West German real effective exchange rates-^^. Average Canadian manufacturing labour productivity was assumed to be an important determinant of manu facturing profits. All other things being equal, lower labour productivity will reduce profits and, thus, lead to a higher number of LFV petitions per year. This study constructed a proxy for the Canadian average labour productiv ity by dividing die Canadian indusuial production index by an index of die number of Canadian manufacturing em ployees. This analysis used the annual rate of change in average labour productivity in the empirical estimation of Expression 13 because it was diought that annual rates would have a sti^onger impact on profits dian annual levels. The rationale for diis assumption was diat it should be die rate of productivity change (radier dian die level) that af fects manufacturing job turn-overs and, thus, net rates of job creation or loss. Given diat job creation or loss rates are closely linked to die overall level of profitability dien, it should be the rate of productivity change diat affect profits and die frequency of LFV complaints. The annual rate of change in die average productivity of Canadian manufacturing labour has fluctuated over the study period. From 1976 to 1982, die rate of improvement in labour productivity fell by 0.9 percent per year. From 1982 to 1984 the rate of productivity improvement improved quickly and grew at 6.7 percent per year. Finally, from 1984 to 1987, rate of improvement in labour productivity fell each year and dropped below a value of one in bodi 1986 and 1987. There was a minus 15.4 percent correlation between the number of Canadian LFV peti tions per year and the rate of change in Canadian average labour productivity. This study used die growdi in real private consumption as a measure of domestic demand in Canada. High growth rates in real private consumption reflect real increases in private spending. In turn, real spending increases re sult in higher profits which provide higher incomes diat promote a higher level of consumption. Thus, high growdi rates in real private consumption are associated with economic expansion and improved profitability, while low growth rates in real private consumption are associated widi economic contraction and declining profitability. The percentage growdi in real private consumption fell from 6.5 percent to 2.3 percent during the period from 1976 to 1981. From 1981 to 1982 growth in real private consumption fell dramatically from 2.3 percent to The correlation between die number of Canadian LFV petitions per year and die ratio of Canadian to West German real effective exchange rates was 31.7 percent minus 2.6 percent Private consumption recovered after the recession and by 1985, the growth rate in private con sumption had risen to 5.2 percent. During the 1985 to 1987 period, die growth in private consumption fell sUghdy to 4.5 percent Over the study period, there was a minus 60.4 percent correlation between the number of Canadian LFV petitions per year and the percentage growth in real private consumption. The real value of Canadian imports was assumed to be an index of potential material injury and, dius, higher imports should lead to a greater number of LFV complaints via a shift in die supply of LFV protection. The real f.o.b. value of Canadian imports ti-ended upward at a rate of 3.5 percent dirough out die study period. Like die U.S., die post-1975 expansion, die recessionary trough in 1982 and the boom during die latter half of die 1980s can be clearly seen in die time series of Canadian imports. There was a minus 56.4 percent correlation between die number of Canadian LFV petitions and the real value of Canadian imports. A trend variable was included in die analysis to account for relative technological changes diat affect the level of import penetration. In addition, indicator variables representing the 1984 change in Canada's trade statutes and federal elections were incorporated into die analysis to test for supply shifts in LFV protection. 2.7.4 Model Results for the Number of Canadian LFV Petitions per Year Table 2.3 presents die results of five different specifications of Expression 13. All models were estimated using OLS. The first model in Table 2.3 tests for die significance of die introduction of the Special Import Measures Act in 1984 and die impact of die federal elections in 1979,1980 and 1984. Aldiough die coefficients for the relative competitiveness, lagged one year (LCGCI), die growdi in real private consumption (CGRPC) and die av erage productivity rate of change of Canadian manufacturing labour (RTCMLAP) are all individually significant and have the expected signs, the regulatory and election year indicator variables (i.e., SIMA and CANELEC) are not sig nificant. Furthermore, die sign of SIMA is negative and, based on conversations with staff of the Canadian International Trade Tribunal, it should be positive. The second model in Table 2.3 indicates diat neidier the real value of imports nor the time trend are sti-ong determinants of die fi-equency of Canadian LFV complaints. Unlike die U.S., die coefficient of the real value of im ports (CRIMP) is not significant, it also has die wrong sign. The time trend variable, included to test for die im pacts of a long-run decline in manufacturing comparative advantage, is also insignificant; however, the sign agrees widi expectations. The relatively high Aux R^ statistics for this model reflect die 79.9 percent correlation between Table 2.3 Regression Results: Canadian LFV Cases per Year for die Years 1976 to 1987 Models Variable 1 2 3 4 5 Lcca 137.61 91.611 118.77 105.14 108.73 (3.63), 0.681 (3.07). 0.542 (3.27), 0.622 (3.79), 0.403 (4.38), 0.276 CGRPC -3.9831 -3.9929 -4.8744 -5.0835 -4.9052 (-2.93), 0.324 (-2.97), 0.389 (-4.03). 0.075 (-3.09), 0.538 (-4.29). 0.071 RTCMLAP -207.54 -150.60 -152.19 -144.26 -128.51 (-2.22). 0.707 (-2.59), 0.330 (-1.77). 0.625 (-2.28), 0.361 (-2.18), 0.284 CANELEC 7.5342 * 3.4403 * * (0.85), 0.611 * (0.40). 0.551 * * SIMA -8.1305 • 4.2577 * (-1.26). 0.373 * * (0.301). 0.863 * CIREND * 0.14381 -1.4014 * * (0.11), 0.758 * (-0.84). 0.826 * CRIMP * -0.000498 * * * * (-0.95), 0.824 * * * Constant 103.14 130.25 69.314 86.037 57.520 (1.57), 0.000 (1.80), 0.000 (1.11), 0.000 (1.51), 0.000 (1.11), 0.000 R2 0.8556 0.872 0.817 0.855 0.813 Durbin Watson 2.63 3.03 2.12 2.67 2.21 Het^ ok ok ok ok ok Degrees of Freedom 6 6 7 6 8 1 Six tests for heteroscedasticity were carried out. Three asterisks indicate rejection of the null hypothesis of homoscedasticity at the 0.01 significance level for at least one of the tests. White's heteroscedasticity-consistent estimates of the variance-covariance ma trix were used to calculate t-statistics in these cases. 2 Values in brackets are t-statistics. The second value is the Aux R^ for the corresponding variable 3 Aux R2 is the R^ for the regression of the i th independent variable on the remaining in dependent variables. CRIMP and CTREND. Furthermore, although inconclusive, the Durbin Watson statistic and a visual inspection of die residuals indicates diat die parameters may be poorly estimated due to die presence of negative autocorrelation. The diird and fourth models in die table of Canadian results test two different combinations of die time trend and supply shifter variables. Once again, CANELEC, SIMA and CTREND are insignificant. Finally, the fifdi model includes only the variables diat were significant in die other four models. This fifdi model explains 81.3 percent of die variability in die annual frequency of Canadian LFV complaints. Furthermore, die results for diis model do not appear to violate the standard OLS assumptions. The results of Box-Cox regression analysis support the choice of a linear functional form. Even though most of the variance in the number of Canadian LFV complaints is explained by the fifth model, we cannot be sure diat diis represents die true model. Nonedieless, diese results are significant and plausible with respect to economic dieory. Figure 2.5 shows die relationship between die actual number of Canadian LFV cases and die predicted number of cases using the fifth model in Table 2.3. In summary, die results in Table 2.3 indicate diat Canadian LFV petitions are negatively correlated with the growdi in private consumption and, dius, widi macroeconomic cycles. In addition, LFV complaints rise widi a dete rioration in Canada's relative competitiveness due to fluctuations in eidier die real effective exchange rate or normal-ized unit labour costs. Finally, high rates of improvement in the average productivity of labour tend to improve the Figure 2.5 Number of Canadian LFV Petitions CLFV=57.520 +108.73*LCGCI CD CO O CM TJ- to h- r- 00 œ 00 00 0> 05 0> 0> O) O) Actual • Predicted profitability of Canadian manufacturers and dius, lower die demand for LFV protection. Combining tiiese factors in dicates diat when manufacturers are faced with shrinking domestic markets and competitive pressures from exchange rates and dechning labour productivity, they tend to increase their demand for LFV protection. The insignificance of die diree supply side shift variables (SIMA, CANELEC and CRIMP), suggests diat the Canadian process for providing LFV protection was quite stable over die study period and that it may be less sus ceptible to the import competing manufacturer's lobby dian in the United States. Aldiough inconclusive, diis premise seems reasonable given die nature of the differences between U.S. and Canadian LFV procedures. For ex ample, in die U.S., a typical ITC hearing often limits die duration of presentations by affected parties to about diirty minutes, lasts only a few days and does not make submitted information public. In Canada, on the other hand, a typical hearing before the International Trade Tribunal does not impose sti-ict time limits on positions, usually lasts weeks and makes all submitted information public^^. As widi the analysis of the U.S., business cycle variables have significant explanatory power in anticipat ing Canadian LFV actions. Consequendy, the results support die hypodiesis diat rent-seeking is among die factors that motivate Canadian industiies to initiate dumping and countervailing duty complaints and diat rent-seeking activ ity in Canada is more vigorous during economic downturns than during economic upturns. All submitted information is made available to affected parties and their representatives; however, only information diat is not confidential widi respect to specific frnns is made available to die general public. 2.8 Discussion and Conclusions This paper focused on die question of what motivates U.S. and Canadian firms to initiate LFV investiga tions. It was presumed diat rent-seeking was among die motivating factors in die sense diat firms make LFV com plaints when it is in dieir interest to do so. However, it is not obvious what rent-seeking implies about the timing of complaints. This paper investigated die hypodiesis diat rent-seeking pressures are more acute at low points in die business cycle. Thus, business cycle variables were expected to have significant explanatory power in anticipating antidumping and countervailing duty actions. Although the premise may sound reasonable, it is very difficult, from a practical perspective, to test if a firm's decision to invest in protectionism is affected by the level of activity in its markets. This is because, at the firm level, investinents in protectionism are rare events. For a firm, or even an industry for diat matter, a petition for import relief occurs only once every so often. That is, for a particular firm or industry, protectionism is the ex ception radier than the rule. Aggregation can solve diis rare event problem. Aggregating across many firms or industi-ies, increases the chance that a rare event will occur in a given time-period. Thus, by grouping import competing firms togedier, there is a greater chance of assessing die relation ship between petitioning for import protection and economic activity. On the other hand, by aggregating across firms or industiies, much of the inherent variance among firms and industiies in terms of, for example, production costs or die timing and frequency of market cycles is lost. Thus, aggregation presents a new set of potential prob lems, many of which can only be solved by adopting simplifying assumptions about the characteristics of the grouped firms. This stiidy used the total number of antidumping and countervailing duty petitions as a measure of protec tionist pressure. This measure is an aggregate not only of LFV petitions, but also of all petitioning firms and in dustries. This is a powerful assumption. It implies that among die petitioners, diere exists a common set of charac teristics that are prominent enough to warrant die aggregation. Are there prominent characteristics diat are common to all or practically all petitioning industiies? This question of the overall validity of aggregating aaoss all petitioning industiies can be assessed, in part, by addressing die assumptions implicit to any aggregation procedure. First, as alluded to above, an optimal aggrega tion or allocation procedure requires a set of characteristics diat maximize the likelihood of differentiating among groups. In other words, from a spatial perspective, some set of characteristics must exist that maximizes the "distance" between groups. There are several characteristics common to all petitioning firms. The most obvious common element is diat diey all initiated LFV investigations, thus diey all were pressuring for special protection from imports. Second, they were all manufacturing firms tiiat compete with imports. Third, many of die investigations were requested by primary producers or traditional smokestack industiies. Fourth, a large proportion of the petitioning industries were large, capital intensive and well organized. Examples of diis representative group are die steel indusdy, die lumber industry, the textile industry, the oil and gas industry, the mining industry, the automotive industry and the agricul tural industiy. Finally, most of die petitioning industiies had similar business cycles, i.e., diey were procyclical. Notwidistanding diese similarities, diere are obviously substantial differences among die population of peti tioning industries in die U.S. and Canada. However, there is value in considering petitioners as a whole and in in vestigating the plausibility of diis assumption. If a sttong positive relationship exists between protectionist pres sure (as measured by the number of initiated LFV cases), and the level of economic activity, and die specification of the relationship is cogent at the firm or industry level, then the implicit assumption that petitioning firms or indus tries are similar in some sense is perhaps reasonable. That is, there is perhaps something to to learned from think ing of petitioners as a whole. Conversely, the lack of a sttong relationship or a cogent specification may mean diat aggregation across all petitioners was inappropriate. Aldiough diis Stody aggregated LFV cases, future work could investigate die effects of industry characteris tics and odier factors diat may significandy affect die supply and demand for LFV protection. For example, LFV cases could be disaggregated into those filed by industiies producing intermediate goods versus those producing goods for final consumption. Another question that could be tested is whedier industries dominated by multinational firms are more or less inclined to seek LFV protection dian industiies less dominated by multinational firms. There may also be a retaliatory component to die decision to seek protection; thus, future work could test die extent to which industiies file petitions in response to previous "attacks" fi-om odier countries. Finally, another factor in die process of LFV protection diat could be tested is die evolution of ITC appointinents. That is, does die composition of die ITC in terms of such factors as die number of economists or die political disposition of die committee members af fect die supply of LFV protection. Ideally, die measures of cyclic economic activity should have corresponded to the pool of LFV petitioners. However, in some cases, indicators of the total economic activity in die U.S. and Canada were used instead. This substitution was made because I assumed diat total economic activity was a reasonable proxy for die aggregate level of activity in die domestic markets of petitioning industries. This assumption is reasonable if these markets repre sent either a relatively fixed proportion or a very significant proportion of total economic activity. Although the manufacturing sectors of the U.S. and Canadian economies have been giving way to service sectors, they still ac count for a large share of total activity and dius die cyclic nature of total activity. The results of this study are based on a general specification of a structiiral model and a somewhat system atic exploration of a database containing several aggregate macroeconomic variables. Consequendy, die final models must be viewed as partially instigated by the data. Leamer (1978) presents both sides of die controversy regarding data dependant model selection processes. He starts widi the basic question of whedier or not there is value in data dependant specification searches. In odier words, are specification searches productive? One view is that diey merely identify relationships diat exist in historical data. That is, searches simply describe the salient economic features of a historical dataset. On die odier hand, Leamer (1978) suggests diat specification searches can be viewed as "dying to lend some data dependant quantitative support to a broadly held qualitatively determined conjecture". He also sug gests that judgement and purpose ultimately are essential ingredients in assessing the validity of data dependent mod els. In diis study, the development of reduced form models represent an attempt to reveal empirically relevant and economically consistent relationships. This search for relevant specifications is legitimate in diat it attempts to re veal "some of die tiiidi diat is buried in die data" (Leamer 1978)^3. This study found that business cycle variables were significant in explaining the frequency of U.S. and Canadian LFV complaints. Thus, the results support the contention that diere is a countercyclical relationship be tween protectionist pressure and macroeconomic variables. The significant explanatory power of business cycle vari ables in anticipating LFV complaints is consistent widi an underlying rent-seeking model of some sort. The find ings in Krueger (1980) also support a rent-seeking model. In her analysis of foreign ti-ade competition Krueger (1980) argues diat examination of the evidence does not support the contention that increased imports were a signifi-Anodier reason to treat these results cautiously is the small number of events. Because the study period covers only two business cycles, part of die reason for die good fit may be due die influential nature of a small number of observations. Techniques are available to address die small number of events problem (e.g., weighted regression techniques), however, diey were not used in diis stiidy. cant determinant of job losses in die United States. Dombusch and Frankel (1987) also maintain diat the adverse ef fects of imports have been overshadowed by economy-wide recoveries. That is, Dombusch and Frankel (1987) found diat although import penetration in die U.S. rose from 6.1 to 8.4 percent between 1972 and 1981, die fraction of shipments exported rose from 5.8 to 9.9 percent over die same period. Bhagwad (1988) and Rugman and Anderson (1987) maintain diat LFV trade disputes often represent a form of harassment aimed at shifting economic benefits toward import competing domestic industiies. According to this rent-seeking model, the technical process of administering LFV statutes in die U.S. and Canada is sensitive to polit ical and economic conditions. Rugman and Anderson (1987) argue that the tendency to down play economic princi ples in LFV investigations is an indication of an inability to depoliticize trade issues. When diis tendency is added to die fact diat trade issues tend to be dominated by a relatively small number of producer organizations (Lenway 1983, and Dombusch and Frankel 1989) and that diese producers face few risks when diey initiate trade actions, it is easy to see why industries resort to lobbying for special restiictions as a way of improving their competitive posi tion. The analysis of Canadian LFV cases suggest that protectionist pressure in Canada is countercyclical. The rate at which private consumption changes is a measure of die overall level of economic activity. When die growdi in Canadian private consumption is low or negative, as it was in 1982, die level of economic activity is low. Conversely, when private consumption is growing quickly, economic activity will be high. Thus, the growdi in Canadian private consumption is positively correlated widi die business cycle. The negative coefficients in Table 2.3 for CGRPC indicates diat die frequency of LFV complaints rises significandy during die downside of business cycles. This finding supports the hypothesis diat rent-seeking is a determinant of the incidence of LFV cases in Canada. Labour productivity is also an important determinant of LFV cases in Canada. This is because beginning in die early 1970s, a fundamental imbalance between real wages and productivity appeared in Canada (Economic Council of Canada 1987). The lack of productivity gains to vaUdate wage increases, coupled widi a decline in inter national competitiveness (due to movements in real exchange rates for example), contributed to die unemployment rate and declining profitability in Canada's import competing manufacturing sector. The imbalance between wages and productivity and the insignificance of real imports in Table 2.3 suggests diat an inability to compete in die do mestic labour market is more important dian die level of imports in explaining unemployment and lost profits. In summary, the results suggest that Canada's import competing industries seek import restricting policies during eco nomic downturns and when competitive pressures are high (due to such factors as changes in exchange rates and an imbalance between wages and productivity). This is consistent with the notion that rent-seeking is an important de terminant in the decision to seek LFV protection in Canada. If die rent-seeking model is correct, dien at least under some circumstance, bodi the political and administi^ative environment surrounding Canadian trade policy favour pri vate business interests over those of die public. I agree widi Cass (1989) that there is no sound justification for in terpretations of the law that result in penalties "when it can be shown that causes wholly extraneous to imports ... caused all but de minimis injury to domestic industry. By raising die requirements and die role of economic analysis in the disposition of LFV cases there may be a greater chance of settling trade disputes in ways that promote the public interest. Methods to depoliticize trade disputes need to be developed so diat die social costs of rent-seeking in the administration of Canadian trade policy can be removed or reduced. Finally, this study did not test for the presence of rent-seeking directiy. However, the significant explana tory power of business cycle variables in anticipating LFV complaints can be interpreted as supporting a rent-seek ing model of some sort. That is, business cycle variables indicate rent-seeking given certain auxiliary assumptions. For example, if it is assumed that during downturns, the opportunity cost of executives and lawyers etc. is low in the sense that they are not engaged in building new plants and doing things that they do during expansions, then the frequency of LFV complaints will tend to rise during recessions because die cost of initiating complaints will be less than during expansions. Or perhaps die marginal benefit of using scarce resources in marketing and odier activities aimed at expanding sales falls during a recession compared to die marginal benefit obtained from seeking LFV pro tection. Another possibility is to assume that the timing of petitions has somediing to do widi die satisficing the ory of the frnn. That is, maybe as long as firms are making a certain amount of revenue or executives and workers are getting paid and firms are not going bankrupt, dien firms and unions satisfice, diey just go along with tilings and do not worry about things. However, if dieir minimum aspiration level is threatened, as might be die case during a recession, dien they take vigorous action. These various timing assumptions were not expliciUy incorporated into die reduced form models estimated in dus study. Consequendy, nothing can be said about die appropriateness of different rent-seeking models for ex plaining the incidence of LFV petitions. A lack of data precluded the testing of possible reasons for the apparent timing of LFV complaints found in diis study. However, the data required for to tests several plausible timing hy-potheses are available so the question of what type of rent-seek motivates industries to seek LFV protection can be the focus of futiire research. The results of diis research would be interesting because it would help to explain why rent-seeking seems to be more vigorous during economic downturns than during economic upturns. 3. IMPORTS AS A CAUSE OF INJURY 3.1 Introduction In 1986, a negotiated setdement to die Canada-U.S. softwood lumber dispute, known as die memorandum of understanding, was signed before the U.S. International Trade Commission (ITC) completed its final injury deter mination. The question of whedier die injury being suffered by American lumber producers at diat time was caused by higher Canadian softwood lumber exports resulting fi-om less than fair value timber pricing pohcies has never been answered. This paper addresses the issue of injury in the Canada-U.S. softwood lumber dispute. That is, it tiies to assess die extent to which the alleged timber subsidy in British Columbia contributed to the injury suffered by American softwood lumber producers. This question of injury causality in the softwood lumber dispute is reasonably well suited to econometiic analysis. Using six different notions of injury, and using the best available data, I undertake a statistical analysis of the relationship betiveen various potential causal factors, including stumpage prices, and "injury" to U.S. producers. My basic finding is that stumpage prices appear to have had littie effect on injury to the U.S. industry. The effects of the "business cycle" and die exchange rate appear to be die most important determinants of the performance of die U.S. industry. Before discussing the six measures of injury used in this econometric analysis, I briefly discuss sev eral general but important aspects of injury. According to U.S. and international law, injury to domestic producers, either by itself or in conjunction widi an unfair subsidy, is a necessary condition for U.S. import protection. Petitions requesting relief from imports involving Escape Clause procedures require a positive ITC injury determination and presidential support for protec tion to be awarded. American softwood shingle and shake producers received "Escape Clause" protection from Canadian exports in June, 1986. The softwood lumber dispute involved die U.S. countervailing duty process. Unlike Escape Clause protec tion, this is a two track system. First, die Department of Commerce, through its International Trade Administration, must find that an unfair subsidy exists ^ and second, like Escape Clause petitions, die ITC must find that the petitioning industry has been injured or is faced with a threat of being injured. More specifically, in countervailing duty cases, die ITC must determine if die allegedly unfair imports are causing material injury to domestic producers. This reference to material injury arises fi-om die Subsidies and Coun tervailing Duties Code signed in 1979 as a result of the Tokyo Round of the multilateral GATT negotiations. However, aldiough die GATT and die Subsidies Code (Article 6:3) refer to material injury, neidier defines precisely what it is. Radier, material injury is described in terms of a series of criteria that must be linked to die alleged sub sidy under investigation^. The ITC must consider material injury to be "harm which is not inconsequential, imma terial or unimportant" (U.S. Tariff Act of 1930 as amended by the Trade Agreements Act of 1979) and include actual or potential declines in such factors as output, sales, market share, profits, and prices in its injury determinations. Furthermore, the ITC is required to take into account all die relevant factors that have a bearing on die state of a do mestic industry and find diat the impact of imports (in terms of volume and price) is significant From diis list of factors, it would seem diat injury has somediing to do with danger or potential danger to domestic claimants (i.e., share holders or wage earners). But a listing of factors does not address die concern over which is the most appropriate measure of injury. A lack of a clear consensus on what the most appropriate measure of injury should be is a concern for several reasons. Trade laws diat define or measure injury in such a way diat ben efits are conferred on one group while imposing a greater tax on anodier group will not serve die public interest. The meaning of injury should relate to a set of well understood and consistent principles for there to be injury deter minations diat will stand up to objective scrutiny. Also, from a more practical perspective, industiies monitor tiade disputes, and a definition or measure of injury that does not properly balance die interests of domestic groups with those of foreign exporters may distort ti-ade flows. Thus, a likely result of a biased definition of injury is a reduction in die profitability of foreign exporters. This idea is at the heart of die argument diat claims that injury determina tions can be (and often are) used as a form of harassment diat discourages foreign exporters from entering or expand ing into a domestic market^. For descriptions of die notion of fairness as it relates to U.S. ti-ade statutes see Rugman and Porteous (1989), Percy and Yoder (1987) and Haufbauer and Erb (1984). For a list of diese criteria see Balassa (1989), Kennedy (1989) and Haufbauer and Erb (1984). See for example Bhagwati (1988) Notions of injury can be found in more than just trade law. For example, competition law views injury and compensation in a way that is analogous to d-ade law. Competition laws define acceptable mediods of competition and award die right to seek redress in cases where injury can be linked to unacceptable methods of competition. Thus, firm's diat have suffered from unfair competition (e.g., predatory pricing) have die right to seek compensation for financial losses. This idea diat a fmn has a right to some form of compensation and protection from an unfair competitive practice is central to Q-ade disputes. Domestic fums in most countries have the right to petition dieir regulatory institutions for protection from foreign imports, and protection will be given when it can be shown diat foreign firms are taking unfair advantage of open markets and that the imports are causing unacceptable losses among domestic producers. Measurement of injury requires a comparison of the injured state with the state before the injury occurred, everydiing else being equal. Although diis comparison may sound simple and straight forward, in practice it usually is not. For example, in a ttade context, it may take several years for the fiiU effects of an allegedly unfair trade prac tice to work their way through the various domestic factor markets to die end product markets. In addition, there may be multiple injuries diat complicate die process of identifying the healthy state. Thus, it may be difficult to identify and specify the timing of a particular cause and effect relationship. Another difficulty relating to the causality problem is that the best measure of injury may be unobservable. If an industiy is being injured, it is possible to assess injury in terms of lost production, employment or market share. However, from a sdicdy economic point of view, the best measure of the injury is the change in the level of social welfare. From a practical point of view diough, the loss of social welfare is very difficult to estimate. Even diough social welfare remains hard to observe, the countervailing duty statutes of some countries (e.g., Canada) include a reference to the public interest^. Thus, it would seem that diere are different opinions about who's interests should be considered in an injury analysis. Who counts is an important consideration in an injury in vestigation. This is because, from a social perspective and excluding sti^ategic concerns, it is hard to see why retalia tion against foreign subsidies is warranted when subsidies result in net economic benefits. This was an issue in die Canada U.S. softwood lumber dispute. Because U.S. trade statutes do not attempt to balance die interests of bodi In die Case of Canada the statute is known as the Special Import Measures Act. Section 45 of diis statute specifically calls for the consideration of the public interest into the unfair trade law process (see Rugman and Porteous (1989) and Giese (1986)). consumers and producers, die ITC must focus only on the concerns of American producers in dieir injury determina tions. According to Kalt (1987), "taking die sum of consumers' and producers' surplus as our yardstick of aggregate national welfare, the ITC might have been instructed to send a note of gratitude to die Canadians, radier dian impose a tariff against diem". This question of who counts is most likely a moot point for any particular U.S. ti-ade dispute because pro cedures are largely fixed by existing laws. However, a focus on existing trade laws begs the question of which of the several measures of injury used by die ITC is the most appropriate for assessing trade impacts to producers? As men tioned above, die ITC must assess impacts on prices, output, market share and profits (among other tilings) as part of its injury determinations. In addition to these measures, the petition in the softwood lumber dispute included lost employment in its injury claims. Are all of diese measures equally capable of measuring the impacts of unfair im ports? Also, what is die relationship among these measures? The next section reviews the literature pertaining to the analysis of causality and the primary literature per taining to die econometiic analysis of die Canada-U.S. softwood lumber dispute. This paper next algebraically de fines six measures of injury and dien shows how diese measures respond to various static changes (i.e., shocks) in their component terms. Following this, an empirical overview of the injury measures is presented. In the following two sections, this paper first presents the theoretical development of a general reduced form injury model and then presents the actual specifications diat are empirically estimated. Next, the data collected for diis study and dieir sources are presented. The next section reports die results from empirically estimating the reduced form models for the six injury measures. Following this, the final injury models are used to assess die economic importance of sev eral key variables. Finally, the results and conclusions of this study are discussed. 3.2 Literature Review This review looks first at the Uterature pertaining to the analysis of causaUty and dien at the primary Utera-ture pertaining to econometric analyses of the Canada-U.S. softwood lumber market. In antidumping and counter vailing duty cases, die ITC is responsible for determining if a domestic indusdy has been materially injured. Morkre and Kruth (1989) review what diey regard as die traditional or dominant approach used by the ITC in dieir assess ments of die causal relationship between allegedly unfair imports and material injury to a domestic industry. In dieir paper, Morkre and Kruth (1989) divided injury investigations into four steps. First the ITC state what are perceived to be die controlling statutory standards. Second, a ti^aditional analysis reviews data on trends of the imports under investigation and often of die domestic industiy's performance. Morkre and Krudi (1989) contend that a negative cor relation between diese ttends points in die direction of a positive determination. Third, die ITC review data on nom inal transaction prices for the imports under investigation and diose for the domestic counterparts. Finally, individ ual transactions where domestic firms have allegedly lost sales or revenue are reviewed. In their critique of this traditional causality analysis, Morkre and Kruth (1989) cite a point made by Commissioner Cass in die 1988 Microdisks dispute diat die ti^aditional ITC approach to causality analysis may not be fully faidiful to die language and history of Tide VII of die U.S. Tariff Act of 1930^. In his review of die ITC's requirement to prove causaUty between unfair imports and injury, Pahneti-e (1986) concluded that cause and effect do not play a particularly important role in die agency's determinations. Widi respect to causality, most economists feel that the approach should compare a domestic industry's actual performance widi die estimated performance in die absence of the unfairly traded import during the time period of the investigation. A number of studies have used diis counterfactiial approach in analyzing die injury suffered by U.S. indus tries. For example, Grossman (1986) looked at die question of imports as a cause of injury to die U.S. steel indus try. In diis study, a log-linear reduced form employment equation was estimated using mondily data from January 1973 to October 1983. Grossman's stiiictural model starts widi a Cobb-Douglas production function widi labour, capital, energy, iron ore and scrap steel as inputs. After specifying supply and derived demand functions for die in puts, Grossman specifies diat die demand for steel is a function of imported steel and, dius, the exchange rate Antidumping and countervailing duty disputes are collectively known and less dian fair value disputes or Tide VII disputes. The term Tide VII disputes arose due to an amendment to the U.S. Tariff Act of 1930 combined antidumping with countervailing duty procedures in section seven of the statute. (among other factors). His study showed that employment in the U.S. steel industry was sensitive to the business cycle and import prices. Employment in die steel industry also suffered because of a time ttend effect interpreted as an economy wide shift in employment to high-technology sectors, an increase in die use of plastics and the adoption of labour saving technologies (among odier things). Grossman conducted a series of counterfactual simulations to assess die relative importance of die explanatory variables in his reduced for model. This study concludes that im ports were not die most important cause of employment loss to the steel industry and were only one fifth of the losses caused by the general secular shift toward odier industiies during die 1976 to 1983 period. Furthermore, dur ing the 1979 to 1983 period, the real appreciation of the U.S, dollar was responsible for all of the job losses associ ated with import competition. Pindyck and Rotemberg (1987) utilized a demand and supply framework in dieir development of a reduced form model of die U.S. copper industiy. Unlike Grossman (1986), they used refinery production, smelter produc tion, mine production and mining employment as indicia of injury. For each of these four measures, real GNP was used to account for demand shifts. Supply shifts were accounted for by a time ttend and die ratio of wages in copper mining relative to average U.S. wages. Foreign competition was captured by including the imported quantity of the appropriate form of copper into each equation. The parameters of the four reduced form models were estimated using annual data for the years 1950 to 1983. Pindyck and Rotemberg found fi-om their simulations that for each index of injury, low GNP and high wages had a greater impact on injury dian did die level of imports. On die basis of these and odier results, diey concluded diat imports "hardly seem to be a 'substantial cause' of injury to the domestic cop per industiy". In a sttidy more closely related to die softwood lumber dispute, Kelly (1988) analyzed die detenninants of injury in the U.S. western red cedar shingles and shakes industiy. Kelly used die level of production as the measure of injury. However, unlike the other studies of injury causality, which were based on estimation of reduced form models, Kelly worked out die algebra for decomposing die changes in domestic production into changes in domestic demand, changes in domestic supply and changes in import supply. This decomposition follows from die assump tion diat tiie demand, supply and import functions are linear. Kelly also showed diat die decomposition can be ex pressed in terms of demand, domestic supply and import supply elasticities. Using estimates of elasticities obtained or estimated by die ITC to establish relevant ranges, diis study found diat imports from Canada were not a substan tial cause of die decrease in domestic shingle and shake production over die period from 1978 to 1984. Furthermore, Kelly found diat the observed changes in price, domestic production and imports showed that the decrease in domestic demand was the greatest cause of injury to die U.S. western red cedar shingles and shakes industiy. When he fo-cussed on just die 1983 to 1985 period, Kelly found that the greatest cause of injury was a shift in die domestic sup ply curve due to a decreasing supply of western red cedar logs suitable for making shingles and shakes. Kelly con cluded that imports of shingles and shakes from Canada were not a cause of injury "but were radier a consequence of the time causes of injury". There has been a considerable amount of research focusing on the North American forest industiy and how it reacts to various poUcy changes. Adams and Blackwell (1973) developed one such model to forecast annual changes in the U.S. lumber and plywood markets. This model accounts for interactions between die U.S. markets for lum ber, plywood, saw logs, veneer logs and stumpage and was estimated using annual data for die years 1949 to to 1969. Adams and Blackwell modelled lumber imports from Canada as a function of U.S lumber consumption, die relative price of lumber in Canada and the U.S. and a time ttend. Because die motivation for diis research pertained to the concern diat the U.S. lumber industiy might not be able to satisfy the requirements of U.S. housing pro grammes at reasonable prices, diis paper focused on the sensitivity of the main U.S. endogenous variables (e.g., the price and consumption of lumber and plywood) to forecasts of housing demand and policies designed to curb lumber price increases. A simulated increase of one billion board feet per year (starting in 1970) in imported lumber from Canada caused the 1975 U.S. price of lumber to fall by less than two percent relative to the 1957 to 1959 average price. Based on diis and other results, Adams and Blackwell concluded diat die U.S. price of lumber was only mod erately sensitive to residential constiuction and price curbing poUcies because of the apparent ability of the industiy to adjust dieir production levels. Various studies have focused on die nature of U.S. lumber demand. Buongiomo and Chou (1983) used mondily data from 1974 to 1980 to estimate the demand elasticities for U.S. forest products. Addressing concerns over die level of Canadian lumber imported into the U.S., this paper found diat lumber imports were more sensitive to die U.S price of lumber than to eidier die import price, the level of residential housing or the U.S. price of all odier commodities. Spelter (1985) also analyzed the nature of U.S. lumber demand. He estimated a logistic product diffusion model diat allowed for a decline in die price elasticity of lumber over time. Spelter simulated die effect of a tariff on lumber imports and concluded that higher domestic prices for lumber would not dramatically affect con sumption. Adams and Haynes (1985) also addressed die increasing controversy in the U.S. over imports of Canadian lumber. For this study, Adams and Haynes used a regional multiple product partial equilibrium market model (known as the Timber Assessment Market Model) to study die importance of factors influencing lumber trade be tween the U.S. and Canada^. Not surprisingly, this study found that production costs, lumber recovery efficiency and a declining accessibility of timber in Canada were critical influences on level of future lumber trade. Adams and Haynes also found that even though lower wood costs in Canada accounted for some of the rise in Canada's market share, tirends in other factors, such as interest rates, demand, Canadian improvements in processing and harvest lev els, also played important roles. Swanson and Jacques (1985) focussed exclusively on the factors diat influence relative shares in tiie U.S. softwood lumber market. Using annual data for the years 1970 to 1983 the ratio of the Canadian to the U.S. share of the U.S. softwood lumber market was modelled as a function of the exchange rate, the ratio of Canadian to U.S. chip revenues, the ratio of Canadian to U.S. average variable costs and the number of U.S. single unit residential housing starts. Swanson and Jacques calculated die ratio of Canadian to U.S. average variable costs using own coun-tiy currency values because they assumed diat "competing foreign producers consider costs and revenues in dieir own respective currencies, and not in the currency of their foreign competitors, before any decisions concerning invest ment are made". This study concluded diat die exchange rate was die most significant factor explaining die rise in Canada's market share. Adams, McCarl and Homayounfarrokh (1986) estimated a regional econometiic model to study the impor tance of the exchange rate on U.S.-Canadian lumber trade. This study of lumber and stumpage markets used data for the years 1950 to 1983 and divided the North American lumber market into a U.S. demand region and five supply re gions: four in the U.S. and one in Canada. Capacity adjustinents over time were also modelled. Based on various exchange rate simulations, die authors concluded diat Canada-U.S. exchange rate played a definite role in die expan sion of die Canada's market share, but a stable exchange rate would not have completely eliminated die rise in Canadian market share over die 1975 to 1979 period. Using a partial equilibrium, regional model of die North American lumber market, Boyd and Krutilla (1987) estimated the production and welfare impacts of U.S. tariff and quota restiictions on imports of Canadian lumber. In The regional and product breakdown of the the Timber Assessment Market Model is described in Adams and Haynes (1980). this spatial market model, die North America lumber market was divided into 26 U.S and eight Canadian supply re gions. The U.S. market was divided into 39 demand regions. The audiors claim that die level of disaggregation was necessary because of die complex stiiicture of inter regional ti-ading patterns. Boyd and Krutilla found diat die U.S. regions diat experienced die biggest relative gains from a ten percent tariff (on imports of Canadian lumber) were those in the west that competed directiy with British Columbian lumber producers. Furthermore, although a ten per cent tariff benefited U.S. producers and the U.S. treasury, the net effect on the North American lumber market was negative, reflecting the efficiency costs associated with market distorting policy actions. Kalt (1987) also studied the welfare impacts of a U.S. tariff on imported Canadian imports. Kalt divided die North American market into U.S. demand, U.S. supply and die supply of Canadian imports and assumed constant elasticity functional forms for these relationships. Based on elasticities cited in the literature, this study estimated the partial equilibrium impacts of a 15 percent tariff The results in this study are similar to those in Boyd and Krutilla (1987), i.e., U.S. producers and the treasury gained more than U.S. consumers lost, but the North American market was found to be worse off (due to losses by Canadian producers). Kalt (1987) also empirically estimated die relationship between the Canadian supply of logs and a measure of die stumpage subsidy using data from 1977 to 1984. Canadian log supply was estimated simultaneously widi the total demand for logs in die U.S. and Canada. Kalt's results, based on constant elasticity functional forms, did not produce any evidence that Canada's system of stumpage subsidies resulted in an increase in Canada's supply of logs. Chen, Ames and Hammett (1988) developed and estimated a model of die U.S. lumber market consisting of four simultaneous equations. The model was estimated, using annual data for die years 1965 to 1985, to assess die impacts of a tariff on imported lumber into die United States. The results of diis stiidy indicated diat lumber im ports from Canada were affected by die U.S. price of lumber, die Canadian price of lumber, and die square footage of U.S. residential construction. The U.S. supply of lumber was found to be inelastic (0.309). Thus, in the short-run, it was concluded that a tariff induced price increase would result in windfall gains to U.S. lumber producers. Roberts (1988) is a departure from diese econometric studies of the U.S. lumber market and die Canada-U.S. lumber dispute. This study looked at the broader question of the relationship between the competitiveness of the Canadian forest products sector and die exchange rate by assessing the impact on indices of international competi-tiveness to a ten percent appreciation in the U.S. value of the Canadian dollar^. The competitive index for softwood lumber was defined as die ratio of U.S. to Canadian average variable costs expressed in a common currency. In his review of the lumber sector, Roberts found that 1987 cost conditions, and an approximate ten percent appreciation of the Canadian dollar caused die effective cost of lumber production in Canada to change from being seven percent lower to three percent higher than in die United States. By separating changes in the lumber competitive index, it was found that changes in relative costs in die U.S. and Canada tend to be greater (in terms of variance) than changes in the exchange rate. Roberts concluded that this suggests that the exchange rate was not the most volatile factor de termining die relative competitiveness of Canada's lumber industiy. Based on diis stiidy, it would seem diat changes in bodi die exchange rate and relative costs of production are important considerations to any competitiveness assess ment of Canada's lumber sector. Buongiomo and Chavis (1988) is another example of research that focuses on the importance of the ex change rate to U.S. imports of Canadian lumber. This study used time-series analysis of monthly data from January 1974 to January 1986. Feed-back measures and long-run multipliers between imports, the exchange rate and die U.S. price of lumber were also calculated. Buongiomo and Chavis found that 68 percent of die rise in imports (during the period smdied) was due to a rise in die U.S. price of lumber, where as the exchange rate was not found to have a significant effect on imports. The audiors also found diat the observed increase in imports did not lead to a decrease in die price of lumber received by U.S. producers. The insignificance of the exchange rate widi respect to its effect on imports was counterintuitive given die results of previous studies (discussed above). The long-run mul tiplier of the exchange rate on imports was reported to be 0.46. This is the same as the elasticity of imports with respect to the exchange rate reported in Adams McCarl and Homayounfarrokh (1986); however die standard error for Buongiomo and Chavis' multiplier was 0.70, hence the conclusion that changes in the exchange rate have not af fected imports. Buongiomo and Chavis conclude diat because die information in their study results from a "model equivalent to a reduced form", they cannot identify die mechanisms diat cause the zero net effect of die exchange rate on die level of U.S. imports. This review of die literature focusing on die North American lumber market is not complete. However, it does present a clear indication of die various econometric approaches diat researchers have adopted to explain Roberts (1988) looked at die softwood lumber, newsprint and market pulp sectors. Canadian and U.S. lumber production, prices and trade flows. Disputes between Canada and the U.S. over softwood lumber have motivated practically all of die research discussed above. However, these smdies have tended to focus on welfare effects and partial equilibrium adjustinent of lumber production and prices to ti^ade restiictions. These studies have not attempted to address die broader question of injury and causality per se. This is not to say that models like those developed by Adams and Haynes (1980) and by Boyd and Krutilla (1987) could not be used to quantify some of the measures of injury addressed in diis study because diey clearly could and they would help to provide insight. However, diese sorts of models were developed to answer more stiategic questions pertaining to changing technologies or forest regulatory policies for example. The question of injury in ITC investigations requires a framework diat is less data intensive than partial equilibrium spatial market models. Furthermore, aldiough diis point may be arguable, injury investigations require a framework that imposes less stiiic-ture on die production and consumption relationships. Approaches similar to diose found in Grossman (1986) and in Pindyck and Rotemberg (1987) are more appropriate than market models for analyzing statutory notions of injury. The reduced form econometiic framework used in diis study is similar to that used by Grossman (1986) and by Pindyck and Rotemberg (1987), but is unique in that it looks at a much broader array of injury measures. The six measures of injury assessed in this study are among the most often cited indicators of performance in die forest sector. A complete evaluation of the primary determinants of these measures will remove much of the ambiguity that has surrounded die sources of die injury cited by U.S. lumber producers in their petition for protection against allegedly unfair lumber imports from Canada. 3.3 Measures of Injury In conducting its injury determinations, the ITC must evaluate all relevant economic factors that have a bearing on the state of a petitioning industry. Secdon 771 of the U.S. Tariff Act of 1930 specifies that the ITC must consider, but is not limited to, at least 15 different economic factors in dieir investigations^. Of diese factors, this paper considers impacts on prices, output, market share, profits (defined as the return on sales and the stock price) and employment in its assessment of the injury suffered by U.S. sawmills during the period from 1951 to 1986. A change in price is die first of die six measures of injury addressed in diis paper. The price of a product in time t (Px,t) is given by its inverse demand function and can be defined as follows: Px,t = Pt(Xt.It.Pst.Pct.Zt). (1) where Xj is the quantity of good X demanded in period t. If is the income of consumers in period t, Pgf and Pcj are vectors of prices for substitutes and complements, respectively, and Zt is a vector of odier exogenous demand vari ables. Assuming diat demand is downward sloping, anydiing diat increases (decreases) die supply of X will decrease (increase) its price. The second measure of injury considered in this paper is a change in the level of output. A firm's level of output (Xf) depends on its production function which can be defined as follows for the general case of N inputs, xt = ttjFj + +...+ ttj^Fj^ , (2) Expression 2 shows that increases (decreases) in the use of factor Fj result in an increase (decrease) in the level of output In addition, technological changes that increase (decrease) die value of «i (die quantity of input i necessary per unit of the output x), also increase (decrease) the level of output. Also, assuming profit maximization, an increase (decrease) in die demand or a decrease (increase) in die marginal cost will increase (decrease) a firm's profit maximizing level of output. The next measure of injury to be considered is a change in market share. Market share can be expressed in terms of quantity or value. For a tiaded good, die market share of domestic and foreign producers (SHRO is equal to Section 771, subparagraph (c), parts (ii) and (iii) of die U.S. Tariff Act of 1930 lists die following 15 economic factors: prices, output, sales, market share, profits, productivity, return on investment capacity utilization, cash flow, inventories, employment, wages, growdi, abiUty to raise capital and investment. the quantity of their respective shipments (consumed in a particular market) divided by the total quantity of ship ments (consumed in the market) and can be defined as follows: SHRi =^ , where (3) ^t M Xt= Ixit .M=l,2 . (4i=l and where xit is die quantity shipped to a particular market in time period t by country i. Anything diat increases (decreases) die quantity of imports relative to die quantity of domestic shipments will decrease (increase) die share of the market held by domestic producers. From an economic perspective, perhaps die most natural measure of injury is lost producer surplus or prof its. The ITC can view a substantial decline in industiy profits caused by unfair imports as an indication of material injury. Although there are many measures of profitabiUty, I consider return on sales (ROS) and stock prices (SP). ROS can be defined as foUows: • <" where TCJ is a firm's profit in period t. A firm's profit in any particular time period is described by its profit function which can be defined as follows: TCt = xt Pt(Xt) - wtLt - rtDt - OCt . (6) where Xj is die quantity of good x produced in period t, Pt(Xt) is the inverse demand for good X, Wj is the wage rate in period t, is die amount of labour used in period t, rj is die interest rate in period t, is die total debt in period t and OCt represents odier costs incurred in period L ROS is a short-run performance measure for a particular time period (e.g., a year). Expression 5 shows diat anything diat increases (decreases) a firm's profits per unit of output wUl increase (decrease) its ROS. The fifdi measure of injury considered is a change to the economic value of a common stock. The eco nomic value of a common stock is equal to its market trading price. Provided we define dividends sufficientiy broadly, dien by adopting a "fundamentalist" view we would expect diat a firm's stock price (SP) to be equal to die present value of future returns as iUusttated by Expression 7. t = oo SP = (1+r)' TTf/Nt ,t (7) t = 0 where is defined by Expression 6, is die number of shares in period t and r is die discount rate. Expression 7 could be modified to allow for different discount rates in each time period. Expressions 6 and 7 indicate diat factors which increase (decrease) a frnn's expected profits will also increase (decrease) its stock price. Lost employment or man-hours is anodier measure of injury. It can be shown diat for a given level of out put during a time period, a cost-minimizing firm should employ its factors of production up to the point where the ratios of the marginal factor productivities to dieir respective marginal oudays are equal. Thus, for a given level of output, an increase (decrease) in the marginal productivity of labour or a decrease (increase) in the marginal outlay for labour will result in an increase (decrease) in the use of labour. Also, changes in the level of output will also affect the demand for labour. A decrease (increase) in the level of output will decrease (increase) the demand for labour. Thus, changes in die use of labour need to be factored into output effects and substitution effects (i.e., movements along die expansion padi versus shifts in die expansion patii). From this Usting and brief discussion, it would appear that die various measures of injury are related to one anodier. For example, a change in die output decision affects die value of die profit function, die return on sales and the distiibution of market shares. However, even diough these measures are related, they are notperfecdy correlated. Consequendy, depending on the nature of the underlying economic shock, the various measures of injury may or may not move together (i.e., in the same direction). The next section explores the relationship between the six mea sures of injury Usted above. 3.4 Relationship Between Measures of Injury This section shows how the six measures of injury listed above respond to various economic shocks. This analysis is done from the perspective of die U.S. and Canadian softwood lumber market The response of die injury measures are determined for two demand shocks (higher U.S. income and higher U.S. housing starts) and two supply shocks (higher Canadian stiimpage prices and a higher Canada-U.S. exchange rate). For each of the four cases, the effects of die shock are illustiated algebraically and diagrammatically. Assume for simplicity that the U.S. and Canada produce softwood lumber and that die U.S. imports lumber from Canada. Furthermore, assume that Canada does not import lumber from the United States. Under diese condi tions, die U.S. will have excess demand (fdled by Canada) and Canada will have excess supply (shipped to the U.S.). Let die U.S. excess demand (XDus) be defined as follows: XDus = XD(PL; Y. H, X) , (8) where PL is die price of lumber in U.S. dollars, Y is die level of U.S. income in U.S. dollars, H is the square footage of new U.S. residential construction and X is a vector of other shift variables. The Canadian excess supply (XSc) can be defined as in Expression 9. XSC = XS(PL;SP,E,Z) , (9) where PL is defined as above, SP is the Canadian stumpage price in Canadian dollars, E is die exchange rate in terms of the number of Canadian dollars per U.S. dollar and Z is a vector of otiier shift variables. Assuming diat in any period the market clears and there are no inventories, then Canadian excess supply will equal U.S. excess demand. Furthermore, changes in Canadian excess supply must be balanced by corresponding changes in U.S. excess demand, and visa versa. Thus, the total derivative of die Canadian excess supply function must equal the total derivative of the U.S. excess demand function. This condition is shown by Expression 10. axs_,^ âxs,^„ 3xs^^ dxD^^ axD , axD -—dPL + -—dSP + r^dE = -—dPL + r—dY + dH (10) aPL asp aE aPL aY an • ^ ^ The Canadian and U.S. markets, as well as the excess supply and demand curves are illusti-ated in Figures 3.1.1 to 3.1.3. In Figure 3.1.1, Sc is die supply of lumber and Dc is die demand for lumber in Canada. Widiout trade, die price and quantity of lumber in Canada would be PLC and QLC- Similarly, in Figure 3.1.3, Sus is die supply of lumber and Dus is the demand for lumber in die United States. Widiout ti-ade, die price and quantity of lumber in the U.S. would be PLUS and QLUS- Figure 3.1.2 shows die Canadian excess supply curve (XSc) and die U.S. ex cess demand curve (XDus) which correspond to Figures 3.1.1 and 3.1.3. With trade, the price of lumber rises in Canada and falls in die Urates States to PLT- Furthermore, with trade, Canadian exports equal AB, and U.S. imports equal CD. Figure 3.1.1 Canadian Lumber Maiket Figure 3.1.2 Equilibrium Condition Figure 3.1.3 U.S. Lumber Market •^LUS 'LT ^us\ / /c N ^us From the equilibrium condition in Figure 3.1.2,1 will now review the effects of a series of economic shocks on the injury measures considered in diis study. For each case, I allow only one of die exogenous variable to change. The results of this analysis are summarized in Table 3.1. For the first shock, I assume a one time increase in U.S. income. From Expression 10, this scenario is characterized algebraically as follows. axs ^ 9XD axD Collecting like terms and factoring yields. dPL axs axD axD laPL aPtJ ay Dividing bodi sides by dY results in, axD aY dY ^ dY "axs axD • (11) aPL aPL Expression 11 shows that the effect of an increase in Y on PL depends on die relative slopes of die excess supply and demand curves with respect to PL and Y. What can be said about die sign of Expression 11? Focusing on the right hand side, we know that the numerator must be positive because demand rises widi an increase in in come. Also, the first term in the denominator must be positive because supply rises with an increase in price. Finally, the second term in the denominator must be negative because demand falls with a rise in price. Thus, it would seem that an increase in income produces a rise in die U.S. price of lumber. This result can also illustrated in Figures 3.2.1 and 3.2.2. The rise in Y shifts Dus to D'us- This causes an upward shift in XDus to XD'us and a rise in the lumber price from PL to P'L. Figure 3.2.2 also shows that the increase in Y causes U.S. lumber imports to rise from AB to CD and U.S. lumber production to rise from OE to OF. Because die rise in imports is greater than tiie rise in U.S. production, Canada's market share rises. Figure 3.2.1 Figure 3.2.2 Equilibrium Condition: Rise in U.S. Income U.S. Lumber Market: Rise in U.S. Income 0 E F The increase in U.S. production will require more labour dius, the increase in Y also increases die number of production hours. Increases in U.S. price and production raise profits by AGHC dius, according to Expressions 2 and 3, die average stock price of the U.S. lumber industry will rise. Finally, die effect of the rise in Y on the return on sales depends on the form of die U.S. supply function respect to price because functional form affects relative profit and quantity responses to an income induced price changes. Thus, die effect of a rise in Y on the return on sales is indeterminate. For the second shock, assume that there is a one time increase in die number of U.S. housing starts due to a low income mortgage subsidy programme for example. An increase in housing starts will increase H, the square footage of U.S. new residential construction. This scenario is depicted by die following expression. axs ^ ÔXD ^ axD ^„ Collecting like terms, factoring and then dividing both sides by dH yields. axD aPb' 3PL Thus, die effect of an increase in H on PL depends on the relative slopes of the excess supply and demand curves with respect to PL and H. The denominator of the right hand side of Expression 12 is the same as in Expression 11 dius, it must be positive, as before. The numerator must also be positive because the demand for lumber is derived from its end uses and residential constiuction utilizes the largest proportion of U.S. lumber con sumption. Consequendy, it would seem diat an increase in the square footage of new residential constiuction raises die U.S. price of lumber. Graphically, this scenario is die same as an increase in per capita income. Consequendy, the impacts on die odier measures of injury are die same. The directions of these effects are listed in Table 3.1. To see die importance of die exchange rate on the performance of the U.S. lumber industry assume diere is an appreciation of die U.S. dollar relative to die Canadian dollar. From die perspective of U.S. lumber producers, an increase in die number of Canadian dollars per U.S. dollar can be symbolized by die following expression. axs,_ axD^^ dxD^^ Collecting Uke terms, factoring and dien dividing both sides by dE yields, axD dE axs ajcD • ^ ^ aPL' aPL As before. Expression 13 states that die effect of the rise in die relative value of the U.S. dollar on the price of lumber depend on die slopes of die excess supply and demand curves. The denominator of die right hand side of Expression 13 is positive; however, in diis scenario, die numerator is negative. This is because die appreciation of die U.S. dollar lowers die U.S. dollar value of Canadian production costs. This result is illustrated in Figures 3.3.1 and 3.3.2. The rise in E shifts XSc to XS'c- This causes an fall in die lumber price from PL to P"L. Figure 3.3.2 also shows diat die inaease in E causes U.S. lumber imports to rise from AB to CD and U.S. lumber production to fall from OE to OF. Because imports rise and U.S. production falls, Canada's market share rises. Lower U.S. production lowers die demand for labour and, dius, die number of production hours will fall. Lower prices and production translate into a drop in profits equal to AGHC dius, according to Expressions 2 and 3, the average stock price of the U.S. lumber industry will fall. Finally, the effect of the rise in E on the return on sales depends on the form of die U.S. supply function. Thus, widiout additional information on nature of U.S. lumber supply, die effect a rise in E on die return on sales is indeterminate. Figure 3.3.1 Equilibrium Conditions: Rise U.S. $ Figure 3.3.2 U.S. Lumber Market: Rise U.S. $ O F E For the final economic shock, consider what would happed as the result of a one time increase in the Canadian stumpage price (SP). Algebraically, this economic shock can be characterizes as follows. axS^^ ôXD^^„ dXD^^ dPr + dSP = dPr aPL asp aPL ' Collecting like terms, factoring and tiien dividing both sides by dSP yields, axD asp dPL dSP axs axD ' (14) aPL aPL The effect of a rise in SP on die price of lumber in die U.S. is positive because bodi die numerator and de nominator on the right hand side of Expression 14 are positive. The numerator is positive because higher stumpage prices raise Canadian production costs and tiiis lowers Canadian excess supply. The price of lumber must rises to ra tion this decrease in Canadian supply. Figures 3.4.1 and 3.4.2 illusti^ate diis result. The rise in SP shifts XSc "P to XS'c and this causes die lumber price to rise from PL to Pi,- Figure 3.4.2 also shows that U.S. lumber imports to fall from AB to CD. U.S. lumber production also rises in diis scenario from OE to OF. Because imports fall and U.S. production rises, Canada's market share falls. Figure 3.4.1 Equilibrium Conditions: Rise Can. SP Figure 3.4.2 U.S. Lumber Market: Rise Can. SP p; XS'^ xs^ ^USN / 'us H C A/ \B G O E F An increase in U.S. production increases the demand for labour and, dius, the number of production hours will rises. U.S. profits rise by AGHC because if higher prices and production. Consequendy, the average stock price of die U.S. lumber industry wUl rise. Finally, as before, die effect of a rise in SP on die return on sales is in determinate. Table 3.1 indicates diat die impact on die return on sales was inconclusive for each of the four economic shocks discussed. As mentioned above, this is because die direction of change in die return on sales depends on the functional form of the cost function. To see this, consider the following special case of a linear supply function. Qs = aPL It can be easily shown that the profit (7t) for a given lumber price implied by this simple supply function is defined by the following expression. n = ap2 Substitiiting lumber output into Expression 3 yields die following expression for the return on sales. ROS = Simplifying yields. R0S=2-Thus, in diis example of a linear supply function widi no intercept, die retiini on sales is 50 percent for all prices. Table 3.1 Review of Impacts on U.S. Lumber Industry Injury Measures Scenario Injury Measure Increase in U.S. Per Capita Income Increase in U.S. Residential Housing Starts Increase in die Can-U.S. Exchange Rate Increase in die Canadian Stumpage Price U.S. Lumber Price + + - + U.S. Lumber Production + + - + Can. Market Share + + + -U.S. Employment + + - + U.S. Remm on Sales 7 ? ? ? U.S. Avg. Stock Price + + - + If the example is changed to a linear supply function with a positive intercept, it can be shown that price increases will result in an increase in the return on sales, but at a decreasing rate. That is, die larger die intercept the lower die return on sales for die initial output, but as output increases, die retiim on sales rises toward 50 percent If the intercept is negative (an unlikely condition), then price increases will result in a decrease in the return on sales. The impacts of the four economic shocks indicate diat while die measures of injury are related, tiiey are not perfecdy correlated. Thus, depending on die nature of an economic shock, dieory shows that the measures may or may not move in the same direction. The question of how diese measures move and whether they actually move to-gedier in practice is an empirical question. This study analyzes diese six measures of injury for the U.S. softwood lumber industiy and attempts to empirically determine die relative importance of factors that have affected diem over time. The next section briefly discusses how these U.S. softwood lumber industry performance measures have changed over time. 3.5 Empirical Overview This section presents a brief historical overview of the six performance measures used to assess the perfor mance of the U.S. softwood lumber industry. Figure 3.5 shows the U.S. softwood lumber price index. Most of die major U.S. recessions are reflected in this plot of die U.S. lumber price index, most notably the recessions in 1970, 1975, and 1982. As one would expect, die lumber price index is related to U.S. lumber production and saw log prices. Figure 3.5 Plot of the U.S. Softwood Lumber Price Index For die 1950 to 1986 period, diere was a 0.61 simple linear correlation coefficients between die U.S. soft wood lumber price index and U.S. lumber production. The simple linear correlation coefficients between die U.S. softwood lumber price index and the real Douglas-fir saw log price in Washington and Oregon was 0.85 for die years 1950 to 1985. This high correlation illustrates the sti-ong linkage between die lumber and stumpage markets in the U.S. and the effect of U.S. saw timber shortages in die U.S. Pacific Northwest on die U.S. lumber market. Figure 3.6 shows the level of U.S. softwood lumber production and die number of U.S. residential housing starts for the years 1950 to 1986. Figure 3.6 shows that lumber production is sti-ongly correlated (76 percent) with the number of housing starts, and housing starts is an important indicator of aggregate economic activity. Conse quendy, lumber production is also procyclical. 36 34 32 30 28 26 24 \ r 2.40 2.20 2.00 1.80 1.60 1.40 1.20 1.00 D S) 3) in o in CD in o CD O) o> in O) o CD o> in 00 a> —D- U.S. Lumber Production (Billion fbm) • U.S. New Housing Starts (Million) Figure 3.7 Plot of Canada's Share of U.S. Softwood Lumber Consumption 35 oinoinoinoin inincDcoh-r-oocD 0)0>0)0>0>0>0>0> Figure 3.7 shows die Canadian share of die U.S. softwood lumber market for die years 1950 to 1986. Canadian market share has been trending upward at a compound rate of 5.1 percent per year diroughout die 37 year time period. A comparison of Figures 3.5 through 3.10 does not suggest a sti-ong positive business cycle relation ship between Canada's market share and aggregate measures of U.S. economic activity. However, if the period after the Canadian dollar was allowed to float in 1970 is considered, dien diere is a stix)ng relationship between Canadian market share and die U.S./Canadian exchange rate. If only die period between 1972 and 1986 is considered, then the correlation coefficient between die Canadian market share and die nominal U.S./Canadian exchange rate is 0.93^. Figure 3.8 Plot of die Weighted Average Return on Sales for U.S. Sawmills 14 T 0 I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I oinoinouooin 0)0)0)0}0>0)0>0) Figure 3.8 shows the weighted average return on sales for the companies comprising the Standard and Poor's Forest Products Index. A comparison of Figures 3.8 and 3.9 indicates that stock prices and ROS do not al ways move in unison with one another. For example, even though the stock prices of forest products companies were falling during the years from 1970 to 1975, die average ROS peaked in 1973 due to a rise in softwood lumber prices relative to costs. The ROS of the U.S. forest products industry shows considerable fluctuation during the period 1950 to 1986. All of these fluctuations correspond widi periods of growth and set backs in die economy as a whole. There were eight main recessionary periods in die U.S. during die 37 years starting in 1950. The post Korean War reces sion in 1954 was due, in part, to reduced defense spending. The recessions of 1959,1961, and 1967 reflect periods of tight money and rising interest rates. The 1970 recession has been attributed to a strike at General motors and a decline in defense spending. The relatively severe recession of 1975 was due to a sharp increase in OPEC oil prices due to an embargo on Arab oil destined for the U.S. The brief recession in 1980 was due to a second oil shock caused by the Iranian revolution in 1979. Finally, the 1981-82 recession is generally attributed to a period of restiic-' The correlation coefficient between the Canadian market share and the real (1980) U.S./Canadian exchange rate is 0.91. live, anti-inflationary monetary policies adopted in the U.S. and elsewhere. Figure 3.8 shows that the ROS of the U.S. forest products industry declined in all but the 1980 recession. Consequendy, the ROS measure of profitability is also very procyclical. For an industry, die stock price is an aggregate concept and measures die present value of aggregate future dividends for an indusdy as a whole. Anodier way to assess industry profit is to aggregate the stock prices of a rep resentative group of firms. This was the approach used in this study. Figure 3.9 Plot of Standard and Poor's U.S. Stock Price Indexes: Forest Products and die'400 Top Indusûials tn o tn o in CD 00 CD o> o> o> o> o> D Forest Products • 400 Industrials Real Price Index Real Price Index (1986=100) (1986=100) Figure 3.9 shows the value of Standard and Poor's Forest Products Stock Price Index and die 400 Industrials Stock Price Index for the years 1965 to 1986. The Forest Products Index is a composite of stock prices for die fol lowing companies: Boise Cascade, Champion International Corporation, Evans Products^^, Georgia Pacific, Louisianna-Pacific Corporation 11, Podatch Corporation, and Weyerhaeuser Company. The forest products index and the industrials index tend to move together, but they are not perfecdy correlated as indicated, for example, by the dif ferent performances during die 1970 business recession. From 1972 onwards, the two stock price indexes move to gether quite closely, although the forest products index shows more variability. The effects of the 1974 and 1982 re cessions can be clearly seen in die forest products index. This is because die softwood lumber and odier forest prod-10 Evans Products is included in this index only up to April 11,1984. Louisianna-Pacific Corporation was formed and included in the Standard and Poor's Forest Products Index in 1973. ucts sectors are widely regarded as being procyclical industries, i.e., their performance is positively correlated to die business cycle. Figure 3.10 Plot of U.S. Sawmill Production Hours 800 T Million Hours oiooinoinoio 0)C)0>0)0)0>0)0) Figure 3.10 shows the number of hours worked by U.S. sawmill production workers^^ Over the 37 years starting in 1950, die number of workers has fallen by about 73 percent. This decline in die use of labour is coinci dent with a steady rise in the real hourly wage of sawmill workers. In fact, over the 1950 to 1986 time period there is a - 0.77 simple linear correlation between the number of workers and dieir wage level. The trend in the number of hours worked is not closely related to the level of activity in the U.S. economy. However, small dips in sawmill employment can be seen for the years 1975 and 1982. Sawmill production workers includes hourly production workers in both hardwood and softwood lumber mills. Salaried staff are not included. 3.6 Theory Development The starting point for a causal analysis of injury suffered by American softwood lumber producers is a spec ification of a reduced form model. In turn, the derivation of a reduced form model requires a structural model. For convenience, unless specified odierwise, lumber will mean softwood lumber. Due to the natiire of die tiade dispute, two products need to be considered. In die petition filed widi die ITC by die American Coalition for Fair Lumber Imports, it was claimed diat Canadian softwood lumber was entering the U.S. at less dian its fair value because (among other tilings) it was manufactured using timber that was being sold at subsidized prices. Consequendy, bodi softwood lumber and softwood logs must be considered in the analysis^^ If it is assumed that logging is conducted by market loggers or by vertically integrated lumber producing firms that transfer logs between dieir logging and milling divisions at market prices, dien Canada's aggregate log ging profit functions can be defined as follows^^^: jtLG = XPLG(XLG)-wL-rD-OCLG + SLGXLG.whCTe (15) jtLG is die profit from producing XLG of logs, PLG(XLG) is die inverse derived demand for saw logs priced in Canadian dollars, w is the logging wage per hour, L is the quantity of labour used to produce XLG of logs, r is the interest rate, D is die total logging debt, OCLG represents other logging cost (e.g., materials, fuel, electiicity etc.) and SLG is the alleged Canadian subsidy per m^ of log. The Canadian derived demand for saw logs will be a function of the price of saw logs, die prices of other factors related to the production of lumber, the price of lumber and the exchange rate. Thus, die demand for logs will have die following general form: X LG= / (PLG. ILM. PLM. E) , where (16) PLG is die saw log price, ILM are factor prices (for factors odier dian logs) used in the production of lumber, PLM is die price of lumber and E is die Canada-U.S. exchange rate^^. Timber and logs are not die same commodity. The term timber refers to the uncut frees (i.e., ti-ee volume on tiie stump), whereas logs are the manufactured product diat result from die harvesting of timber. Thus, logs are die input (not timber) in the production of lumber. The time period subscript has been omitted for convenience. This specification assumes tiiat lumber is priced in U.S. dollars. A subsidy per of harvested and scaled timber is equal to die difference between die true economic value of the right to harvest timber (EVRT) and die appraised value (i.e., die stumpage price SP). Thus, a timber subsidy can be represented generally as follows: SLG = EVRT - SP , (17) where SLG is defined as above and SP is die appraised stumpage price per m^ of harvested and scaled timber. However, in practice, die variable EVRT cannot be observed. An observable proxy variable diat is a function of SLG must be specified. The total rent per m^ of log delivered to a sawmill equals the market price of logs minus the appraised stumpage price and costs. Thus, die rent per m^ of log can be illusti-ated as follows: RLG = PLG - HC - TC - OCLG - SP , where (18) RLG is die rent per m^ of log, PLG and SP are defined as above, HC is die harvest cost per m^, TC is die ti-ansporta-tion cost m^ and OCLG are odier costs (e.g., the return to capital) per m^. Expression 17 can be rewritten in terms of the stumpage price as follows: SP = EVRT-SLG (19) Substituting Expression 19 into Expression 18 yields die following relationship: RLG = PLG - HC - TC - OCLG - EVRT + SLG • (20) Letting the variable C equal die sum (HC + TC + OCLG). Expression 20 can be rewritten as follows: RLG = PLG-C-EVRT + SLG- (21) Thus, the rent per m^ of log delivered to a sawmill is correlated with the timber subsidy. That is, everydiing else being equal, an increase (decrease) in die subsidy will result in an increase (decrease) in die rent per m^ of log. This idea of using leaked rents as a proxy for die timber subsidy is based on Kalt (1987) who used a similar definition for the alleged subsidy in his evaluation of the lumber dispute. Substituting leaked rent for die subsidy in Expression 15 yields Canadian regional logging profit functions of the following form: nLG = XLGPLGPfLG)-wL-rD-OCLG + RLGXLG- (22) Consequendy, logging profits in Canadian regions will increase (decrease) with an increase in die level of leaked rents. Widi respect to lumber markets, Canadian lumber profit can be defined as follows: îtClJ^ = xcLMPLM(XLM)E-wcLMLCLM-rDCLM-OCcLM . (23) where 7CCLM is the profit from producing xcLM of lumber, XLM is the total quantity of lumber consumed in the U.S. and Canada, PLM(XLM) is die inverse derived demand for lumber priced in U.S. dollars, E is die exchange rate (as above), WCLM is die average hourly wage in Canadian saw mills, LCLM is the quantity of labour used in Canadian saw mills, r is the discount rate, DcLM is the total saw mill debt and OCcLM represents other Canadian lumber manufactiiring costs (e.g., the cost of logs, fuel, electiicity and odier materials). Similarly, the U.S. lumber profit function can be defined as follows: rtUU^ = xuLMPlJ^(XLM)-wuLMLuLM-rDuLM-OCuLM, (24) where die variables TCULM . xuLM. PLM(XLM). WULM, 1-UIM< r. ^ULM and OCuu^ are defmed as above. The derived demand for lumber will be a function of the price of lumber and demand shifters (e.g., U.S. and Canadian income, the interest rate, prices of substitutes and complements etc.). Thus, the demand for lumber will have die following general form: XLM = / (PLM. IC IU. r. E, d^), (25) where PLM is the price of lumber, Ic and ly are the incomes in Canada and the U.S. respectively, r is the interest rate, E is the exchange rate and dm are other demand shift variables. In summary. Expressions 15 through 25 imply tiiat die alleged injury to American lumber producers can be modelled as a function of exogenous variables pertaining to Canadian regional log markets, the Canadian lumber market, die U.S. lumber market and die U.S. economy!^. In general, an index of die injury to American lumber produces can be represented as follows: iNJi = / (PLG. RLG. PLM. E, DLG. DCLM. DULM. WLG. WCLM. WULM . OCLG. OCCLM. OCCLM. IC. IU. dm ) . (26) Note diat die sfructural model in diis study excludes international effects. Softwood lumber is traded internationally; consequendy, the price of lumber in the U.S. is a function of tiie world price of lumber. Similarly, die odier U.S. sawmill performance measures are also functions of international conditions. However, including international variables in die stiiictural model would not add significantiy to the analysis because of die small fraction of U.S. and Canadian softwood lumber consumption supplied from outside of the U.S. and Canada. Furthermore, from a statistical perspective, because of collinearity between North American and international variables, it would be hard to separate die effects of conditions in the U.S. and Canada from those in international markets. where INJj are the various measures of injury (described by Expressions 1 through 7) and all the independent vari ables are defined as above. These reduced form equations do not contain die volume of Canadian lumber imported into the U.S. expliciUy. This is because the volume of U.S. lumber imports from Canada is an endogenous vari able, influenced by developments in the lumber sectors in the U.S. and elsewhere. A possible interpretation of die U.S. ITC mandate to determine die presence of material injury is that there be a statistically significant shift in a foreign excess supply function due to an alleged unfair trading practice. That is, determinations of material injury should be limited to cases where a significant quantitative link to an unfair prac tice can be shown. Injury caused by die changes in U.S. markets (e.g., changing preferences or relative increases in factor prices) or by increased "fair" competition should not be included in material injury determinations. Expression 26 is a reduced form equation for the six measures of injury. These six equations were estimated using annual data for die years 1951 to 1986. The next section describes the choice of exogenous variables included in the specifications for the reduced form models of the six injury measures. 3.7 Model Specificadon Specificadon of Expression 26 depends on die factors influencing supply and demand for lumber in the U.S. and Canada. Lumber is an intermediate good; consequendy, the demand for lumber in die U.S. and Canada is derived from the end uses of lumber. The main end uses for lumber are residential housing, home repair and remodelling, and other construction. Historically, residential constiuction alone has accounted for about 40 percent of U.S. and Canadian lumber consumption. In 1989, U.S. residential constiTiction and the activity in residential repair and re modelling accounted for 35 percent and 31 percent, respectively, of U.S. lumber consumption. Odier end uses for lumber in 1989 were nonresidential constiuction at 14 percent, materials handling at ten percent and all other uses at ten percent (Paper Tree Letter, October 1989 and Western Wood Products Association 1985)^'^. The derived demand for lumber can be determined from the cost functions of the primary end uses. Using this approach, die demand for lumber wdl be a function of its price, the prices of odier inputs and the level of die outputs. Focussing on all end uses would lead to an appropriate specification of die totid demand for lumber. However, for simplicity, I focus on residential housing construction, although the same estimation structure would arise from considering all end uses. Consequendy, the demand for lumber is derived from aggregating the production functions of individual house builders. By utilizing the duality between production and cost function (see Varian 1984), lumber demand can be obtained from the total cost function of residential construction. The determinants of the cost of residential housing constiuction are labour, capital, lumber, wood panel products (e.g., plywood), other materials (steel, bricks and cement etc.) and die number of residential housing starts. Thus, die cost of residential constiTiction (CR) is given by the following expression. CR = / (W, k, PL, Pp, M, USFOOT) , where w is the price of labour, k is the price of capital, PL is die price of lumber, Pp is the price of wood panel products, M is die price of other materials, and USFOOT is die square footage of residential construction. The square footage of residential construction or the number of residential housing starts can be used as measures of the output product The square footage of residential construction is a better indicator of lumber demand than housing starts because it weights die various kinds of residential housing (e.g., single family, multi-family and mobile homes etc.) differenUy whereas the number of housing starts gives these housing types equal weight ^ ^ Materials handling includes such uses as boxes, crates, pallets and dunnage etc. and all odier uses includes the use of wood by raihoads, mining and in furniture. The derived demand for lumber (QDL) using Shephard's lemma is given by die following expression. Thus, die demand for lumber derived from residential construction is given as follows: QDL = /(USFOOT, PL, W, k, Pp, M). The square footage of residential housing (USFOOT) is an end product and is determined by aggregating die demands of individuals. The demand for housing is a function of income, die price of housing and die prices of odier consumer goods. The decision to invest in a home is also quite sensitive to the interest rate. The total demand for housing will rise widi an increase in aggregate income and an expansion of die economy. Furthermore, total demand for housing will fall with an increase in the relative price of housing. Because the interest rate is an important de terminant of the price of a house, the demand for housing will rise widi a decrease in the interest rate. To test the importance of diese factors as determinants of housing demand, diey were regressed against die square footage of U.S. housing starts. Table 3.2 Summary of Regression Results Dependent Variable: Square Footage of U.S. Residential Construction (Million Square Feet) USFOOT = - 4.0414 + 0.92523E-03*USGNP -i- 4.5691*CUSGNP - 0.11819*USINT (-2.04) (4.78) (2.44) (-3.54) R2 = 0.746 DW = 1.79 DF = 32 Note: Values in parendieses are t-statistics The results of this regression are summarized in Table 3.2 and demonsfrate die importance of aggregate in come, the business cycle and die interest rate to die demand for residential housing. Using annual data for the years 1951 to 1986, and a first-order autocorrelation model, about 75 percent of die variance in die annual square footage of U.S. residential construction was explained by total real U.S. income (USGNP), the business cycle (CUSGNP), and the interest rate (USINT)^^. The business cycle effect is represented by the annual rate of change in real U.S. GNP. Substituting USGNP, CUSGNP and USINT for USFOOT means diat die derived demand for lumber in die U.S. (QDLU) can be specified as follows. QDLU= /(USGNP, CUSGNP, USINT, PL, W, Pp, M) Assuming that lumber demand in Canada is sensitive to die same factors as in the United States, die derived demand for lumber in Canada (QDLC) can be specified as follows. QDLC= /(CANGNP, CCGNP, CANINT, PL. W, Pp, M) For reasons of parsimony, I assumed diat die U.S. and Canadian demand could be reasonably specified as functions of the price of lumber, the level of income and the business cycle. Relative prices of other inputs in resi dential constiuction were not explicidy included so dieir effect will be part of die "error" term of die estimated mod els. Thus, die demand for lumber in die U.S. and Canada is given by the following two expressions: QDLU= /(PL. USGNP, CUSGNP). and QDLC= /(PL. CANGNP. CCGNP). Although the performance of the U.S. lumber industiy is affected by changes in both U.S. and Canadian demand, diese two effects are hard to separate due to their collinearity. For the years 1951 to 1986. there was a 99.6 percent correlation between real U.S. GNP and real Canadian GDP, and a 67.2 percent correlation between the annual rate of change in real U.S. GNP and in real Canadian GDP. Because of diis high level of multicollinearity and be cause of this study's focus on die U.S. lumber industry, I used only U.S. measures to account for changes in de mand. Consequendy, changes in real U.S. GNP and the business cycle in die U.S. account for shifts in both U.S. and Canadian demand. The quantity of lumber supplied is determined by summing the output decisions of individual sawmills. A firm's lumber output is a function of demand conditions and production costs. The cost of manufacturing lumber is a function of labour, capital, logs, odier materials (fuel, electricity, etc.) and the quantity of lumber produced. Thus, die total cost of producing lumber (CL) is given by die following expression. CL = /(W, k,PG, M, QSL). The nominal U.S. prime interest rate was used in this regression. The choice between real and nominal interest rates is not obvious. Using die nominal interest rate is appropriate if new housing purchases are more responsive to nominal than real interest rates. where w is the price of labour, k is the price of capital, PQ is the price of logs, M is the price of other materials and QSL is die quantity of lumber produced. The marginal cost of producing lumber (MCL) is equal to die derivative of die total cost function widi re spect to lumber output and is given by die following expression: The specification of the marginal cost function depends on die form of die total cost function. However, in die ab sence of stiict global constant returns to scale, die marginal cost of producing lumber is a function of the level of production and the prices of die inputs. MCL = /(QSL. W, k,PG,M) Increases in factor prices raise die marginal cost of producing lumber. Thus, die total supply of lumber will fall widi a rise in factor prices. However, because complete factor price data was unavailable, diis study used the average variable cost of production instead to account for aggregate changes in lumber factor prices. For diis rea son, lumber supply will rise (fall) with a fall (rise) in die average variable cost of producing lumber. An industry's international competitiveness rises when its production costs (measured in a common cur rency) fall relative to its foreign competitor's costs. Accordingly, Canadian lumber producers will have improved dieir competitive position relative to U.S. producers if dieir average variable cost of production, measured in U.S. dollars, fall relative to diose of U.S. producers (Roberts 1988 and Adams et al. 1986). Thus, Canada's competitive ness improves relative to tiie U.S. if die following condition holds: [CAVCt * REXCHRTJ [UAVCt] [CAVCt-i * REXCHRTt.i] ^ [UAVCM] ' where CAVCt Js die real average variable cost of production in Canada, REXCHRT is the real Canadian-U.S. ex change rate (in terms of Canadian dollars per U.S. dollar) and UAVCt is the real average variable cost of production in the United States. The real Canadian-U.S. exchange rate is defined as NEXCHRT * (UDEF/CDEF) , where NEXCHRT is die nominal Canada-U.S. exchange rate, UDEF is the U.S. GNP implicit price index and CDEF is die Canadian GDP implicit price index. Thus, Canada's excess supply of lumber to die U.S. should rise (fall) widi a fall (rise) in die U.S. dollar value of Canadian real average variable costs relative to those in the United States. Similarly, U.S. lumber supply should rise (fall) with a fall (rise) in U.S. real average variable costs relative to diose in Canada when expressed in U.S. dollars. In diis study, Canadian real average variable costs were kept in terms of Canadian dollars so that changes in relative competitiveness could be decomposed into changes in average variable costs and changes in the exchange rate. Separating these two effects is important; for example, Roberts (1988) found diat over the period from 1971 to 1987, changes in the relative costs of lumber production in Canada and the U.S. were 2.5 times greater than changes in the real Canada-U.S. exchange rate. Thus, the exchange rate was not the only factor determining die relative competitiveness between die U.S. and Canada. Furthermore, diere is not a clear understanding of die effect that real exchange rate changes have had on die performance of the U.S. lumber industiy. In summary, assuming everything else stays the same, Canada's excess supply of lumber to die U.S. will rise (fall) with a fall (rise) in real average variable costs denominated in Canadian dollars or a rise (fall) in die real Canada-U.S. exchange rate. From die per spective of die U.S. producers, dieir supply of lumber will rise (fall) given a fall (rise) in their real average variable costs or a fall (rise) in the real Canada-U.S. exchange rate. Finally, Canada's excess supply of lumber to the U.S. was assumed to be a function of die price of timber and a "timber subsidy" in British Columbia. The situation in British Columbia was used, versus that in Canada, be cause British Columbia has consistendy supplied about 80 percent of Canada's softwood lumber exports to die United States and because much of the 1986 debate focussed on the price of timber in British Columbia. U.S. lumber producers alleged diat the price of timber in British Columbia was injuring diem because the method used to calculate the price of timber allowed rents to be leaked to British Columbia's lumber producers and because of diis, the price of timber was "too low" relative to U.S. timber prices. Even diough die relationship be tween the price of timber and Canada's excess lumber supply is stiaight forward in that an increase in die price of timber should decrease supply due to a lower timber harvest, die U.S. concern over leaked rent is not. The term "leaked" rent means rent diat flows through to firms. In economic terms, die rent per m^ of log delivered to a sawmill is equal to die residual value after subducting harvesting, transportation and otiier costs (including a normal rate of retiim) from die price of logs at die mUl gate^'. Stumpage fees are the administered In general, economic rent is die payment to die fixed factor of production required to keep it in a particular end-use. Hence, there is a particular rent associated with each end-use. In this study, rent pertains to die use of timber for producing lumber. charge for the right to harvest timber and equal a portion of the total rent per m'^ of log. The portion of the rent diat flows to the users of timber (i.e., sawmills) is said to be leaked. Two aspects of this issue of leaked rent are important widi respect to injury. First, die question of who gets die economic benefit of timber is a distributional question. As long as die rent on die last log harvested is zero, redistiibuting rent from timber owners to timber users does not distort die efficiency of harvest decisions. This is because die rent diat is being redisttibuted is infra marginal. Second, as demonsti^ted by Expression 14, leaked rent is not equal to die timber subsidy, it is only correlated widi die subsidy. The leaked rent in British Columbia was estimated by subtracting the sum of the weighted average B.C. smmpage price and the weighted average variable log ging cost in B.C. from the weighted average B.C. deUvered wood cost^^. Because of the relationship between leaked rent and die timber subsidy, it was assumed diat leaked rent was a reasonable proxy for the timber subsidy. If every thing else is held constant, a rise in the timber subsidy will result in a rise in leaked rents. In summary, Canada's excess supply of lumber to the U.S. should rise with an increase in British Columbia's leaked rents and fall widi an increase in its stumpage price. These demand and supply assumptions mean diat the six indices of injury (INJj) given by Expression 26 can be rewritten as follows. INJi = / (USGNP, CUSGNP, REXCHRT, CAVC, BCSP, BCLRENT, UAVC) , (27) where BCSP is die British Columbian smmpage price and BCLRENT is the level of infra marginal leaked rent in British Columbia. Accordingly, I am assuming that changes in die six measures of injury considered in this study can be traced to changes in factors affecting die excess demand and excess supply of lumber in the United States. Demand shifts are traced to changes in aggregate income and die business cycles in the U.S. and Canada. Supply shifts, are tiBced to adjustinents in the real exchange rate, modifications in U.S. and Canadian production costs, and changes in die level of British Columbia's stumpage prices and leaked rent The next section reports the results of empirically estimating these six reduced form models using data for die years 1951 to 1986. Weighting was done in terms of Coastal and Interior harvest levels. 3.8 Data Description Tables 3.3 and 3.4 list the variable names for die data diat were collected for diis study of die alleged injury to American lumber producers caused by unfair Canadian dmber pricing poUcies and the sources of the data. The Canadian data was compUed on a regional basis and dien aggregated for die estimation of die models specified above. The data used to estimate die reduced form equations for die six measures of injury are Usted in Appendix 3.1 Table 3.3 Usts the regional variables in die Canadian data base. Canadian log producing sectors were di vided into die B.C. Coast (BCC). die B.C. Interior (BCI) and die rest of Canada east of die Rockies (EOR). For each of these three regions, time-series for the annual softwood harvest, the average stumpage price, die average wage and die average variable cost of production were constiiicted for die years 1950 to 1986. Canadian weighted averages were consttucted from regional variables. Aldiough only British Columbia's logging sector data were used in the estimation of the injury models, logging sector data for aU the Canadian regions were coUected. All logging sector data were obtained from Statistics Canada catalogues, except for stumpage prices. British Columbia average softwood stumpage prices were obtained by aggregating up from regional, administrative data obtained from B.C. Ministiy of Forests annual reports. Aver age stumpage prices (weighted by die total harvest volumes) for EOR were derived from total harvest and royalty data obtained from annual forestry reports for Alberta, Ontario, Quebec and New Brunswick. Consequendy, EOR stumpage prices can only be considered as proxies for softwood sawlog prices because the aggregate harvest and roy alty data contain some hardwood and pulpwood values and because the definition of royalties varies across these provinces. Canadian regional logging sector wages for die years 1967 to 1986 were constiTicted from Statistics Canada Catalogue 25-201 by dividing estimates of die total payroU by die total number of hours worked^!. Regional log ging wages for the years 1950 to 1986 were estimated using regression analysis. For die BCC, logging wages were estimated as a function of die average price of softwood sawlogs pur chased by B.C. Coast sawmills and Canadian GNP. For die BCI, logging wages were estimated as a function of the average price of softwood sawlogs purchased by B.C. Interior sawmills and Canadian GNP. FinaUy, EOR logging wages were estimated as a function of die Canadian GNP. The results of diese models are presented in Table 3.5. Logging payroU and hours worked were for production workers only. Canadian Regional Data and Sources Collected for Injury Investigation Variable Name B.C. Coast Delivered Wood Cost (S/m^) B.C. Coast Harvest Level (m^) B.C. Coast Logging Average Variable Cost (S/M^) B.C. Coast Logging Wage ($/hr) B.C. Coast Sawmill Average Variable Cost (S/m^) B.C. Coast Sawmill Wage ($/hr) B.C. Coast Sawmill Production (m^) B.C. Coast Stumpage Price (S/m^) B.C. Interior Delivered Wood Cost (S/m^) B.C. Interior Harvest Level (m^) B.C. Interior Logging Average Variable Cost (S/M^) B.C. Interior Logging Wage ($/hr) B.C. Interior Sawmill Average Variable Cost (S/m^) B.C. Interior Sawmill Wage ($/hr) B.C. Interior Sawmill Production (m^) B.C. Interior Stumpage Price (S/m^) East of the Rockies Delivered Wood Cost (S/m^) East of the Rockies Harvest Level (m^) East of the Rockies Logging Average Variable Cost (S/M^) East of the Rockies Logging Wage ($/hr) East of the Rockies Sawmill Average Variable Cost (S/m^) East of the Rockies Sawmill Wage ($/hr) East of the Rockies Sawmill Production (m^) East of the Rockies Stumpage Price (S/m^) B.C. Forestry Strike Days Canadian GNP Canadian GNP Annual Percentage Change Can. Implicit Price Deflator Canadian Population Can. Share of U.S. Lumber Market (%) Variable Label BCCDWC BCCHARV BCCLGAVC BCCLGWG BCCSMAVC BCCSMWG BCCSMPROD BCCSTUMP BCIDWC BCIHARV BOLGAVC BdLGWG BQSMAVC BQSMWG BQSMPROD BQSTUMP EORDWC EORHARV EORLGAVC EORLGWG EORSMAVC EORSMWG EORSMPROD EORSTUMP BCSTRIKE CANGNP CGNPGR CANDEF CANPOP CANMSHR Source SC 35-204 SC 25-201 SC 25-201 SC 25-201 SC 35-204 SC 35-204 SC 35-204 B.C. MOF SC 35-204 SC 25-201 SC 25-201 SC 25-201 SC 35-204 SC 35-204 SC 35-204 B.C. MOF SC 35-204 SC 25-201 SC 25-201 SC 25-201 SC 35-204 SC 35-204 SC 35-204 Alberta, Ontario, Quebec, & New Brunswick MOF Aimual Reports B.C. FIR IMF-IFS Transformation IMF-IFS IMF-IFS USPS SC refers to Statistics Canada, B.C. MOF refers to B.C. Ministry of Forests Annual Reports, B.C. FIR refers to die B.C. Forest Industrial Relations Limited, IMF-IFS refers to IMF International Financial Statistics and USFS refers to the U.S. Forest Service Publication "U.S. Timber Production, Trade Consumption and Price Statistics 1950-1986". U.S. Data and Sources Collected for Injury Investigation Variable Name Variable Label Source Exchange Rate ($Can/$U.S.) EXCHRT IMF-IFS Real Exchange Rate (Constant 1986 $Can/$U.S.) REXCHRT Transformation U.S. ImpUcit Price Deflator (1986=100) USDEF IMF-IFS Producer Price Index (All Commodities, 1967=100) USPPI USDL U.S. GNP (Constant 1985 $U.S.) USGNP IMF-IFS U.S. Prime Interest Rate (%) USINT Bank of Canada U.S. Population USPOP IMF-IFS U.S. Manufacturing Wage ($U.S./hr) USMANWG USDL U.S. Construction Expenditure (Billion $U.S.) USCONEXP USFSl U.S. New Construction Square Footage (Million ft^) USFOOT USFSl U.S. Housing Starts (Thousands) USHSTART USFSl U.S. Plywood Price Index (1976=100) USPWPI USFSl Forest Products Return on Sales (%) USFPROS MIM U.S. Forest Products Stock Price Index (1986=100) SPFPI S&P Stock Price Index of 400 U.S. Industrials (1986=100) SP400 S&P U.S. North West Saw Log Price ($U.S./MBF) USSLOGPR USFSl U.S. North West Veneer Log Price($U.S./MBF) USVLOGPR USFSl U.S. Douglas Fir Stumpage Price ($U.S./MBF) USDFSTUMP USFSl U.S. Southern Pine Stumpage Price ($U.S./MBF) USSPSTUMP USFSl U.S. Logging and Hauling Cost ($U.S./MBF) USDWC USFSl U.S. Softwood Timber Harvest (Million ft^) USHARV USFSl U.S. Softwood Lumber Price Index (1986=100) USSLPI USFSl U.S. Softwood Lumber Production (Billion BP) USLMPROD USFS2 U.S. Softwood Lumber Capacity (Billion BF) USLMCAP USFS2 U.S. Sawmill Wage ($U.S./hr) USSMWG USFSl U.S. Sawmill Person Hours (Million Hours) USSMHRS USASM U.S. Logging Average Variable Cost (SU.S./ft^) USLGAVC USASM U.S. Sawmill Average Variable Cost ($U.S./MBF) USSMAVC USASM * IMF-IFS refers to IMF International Financial Statistics, USDL refers to U.S. Department of Labour Publication "Producer Prices and price Indexes", USFSl refers to die U.S. Forest Service Publication "U.S. Timber Production, Trade. Consumption and Price Statistics 1950-1986", MIM refers to "Moody's Industiial Manual: American and Foreign", S&P refers to "Standard and Poor's Analyst's Handbook Official Series", USFS2 refers to the U.S. Forest Service Publication "Production Consumption and Prices of Softwood Products in Nordi America: Regional Time Series Data 1950 to 1985", USASM refers to die Bureau of die Census Publication "U.S. Annual Survey of Manufacturers". ** For a discussion of real exchange rates see page 123 in R. I. McKinnon's book entided "Money in International Exchange: The Convertible Currency System". Regression Results: Canadian Regional Logging Wages Per Hour Dependent Variable: Regional Logging Wages per Hour ($C/hr) B.C. Coast Model B.C. Interior Mpdgl East of die Rockies Model Variable Coeff (t-stat) Coeff (t-stat) Coeff (t-stat) Const. 1.8444 4.7206 1.3431 2.4873 0.8955 3.5158 BCCDWC 0.0618 2.1335 • • • • BCIDWC • • 0.1374 1.8876 • • CANGNP 0.0264 8.6457 0.0195 5.2840 0.0273 29.0000 R2 0.982 0.966 0.979 Durbin Watson 1.358 1.071 0.446 F-stat 458.1723 240.5146 837.9507 Observations 20 20 20 Estimates of Canadian regional average variable logging costs for the years 1967 to 1986 were constiiicted by first summing the regional costs for logging sector payrolls, purchased materials, and fuel and electiicity. These estimates of total regional variable costs were then divided by the corresponding regional harvest levels to obtain logging costs per harvested cubic meter. As was the case for logging wages, complete time series for average variable logging costs were unavail able. Consequendy, Canadian regional average variable logging costs for the years 1950 to 1966 were estimated us ing regression analysis. For the BCC, average variable logging cost was estimated as a function of die Canadian im plicit price index, die BCC delivered wood cost and die BCC harvest level. For die BCI, average variable logging cost was estimated as a function of die BCI delivered wood cost and die BCI harvest level. Finally, EOR average variable logging cost was estimated as a function of the EOR delivered wood cost and the EOR harvest level. The results of tiiese models are presented in Table 3.6. Canadian regional delivered wood costs represent die average mill-gate prices (per m^) of softwood sawlogs purchased by sawmills. DeUvered wood cost includes harvesting, transportation, stiimpage and odier logging costs. Time series for delivered wood cost were constructed for each of die diree Canadian regions by dividing the value of softwood sawlogs purchased by sawmills by die corresponding quantity. Regression Results: Canadian Regional Average Variable Logging Cost Per Cubic Meter Dependent Variable: Regional Average Variable Logging Cost Per Cubic Meter ($C/w?) B.C. Coast Model B.Ç, Intgripr Model East of die Rockies Model Variable Coeff (t-stat) Coeff (t-stat) Coeff (t-stat) Const. 6.4898 1.8954 0.7158 0.6495 10.1674 2.3482 CANDEF 0.0499 1.4787 • • • • BCCDWC 0.8654 11.0000 • • • BCIDWC • • 0.6657 8.5691 • • EORDWC • • • • 0.8628 8.9278 BCCHARV -0.000361 -3.2032 • • • • BCIHARV • • 0.0000534 0.9228 • • EORHARV • • • • -0.000141 -1.8205 R2 0.991 0.964 0.940 Durbin Watson 2.259 0.810 1.061 F-stat 593.8035 230.0614 133.780Observations 20 20 20 Canadian regional timber harvests were obtained from Statistics Canada^^; however, a breakdown of the British Columbian harvest into die Coast and Interior regions was only available for years after 1965. For the years 1950 to 1965, die total British Columbian harvest was divided into die Coast and Interior regions by using ad-justed^^ British Columbian regional softwood lumber production as weights. As widi logging sector data, Canadian sawmill data were collected on a regional basis and dien aggregated. Canadian sawmill data originated from Statistics Canada catalogues. Information for die years 1950 to 1986 was available for all variables except die number of sawmill production hours. Regression analysis was used to estimate the total number of sawmill production hours for the three Canadian regions. For the BCC, production hours were estimated as a function of a time frend. Coastal softwood lumber production, die number of production hours in U.S. Regional harvest levels were converted from board feet (log tally) to cubic meters using die following conversion factors: 5.8 fbm/ft^ for die B.C. Coast, 5.5 fbm/ft^ for elsewhere and 2.83168 m^/cunit. The adjustinent to die weights was equal to die 1966 difference between the Coastal share of die B.C. harvest and die Coastal share of B.C.'s softwood lumber production. This adjustinent (6.52 percent) was added to die Coastal share of the total B.C. softwood lumber production and subtiacted from die Interior share. Regression Results: Canadian Regional Sawmill Production Hours Dependent Variable: Canadian Regional Sawmill Production Hours (Thousands of Hours) B.C, Coast Modgl B.C, Interior Model EastQfdieRoçkiggModgl Variable Coeff (t-stat) Coeff (t-stat) Coeff (t-stat) Const. 940223.254 5.0859 590944.338 1.2715 -29758.816 -1.7715 YEAR -481.0726 -5.0957 -298.4443 -1.2702 • • BCCSMPROD 4.9770 15.2576 • • • • BCISMPROD • • 3.4443 6.2766 • • EORSMPROD • • • • 1.0473 2.1333 USSMHRS 4.2215 0.4302 11.7137 0.6080 • • EXCHRT 236.5962 0.0971 -16292.160 -4.4374 • • BCCSMAVC 77.9856 7.4208 • • • • BCISMAVC • • 205.9413 7.0633 • • CANGNPGR • • • • 976.2060 3.7928 REXCHRT • • • • 43877.0888 3.3796 USSLPI • • • • 92.0491 2.3732 R2 0.967 0.956 0.826 Durbin Watson 2.307 1.186 1.150 F-stat 119.0027 87.6034 24.8483 Observations 26 26 26 sawmdls, die nominal exchange rate and die average variable cost of producing lumber on the B.C. Coast. For die BCI, production hours were estimated as a function of a time trend. Interior softwood lumber production, the number of production hours in U.S. sawmills, the nominal exchange rate and the average variable cost of producing lumber in die B.C. Interior. The results of diese models are presented in Table 3.7. Canadian regional sawmill sector hourly wages and average variable manufacturing costs for the years 1950 to 1966 were constinicted in the same manner as for die logging sector. The hourly sawmill wage for production workers was constructed by dividing die annual total payroll by die total number of hours worked^. Average vari able sawmUl costs were constructed by summing the regional costs for sawmill payrolls, purchased materials^^ and Sawmill payroll and hours worked were for production woiicers only. The cost of purchased materials includes stumpage payments. 84 Table 3.8 Representative U.S. Forest Products Companies U.S. Ranking Company Number of Lumber Mills Head-quartos %of'85 Softwood Lumber Production Hardwood Lumber Production MBF Softwood Lumber Production MBF 1 Weyerhaeuser Co. 30 Wash 8.16% 105,000 2,504,000 2 Louisiana-Pacific Corp. 64 Ore 6.42% 26,000 1,971,000 3 Champion International Corp. 30 Conn 5.92% 20,750 1,816,670 4 Georgia-Pacific Corp. 37 Ga 4.86% 192,000 1,492,000 5 Boise Cascade Corp. 17 Idaho 2.52% 29,373 772,619 8 Pope & Talbot Inc.l 5 Ore 2.00% 0 613,575 30 Podatch Ltd. 3 Calif 0.71% 0 217,044 31 Evans 3 B.C. 0.70% 0 216,155 TOTAL 31.29% 373,123 9,603,063 U.S. TOTAL 100.00% 6,474,000 30,690,000 1 Not included in Standard and Poor's Forest Products Stock Mce Index. fuel and electiicity usage. These estimates of die total regional variable cost were then divided by the conesponding regional lumber production to obtain average variable manufactiiring cost per cubic meter of lumber. Regional lumber production for the years 1950 to 1986 were obtained from Statistics Canada Catalogues 35-204 and 35-250. When necessary, lumber production in board feet (lumber tally) was converted to cubic meters using a board foot to cubic meter conversion factor of 2.3597. This factor assumes full-sawn lumber^^ (i.e., the lumber's actual dimensions are equal to its nominal dimensions). Table 3.4 lists die variables in die U.S. data base. Most of the data were obtained from a U.S. Forest Service Publication entitied "U.S. Timber Production, Trade, Consumption and Price Statistics 1950 to 1986". The data from this publication are updated annually from various sources, but chiefly from annual survey data published by the U.S. Departments of Commerce and Labor. Only full-sawn lumber has actual dimensions equal to its nominal dimensions. In Nordi America, actual dimensions do not equal nominal dimensions. For example, a 2 x 4 has actual dimensions of 1.5 inches by 3.5 inches. The conversion factor from Hartman et al (1981) of 2.3597 m^/MBF assumes full-sawn lumber. 85 Table 3.9 Years of Inclusion in the Computation of the Annual Average Return on Sales Company Data Range Weyerhaeuser Co. 1950 to 1986 Louisiana-Pacific Corp. Champion International Corp. Georgia-Pacific Corp. Boise Cascade Corp. 1972 to 1986 1950 to 1986 1950 to 1986 1953 to 1986 Pope & Talbot Inc. Podatch Ltd. 1950 to 1986 1951 to 1986 Evans 1950 to 1982 U.S. average variable costs were constructed using die same method that was used for constructing the Canadian data base. U.S. average variable logging costs for die years 1950 to 1986 were constiiicted by first sum ming the values of logging payrolls and purchased materials, and dien dividing this total by die U.S. softwood tim ber harvest U.S. average variable sawmUl costs for die same time period were constiiicted in a similar manner, ex cept diat the annual estimates of total variable sawmill cost were divided by die total U.S. production of softwood lumber. The annual return on sales of U.S. forest products firms was constiiicted as a weighted average based on firm level net income and gross revenue data obtained from Moody's "Industiial Manual: American and Foreign". Income and revenue data, where available, was collected for the eight U.S. forest products companies listed in Table 3.8. The choice of these eight companies was based on their lumber production and on their inclusion in the Standard and Poor's Forest Products Stock Price Index. The Forest Products Stock Price Index does not include Pope & Talbot; however, this company was included because of its national and regional importance. As can be seen in Table 3.8, the Forest Products Stock Price Index includes die top five U.S. softwood lumber producers and seven of the top 31 producers. Taken togedier, diese eight companies used to estimate die annual return on sales of U.S. for est products firms accounted for 31 percent of softwood lumber production and six percent of hardwood lumber pro duction in 1985. Income and revenue data for the years 1950 to 1986 were available only for four of the eight companies listed in Table 3.8. Consequendy, the number of firms included in the average return on sales varies by year. Table 3.9 lists die years in which each of die eight firms listed in Table 3.8 were included in die computation of die annual return on sales. U.S. average annual logging and hauling costs per thousand board feet (MBF) log tally (i.e., the average U.S. delivered wood cost) represent die average unit cost of harvesting and transporting felled and bucked sawlogs to a saw mUl. The estimate of die U.S. logging and hauling cost for the years 1950 to 1985 was computed from data representing six different U.S. logging regions. A weighted average of these costs was calculated using regional softwood harvests as weights. The U.S. logging and hauUng cost does not include stumpage payments. All prices in Tables 3.2 and 3.3 were converted into real terms. In the case of Canadian prices, the Canadian GNP Implicit Price Deflator was used to convert prices into constant 1986 dollars terms. U.S. prices were converted into real terms using die U.S. GNP Implicit Price Deflator. The Canada-U.S. exchange rate was converted into real terms using die Canadian and U.S. GNP Implicit Price Deflators. 3.9 Estimation and Results The six measures of injury were estimated in an effort to establish die relative importance of changes in Canadian and U.S. economic conditions to the performance of American softwood lumber producers. The ordering of the results in diis section is based on the order of die industiy injury impacts listed in section 771 (C) of die U.S. Tariff Act of 1930. All estimation was done using Version 6.2 of SHAZAM (White et al. 1990). Unless stated otherwise, the data used in this section were for die years 1951 to 1986. Each model was es timated first using ordinary least squares (OLS). In most instances die OLS results indicated die presence of autocor relation among die residuals and an autocorrelation routine based on the standard Cochrane-Orcutt iterative procedure (AUTO) was used to estimate the models. In each of the tables that summarize the results of estimating the six re duced form models, the use of OLS as a column tide means diat the model was estimated using ordinary least squares. The use of AUTO as a column tide means that the model was estimated using an autocorrelation routine. Most of die tables in diis section list results from estimating a model using both OLS and AUTO proce dures. When this is the case, the column tides will have the same number. In Table 3.10, for example, the columns labelled "1 OLS" and "1 AUTO" report die results of estimating die same model using OLS and AUTO procedures. Throughout diis section, the term "full model" means a reduced form model containing all of the hypothe sized explanatory variables. For each of the six measures of injury diere are two full model specifications. That is, there is a full model diat tests for the significance of BCLRENT and one that tests for the significance of BCSTUMP. When discussing die results, the term "rent model" means the model diat tests for the significance of BCLRENT. Similarly, die term "smmpage model" means the model that tests for die significance of BCSTUMP. Unless stated odierwise, significance means statistical significance at die 95 percent level of confidence. Aldiough significance tests of variables assume an underlying random sampling process, random sampling was not part of the process used to develop the reduced form models. That is, variables were removed and, m some cases, new variables were added to improve die explanatory power of a model. Consequendy, die sttength of the statistical inferences based on the magnitude of t-statistics is diminished. An alternate way of viewing die results reported in this section is to consider diem as plausible explanations of what happened in the real world based on an analysis of die best available data. The results summarized in this section are listed in Tables 3.9 to 3.15. For each model, six tests for het-eroscedasticity were carried out. Three asterisks in die tables indicate rejection of die null hypodiesis of homoscedas-ticity at die 0.01 significance level for at least one of die six tests. In these cases and for OLS models. White's het-eroscedasticity-consistent estimates of die variance-covariance matiix were used to calculate t-statistics. In each table, die row labelled RESET shows die results of performing die diree Ramsey Reset specification tests^^. Three asterisks indicate rejection of die null hypodiesis of no misspecification at die 0.01 significance level for at least one of die tests. Because die Ramsey Reset tests are only defined for OLS regressions, diese tests were not available (na) to assess the presence of specification error in die models estimated by the AUTO routine. The Aux R^ stiitistics re ported in diis section represent die R^ value for die regression of die i di independent variable on die remaining inde pendent variables. The Aux R^ statistics is included as a measure of multicollinearity. 3.9.1.0 U.S. Softwood Lumber Price Index Changes in die U.S. real softwood lumber price index (1986=100) was die fu-st measure of injury evaluated in this study. The real price index of United States softwood lumber (USLMPI) is an aggregate index of all soft wood lumber prices. The time series for USLMPI can be divided into diree periods. During the fu-st period from 1950 to 1967, USLMPI trended downward slowly. However, between 1968 and 1979, USLMPI became quite volatile and ti-ended upward at a rate of 3.0 percent per year. From 1980 to 1986, USLMPI trended downward at a brisk 2.0 percent per year. Given the dieoretical development of Expression 27, die USLMPI was assumed to be a function of die var ious shift variables accounting for changes in demand and supply conditions in the U.S. and Canada. Total income and the business cycle were specified as demand shifters. Widi respect to income, USLMPI should rise with a rise in die level of U.S. GNP (USGNP). During die years 1951 to 1986, diere was a 31.2 percent correlation between die USLMPI and USGNP. Lumber demand rises during periods of economic expansion. USLMPI should also rise with These tests involve running duee OLS regressions of a dependent variable on die independent variables and on different powers (i.e., second, third and fourth) of the predicted dependent variable. The RESET tests are F-tests diat test if die coefficients for the powers of die predicted dependent variable are significandy different from zero (White et. al. 1990). a rise in annual rate of change in U.S. GNP (CUSGNP). Over the study period, diere was only a 0.1 percent correla tion between die USLMPI and CUSGNP. Increases in U.S. production costs will cause an increase in the excess demand for lumber in the United States. The real variable cost of producing lumber in the U.S. (RUSAVC), measured in U.S. dollars per diousand board feet, was specified as die aggregate measure accounting for shifts in die U.S. supply of lumber. Consequendy, USLMPI should rise witii a rise in RUSAVC. There was a 80.5 percent correlation between die USLMPI and RUSAVC. Widi respect to die Canadian supply of lumber, increases in Canadian production costs will cause a decrease in the Canadian excess supply of lumber and, thus, a rise in USLMPI. Given the specification in Expression 27, shifts in Canadian production costs are attributed to changes in Canadian average variable costs (RCAVC), changes in the real Canada-U.S. exchange rate (REXCHRT), changes in the level of rent that flows through to British Columbian lumber producers (BCLRENT) and changes in the British Columbian stumpage price (BCSTUMP). Thus, USLMPI should rise with a rise in RCAVC (measured in Canadian dollars per diousand board feet), a fall in REXCHRT (measured in terms of Canadian dollars per U.S. dollar), a fall in BCLRENT (measured in real Canadian dollars per cubic mette) and a rise in BCSTUMP (measured in real Canadian dollars per cubic metre). The correlation between USLMPI and die Canadian supply shifter variables were as follows: RCAVC 77.9 percent, REXCHRT -22.0 percent, BCLRENT -10.2 percent and BCSTUMP 45.1 percent. Finally, it was assumed diat there has been a decline in USLMPI over die study period due to a ttend away from using lumber in its ttaditional end-use markets. Changes in budding codes, consttuction techniques and a rise in die use of substittite products (e.g., plastics, plywood, aluminium) have lowered die demand for lumber in bodi residential and nonresidential consttuction markets. Thus, diere should be a negative ttend in USLPI over time. 3.9.1.1 Model Results for die U.S. Softwood Lumber Price Index Table 3.10 lists die results of die reduced form model of die real (1986=100) U.S. softwood lumber price index. The models in columns one and two of Table 3.10 represent the initial OLS estimation of die full models and test die significance of BCLRENT and BCSTUMP as Canadian supply shifters. The pattern of significance and die signs and magnitudes of die coefficients were encouraging, but the low Durbin-Watson statistics indicated tiiat there was autocorrelation in the error term. Consequentiy, the two full variable models were estimated using the AUTO routine. The choice of the second-order autocorrélation model was based on die significance tests of die first and second-order autocorrelation parametCTS^^. The results of estimating the two full models using AUTO are reported in columns three and four of Table 3.10. The Durbin-Watson statistics indicate diat diere is no reason to reject die assumption of no fu-st-order autocor relation among die residuals. A review of die t-statistics indicates diat RCAVC and TREND do not add significandy to the results. In addition, the collinearity between RUSAVC and RUSAVC suggest diat the parameter estimates of diese two variables may be unstable. Columns five and six in Table 3.10 show die impact of removing RUSAVC and TREND from die rent and stumpage models. The t-statistic for the BCLRENT parameter estimate implies that rents which flowed dirough to British Columbian lumber producers did not significandy affect the U.S. price of lumber. The model in column six is the preferred specification for explaining the changes in USLMPI. The tests for heteroscedasticity and Durbin Watson statistics for diis model imply that the assumption of normally distiibuted and uncorrelated residuals is reasonable. The Aux R^ statistics indicate diat multicollinearity among die regressors has inflated die parameter error estimates. However, widi die possible exception of the parameter for USGNP, die estimation and inference problems due to multicollinearity are acceptable. The results of various Box-Cox transfor mations (e.g., log-log) did not indicate die presence of model misspecification. A review of plots of USLMPI ver sus the explanatory variable did not indicate substantial stiiictural changes in die data. The signs of the coefficients in column six agree with economic theory. Higher lumber prices are associ ated with higher income, die upside of business cycles, a stronger Canadian dollar (relative to the U.S. dollar), higher Canadian average variable lumber production costs and higher British Columbian smmpage fees. Generally, the pre ferred model indicates diat USLMPI is primarily a function of die business cycle and die exchange rate. The relative significance of RUSAVC and the Canadian supply shifters suggests that USLMPI is more sensitive to changes af fecting Canadian supply dian to changes affecting U.S. supply. The insignificance of the ti-end variable means diere was no unexplained downward trend in prices over time. The t-statistics for die fu-st and second order autocorrelation parameters for die model in column three of Table 3.9 were 4.44 and -2.22 respectively. The corresponding t-statistics for the model in column four of Table 3.9 were 5.30 and-1.84. Dependent Variable: U.S. Softwood Lumber Real Price Index (1986=100) Time Series: 1951 to 1986 Estimation: OLS and Second-order Autocorrelation Models lOLS 2 OLS lAUTO 2 AUTO 3 AUTO 4 AUTO USGNP Beta 2.23E-02 2.15E-2 1.31E-02 1.93E-02 1.30E-02 l.llE-02 t-stat 1.63 1.81 0.424 0.64 3.23 2.39 AUXR2 0.994 0.994 0.994 0.994 0.749 0.638 Elasticity 0.522 0.503 0.307 0.451 0.303 0.259 CUSGNP Beta 166.92 161.21 135.29 117.07 151.82 140.06 t-stat 3.12 3.05 2.35 2.34 3.56 3.85 AUXR2 0.281 0.261 0.281 0.261 0.110 0.114 Elasticity 1.536 1.484 1.245 1.077 1.397 1.289 REXCHRT Beta -73.694 -86.178 -71.477 -56.563 -80.580 -59.035 t-stat -2.66 -4.01 -1.98 -1.63 -2.55 -1.95 AUXR2 0.662 0.667 0.662 0.667 0.570 0.584 Elasticity -0.740 -0.866 -0.718 -0.568 -0.809 -0.593 RUSAVC Beta 0.21380 0.20842 6.39E-02 6.02E-02 t-stat 3.18 4.18 1.05 1.03 AUXR2 0.729 0.650 0.729 0.65 Elasticity 0.551 0.537 0.165 0.155 RCAVC Beta 0.19675 0.073977 0.37060 0.20774 0.41991 0.22572 t-stat 2.09 0.99 2.97 1.60 4.14 1.96 AuxR 0.722 0.771 0.722 0.771 0.424 0.549 Elasticity 0.380 0.143 0.716 0.401 0.811 0.436 BCLRENT Beta 0.20418 -0.77556 -0.83438 t-stat 0.31 -1.07 -1.25 AUXR2 0.729 0.729 0.594 Elasticity -0.003 0.010 0.010 BCSTUMP Beta 1.4736 1.8165 1.8479 t-stat 2.81 2.98 3.30 AUXR2 0.451 0.451 0.450 Elasticity 0.079 0.097 0.099 TREND Beta -1.0635 -0.734 -0.10942 -0.67824 t-stat -0.97 -0.82 -0.05 -0.29 AUXR2 0.994 0.993 0.994 0.993 Constant Beta -120.01 -85.010 -79.131 -56.292 -79.859 -55.029 t-stat -1.80 -1.41 -1.26 -1.01 -1.43 -1.06 R2 0.810 0.840 0.855 0.884 0.851 0.880 DF 28 28 28 28 30 30 DW 1.45 1.41 2.09 2.00 2.04 1.98 HHl' *** *** ok ok ok ok RESET ok ok na na na na The results in Table 3.10 indicate that lumber prices are more sensitive to real exchange rate changes than to changes in U.S. or Canadian production costs. This is plausible, especially if one considers die period between 1975 and 1986 when the real value of die Canadian dollar (relative to a U.S. dollar) was ti-ending upward at 2.2 per cent per year, while die ratio of U.S. to Canadian average variable costs was tiending upward at only 0.1 percent per year. Table 3.10 also shows that the Canadian excess supply of lumber is more elastic than the U.S. supply of lumber. This is indicated by die coefficient of Canadian average variable cost which is greater dian die coefficient of U.S. average variable cost (see column four). In turn, this means that a unit increase in the Canadian average vari able cost will cause die price of lumber to rise by more than a unit increase in the U.S. average variable cost. For this to be true, die slope of die Canadian excess supply function must be less than the slope of die U.S. supply function. In summary, die results show that the variation in the U.S. price of lumber can be explained, to a very large extent, by the level of economic activity in lumber's end-use markets. Furthermore, shifts in the factors affect ing the Canadian excess supply of lumber are also significant determinants of the U.S. price of lumber. There is no statistical evidence to support die contention diat rents, which may have flowed through to British Columbian sawmills, were responsible for lower U.S. lumber prices due to a positive shift in die supply of Canadian lumber exported to the U.S. market. 3.9.2.0 U.S. Softwood Lumber Production U.S. softwood lumber production (USLMPROD) was assumed to be a function of the same explanatory variables as U.S. softwood lumber price. Consequendy, on die demand side, USLMPROD should rise widi an in crease in USGNP and CUSGNP. The correlations between USLMPROD and diese two demand shifters were 29.3 percent and 45.9 percent, respectively. On die supply side, USLMPROD should rise with a fall in RUSAVC or a rise in RCAVC. USLMPROD should also increase widi a decrease in REXCHRT, a decrease in BCLRENT or an increase in BCSTUMP. Finally, USLMPROD should decline over die study period due to falling demand in tradi tional end-use lumber markets. The correlation between USLMPROD and RUSAVC was 15.3 percent The correla tions between USLMPROD and the Canadian supply shifters were as follows: RCAVC 25.2 percent, REXCHRT 15.9 percent BCLRENT -2.1 percent and BCSTUMP 30.1 percent. Among die primary reasons for the declining trend in lumber demand are changes in the composition of housing starts, floor space, construction techniques and the use of substitute products. Residential construction ac counts for between 40 and 50 percent of U.S. lumber consumption. From 1950 to 1986, diere was an 18 percent shift in floor area away from single family houses toward multi-family buildings and mobile homes. In 1970, sin gle family homes required about diree times more lumber as multi-family buildings (Rinfret Boston Associates, 1974). Thus, die change in die composition of housing starts away from single famUy dwellings lowered die U.S. demand for lumber. Although the square footage of one and two family houses has tended to increase over die years, the total use of lumber in diese housing starts has fallen. From 1962 to 1982, die total amount of lumber (including hard wood) used in one and two family houses dropped 12.5 percent from 11.2 MBF to 9.8 MBF. Changing construction techniques and a wider use of non-lumber substitutes are responsible for this decline. With respect to changing con stiuction techniques, the increased use of slab construction (building on a concrete floor), a tiend toward the use of masonry versus wood frame construction in exterior walls and préfabrication have lowered consumption and reduced waste on constiuction sites. Furthermore, relaxation of U.S. stiuctural building codes has led to lower lumber uti lization. Lumber utilization in residential constiTiction has also fallen because of the increased use of non-wood and non-lumber substitute products. The greater use of non-wood products in roofs, interior walls, windows and doors and of panel products (e.g., plywood) for sheadiing has caused lumber's traditional end-use markets to shrink. 3.9.2.1 Model Results for U.S. Softwood Lumber Production The estimation results for die reduced form models of USLMPROD are Usted in Table 3.11. The first two columns of Table 3.11 show die OLS results of estimating the full rent and stumpage models. As with lumber price, the signs and significance of die parameter estimates were encouraging, but the low Durbin-Watson statistics indicated diat die presence of autocorrelation. The full models were estimated again using die second-order AUTO routine. As before, die choice of die second-order Cochrane-Orcutt iterative procedure was based on die significance tests of die first and second-order autocorrelation parameters'^. The Durbin-Watson statistics in columns three and four indicate die assumption of no first-order autocorre lation among die residuals from the AUTO models is reasonable. Except for RCAVC, all of the coefficients in columns three and four are significant and have the expected sign. Aldiough we should expect the same variables to explain bodi price and quantity, it appears diat USLMPROD is more sensitive to Canadian supply than USLMPI. The significance and negative sign of die ti-end variable supports die assumption that lumber has been losing ground to substitute products in its end-use markets. The Aux R^ values indicate diat die standard errors of the parameter es timates are inflated. This is especially true for USGNP and TREND which have values of 0.99. Even diough diis The t-statistics for the first and second order autocorrelation parameters for die model in column diree of Table 3.10 were 6.25 and -4.52 respectively. The corresponding t-statistics for die model in column four of Table 3.10 were 5.68 and -2.83. Dependent Variable: U.S. Softwood Lumber Production (Billion Board Feet) Time Series: 1951 to 1986 Estimation: OLS and Second-order Autocorrelation Models lOLS 2 OLS 1AUT0 2AU1U 3 ALTO 4 AUTO USGNP Beta 1.56E-02 1.71E-02 1.37E-02 1.52E-02 1.45E-02 1.40E-02 t-stat 3.77 4.51 2.92 3.01 3.89 3.01 AUXR2 0.994 0.994 0.994 0.994 0.992 0.992 Elasticity 1.406 1.538 1.235 1.368 1.306 1.255 CUSGNP Beta 30.014 25.551 20.561 16.755 19.595 18.757 t-stat 2.66 2.44 2.51 1.98 2.65 2.44 AUXR2 0.281 0.261 0.281 0.201 0.270 0.251 Elasticity 1.062 0.904 0.727 0.593 0.693 0.663 REXCHRT Beta -13.304 -14.572 -15.050 -14.023 -15.352 -14.041 t-stat -2.45 -2.83 -2.96 -2.46 -3.21 -2.48 AUXR2 0.662 0.667 0.662 0.667 0.629 0.619 Elasticity -0.514 -0.563 -0.581 -0.541 -0.593 -0.542 RUSAVC Beta 1.65E-02 7.52E-03 -2.94E-02 -2.66E-02 -2.88E-02 -2.89E-02 t-stat -1.34 -0.74 -3.71 -2.83 -3.77 -3.25 AUXR2 0.729 0.650 0.729 0.650 0.617 0.535 Elasticity -0.163 -0.074 -0.291 -0.264 -0.285 -0.286 RCAVC Beta -1.67E-02 -4.46E-02 5.08E-03 -9.97E-03 t-stat -0.78 -2.03 0.28 -0.48 AUXR 0.722 0.771 0.722 0.771 Elasticity -0.124 -0.331 0.038 -0.074 BCLRENT Beta -0.19994 -0.28067 -0.28094 t-stat -1.43 -2.81 -2.85 AUXR2 0.729 0.729 0.726 Elasticity 0.009 0.013 0.013 BCSTUMP Beta 0.27374 0.25858 0.24196 t-stat 2.45 2.40 2.49 AUXR2 0.451 0.451 0.321 Elasticity 0.056 0.053 0.050 TREND Beta -1.0125 -1.1443 -0.84030 -0.99584 -0.89805 -0.89917 t-stat -3.18 -4.09 -2.35 -2.61 -3.11 -2.55 AUXR2 0.994 -0.725 0.994 0.993 0.992 0.992 Constant Beta -0.99955 5.6944 11.471 14.565 12.736 12.600 t-stat -0.08 0.49 1.27 1.56 1.72 1.59 R2 0.629 0.672 0.789 0.777 0.788 0.776 DF 28 28 28 28 29 29 DW 1.18 1.31 1.77 1.77 1.79 1.75 HbT' ok ok ok RESET* ok ok na na na na degree of collinearity means that parameter estimates are unstable, this is unavoidable when theory and structural characteristics of die market suggest diat diese variables are necessary. Columns five and six in Table 3.11 show die effects of removing RCAVC from tiie rent and stumpage models estimated widi die second-order AUTO routine. The results in diese last two columns indicate that diese are die preferred models for explaining die changes in the level of U.S. softwood lumber production. Of the six het-eroscedasticity tests performed, only the chi-squared goodness of fit test for normality was significant at the 0.01 level of confidence. These test statistics became insignificant, however, if 1986 is removed from die data set. All of die signs of die coefficients in the preferred models agree with economic dieory. Higher U.S. lumber production is associated widi higher income, die upside of die business cycle, a sfronger Canadian dollar, lower U.S. average variable manufacturing costs, a lower level of rent flowing through to British Columbian sawmills and higher British Columbian stiimpage fees. The significance of BCLRENT and BCSTUMP is sensitive, however, to lengdi of the study period^^ A review of die elasticities in Table 3.11 suggests diat most of the change in die level of U.S. lumber pro duction is due to changes in income and die business cycle. The positive elasticity for BCLRENT (which has a neg ative coefficient), reflects the fact that BCLRENT has negative values for some of the years in die time series and, thus, diat die elasticity is unstable widi respect to signal. The average elasticity for BCLRENT is -0.017 if only the 15 years when BCLRENT is positive are considered. For the years 1975 to 1986, die average elasticity for BCLRENT is -0.010. Thus, concerning Canadian supply shifters, the exchange rate effect appears to be die domi nant factor explaining shifts in Canadian lumber supply. That is, die elasticity for REXCHRT is 12 times larger dian die elasticity for BCSTUMP, and at least 30 times larger diat die elasticity for BCLRENT. Finally, aldiough USLMPROD is affected by changes in U.S. average variable costs, die results show that USLMPROD is inelastic widi respect to aggregate changes in factor prices. Broadly speaking, die results reported above are simUar to diose obtained from die analysis of U.S. lumber prices. Generally, most of the changes in the price and and the quantity of lumber produced in the U.S. are explained by factors affecting the demand for lumber in its end-use markets (e.g., residential and non-residential construction). That is, rising income and the general level of economic activity explain most of die annual variation in lumber prices and output levels. Shifts in Canadian excess supply appear to have the next most important role among the determinants of U.S. lumber prices and output levels. The real exchange rate was die most important Canadian sup-For example, during the period from 1960 to 1986, die parameter estimate for BCLRENT was insignificant Negative values mean diat die flow dirough rents to British Columbian sawmills were negative. Given the assumption diat BCLRENT is a proxy for the alleged subsidy, dien negative values mean die alleged subsidy was negative, i.e., that British Columbian sawmills were being taxed. BCLRENT has a negative value in 22 of die 37 years between 1950 and 1986. ply shifter in both the price and output models. Unlike the preferred model of U.S. lumber prices, the coefficient for BCLRENT was significant in the model of U.S. lumber production. A complete explanation of diis asymmetry is unavailable widiout better information. Finally, aldiough die coefficients for BCLRENT and BCSTUMP were sig nificant, the elasticities for these variables are substantially smaller, by an order of magnitude, than die elasticities for the other régresser variables. 3.9.3.0 Canada's Share of die U.S. Softwood Lumber Market Canada's share of die U.S. softwood lumber market (CMSHR) is die next of the six injury measures ad dressed in diis study to be Usted in die U.S. Tariff Act of 1930. Given die specification of Expression 27, CMSHR should rise widi a rise in USGNP and CUSGNP. During the study period, die correlation between CMSHR and USGNP was 98.2 percent. This strong linear correlation mirrors die fact that CMSHR has trended upward at an an nual compound rate of 5.1 percent The correlation between CMSHR and CUSGNP was -10.6 percent CMSHR should also rise with a rise in Canadian competitiveness relative to the United States. Consequendy, CMSHR should rise widi a rise in RUSAVC, a fall in RCAVC, a rise in REXCHRT, a rise in BCLRENT or a fall in BCSTUMP. The correlations between CMSHR and diese supply variables were as follows: RUSAVC 18.4 percent, RCAVC 13.0 percent REXCHRT 65.6 percent, BCLRENT 68.5 percent and BCSTUMP -28.0 percent. A ti-end variable was also included in die specification to test for unexplained linear patterns in die dependent variable. 3.9.3.1 Model Results for Canada's Share of die U.S. Softwood Lumber Market Table 3.12 lists die results of estimating various specifications of die reduced form Canadian market share model. As before, die Durbin-Watson statistics for the full OLS rent and stumpage models (in columns one and two), indicate the presence of serial correlation in die error term, and so die models were reestimated using the AUTO routine. Even diough a review of fu^t and second-order autocorrelation parameters suggested that a fu-st-order model would correct for diis serial correlation, the Durbin-Watson statistics in columns diree and four lie between the upper and lower significance points for die 0.05 level of confidence^'. The t-statistic for die fu-st order autocorrelation parameters for the models in columns diree and four of Table X were 2.59 and 2.69, respectively. Dependent Variable: Canadian Share of U.S. Lumber Consumption (%) Time Series: 1951 to 1986 Estimation: OLS and First-order Autocorrelation Models lOLS 2 OLS 1 AUlU 2 AUTO 3AU1U 4 AUTO 5 ALIO USGNP Beta t-stat AUXR2 Elasticity 7.09E-4 0.20 0.994 0.102 5.19E-04 0.15 0.994 0.075 6.08E-04 0.15 0.994 0.087 3.83E-04 0.09 0.994 0.055 CUSGNP Beta t-stat AUXR2 Elasticity 15.415 1.63 0.281 0.871 15.769 1.69 0.261 0.891 12.711 1.43 0.281 0.718 12.602 1.45 0.261 0.712 USFOOT Beta t-stat AUXR2 Elasticity 1.5578 3.84 0.376 0.199 1.6409 3.91 0.402 0.210 1.5485 3.85 0.196 REXCHRT Beta t-stat AUXR2 Elasticity 11.363 2.50 0.662 0.701 11.175 2.45 0.667 0.689 9.2686 1.76 0.662 0.572 9.0212 1.70 0.667 0.557 14.090 3.59 0.377 0.884 14.919 3.86 0.372 0.936 13.674 3.81 0.844 RUSAVC Beu t-stat AUXR2 Elasticity 2.85E-02 2.77 0.729 0.451 2.74E-02 3.01 0.650 0.433 2.13E-02 2.01 0.729 0.338 2.12E-02 2.06 0.650 0.335 3.08E-02 3.49 0.730 0.495 3.04E-02 3.77 0.657 0.489 2.90E-02 3.55 0.459 RCAVC Beta t-stat AuxR Elasticity -2.36E-02 -1.32 0.722 -0.280 -2.37E-02 -1.21 0.771 -0.281 -1.73E-02 -0.91 0.722 -0.205 -2.07E-02 -0.99 0.771 -0.245 -2.38E-02 -1.66 0.634 -0.287 -1.99E-02 -1.23 0.685 -0.241 -2.35E-02 -1.70 -0.279 BCUŒNT Beta t-stat AUXR2 Elasticity 0.028124 0.24 0.729 -0.002 0.006844 0.06 0.729 -0.001 0.040561 0.43 0.694 -0.003 BCSTUMP Beu t-stat AUXR2 Elasticity 0.008146 0.08 0.451 0.003 0.04122 0.41 0.451 0.014 -0.05887 -0.65 0.396 -0.020 TREND Beta t-stat AUXR2 0.61310 2.30 0.994 0.63683 2.56 0.993 0.63321 2.02 0.994 0.65690 2.17 0.993 0.58757 12.56 0.769 0.58475 14.65 0.600 0.59934 18.41 Constant Beta t-stat -26.745 -2.66 -26.573 -2.57 -21.069 -2.18 -19.988 -2.02 -15.784 -2.80 -17.276 -2.97 -15.119 -3.05 R2 DF DW HET RESET 0.979 28 1.31 ok ok 0.979 28 1.29 ok ok 0.982 28 1.67 ok na 0.982 28 1.66 ok na 0.987 29 1.78 '^** na 0.987 29 1.76 *** na 0.986 30 1.75 *** na The tests for heteroscedasticity indicated that the assumption of normally distiibuted errors could not be re jected. All of die coefficients in columns diree and four have the expected sign; however, the coefficients for USGNP and CUSGNP are insignificant. Multicollinearity among die regressor variables means diat the errors of die parameter estimates are inflated; diis problem is die most severe for USGNP and TREND which bodi have Aux R2 values of 0.99. Because collinearity in the data does not permit the separation of the trend and demand shifter ef fects, USGNP and CUSGNP were dropped from the analysis and replaced by die square footage (in millions of square feet) of new U.S. residential construction. Aldiough residential construction accounted for only 35 percent of U.S. lumber consumption in 1989, die square footage of U.S. residential constiTiction (USFCX)T) is a reasonable proxy for total constiuction activity and therefore of the total lumber demand because construction and remodelling activity in the residential and nonresiden tial sectors (which together accounted for 80 percent of U.S. consumption in 1989) are highly correlated. Columns five and six in Table 3.12 show die impacts of replacing USGNP and CUSGNP by USFOOT as an indicator of changes in die derived demand for lumber. The Durbin-Watson statistics imply diat die assumption of zero autocor relation in the error term cannot be rejected. Four of the heteroscedasticity tests for die model in column five and three of the heteroscedasticity tests for die model in column six were significant at the 0.01 level of confidence. However, all of die heteroscedasticity tests statistics for diese models became insignificant if 1986 was excluded from the time series. Even though 1986 was likely special because of record U.S. consumption and the occurrence of a long International Woodworkers of America strike in British Columbia, removal of this year did not resuh in substantial changes in the parameter estimates or die pattern of significance-^^. Consequendy, an indicator variable for 1986 was not intioduced in the analysis. The t-statistic for die BCLRENT and BCSTUMP coefficients indicate diat neidier die rents diat flowed du-ough to British Columbian lumber producers nor die British Columbian stumpage price had a significant impact on Canada's market share. Column seven shows die results of dropping BCLRENT and BCSTUMP from the analy sis. This last model is the preferred specification for explaining changes in Canada's market share. One of the tests for heteroscedasticity faded at die 0.01 percent level; however, all six tests passed when 1986 was excluded from the analysis. Although the Durbin-Watson test for this model was inconclusive at die 0.01 percent level, the stability of the coefficients and elasticities among die last du-ee models suggests diat die estimation problems due to serial The International Woodworkers of America stiike started on July 23 and lasted four and a half mondis. This stiike was at least twice as long as previous post-war FWA stiikes and it shut down about 40 percent of die province's lumber producing capacity. By die end of 1986, the stiike had lowered lumber production by seven percent and made B.C.'s Coastal region North America's highest cost lumber producer (Paper Tree, September 1986). correlation are not serious. The Aux R' statistics estimation suggest that the estimation and inference problems due to multicollinearity are moderate. The results of various Box-Cox transformations (e.g., log-log) did not indicate the presence of model misspecification. Plots of CMSHR versus the explanatory variable did illusti^te any obvious pat terns diat indicated smictiual changes in die data. The signs of the coefficient in the preferred model agree with economic theory. Higher Canadian market share is associated widi higher constiuction activity, a weaker Canadian dollar (relative to die U.S. dollar), higher U.S. manufacturing costs and lower Canadian manufactiiring costs. The t-statistic for die trend variable is large and positive in part because USGNP (which is collinear widi TREND) was removed from die model. Thus, die trend variable is picking up both the income effect and odier unexplained linear patterns in die dependent variable. The elasticities in column seven indicate diat CMSHR is considerably more sensitive to supply factors dian to demand factors (aldiough some of the sensitivity to die demand side has been lost to die trend variable). As with the other measures of injury discussed above, REXCHRT continues to be the most important determinant of change with respect to both the measure of injury and Canadian supply. That is, CMSHR and Canadian supply are the most sensitive to a unit change in REXCHRT as compared to a unit change in the other explanatory supply vari ables. Over the study period, CMSHR has been more sensitive to changes in RUSAVC than to changes in RCAVC. Thus, everything else being equal, aggregate changes in U.S. factor markets have been more important de terminants of Canada's market share dian aggregate changes in Canadian manufacturing costs. This suggests that there is more production capacity at die margin in the U.S. dian in Canada. In summary, the results show that most of the variation in Canada's market share can be explained by changes in the real exchange rate. Furthermore, it appears that die business environment plays a less important role in determining Canada's market share than it does in determining the price or output level in the U.S. market. Finally, there is no statistical evidence to support die contention that rents which flowed dirough to British Columbian lumber producers significandy affected Canada's share of the U.S. softwood lumber market. 3.9.4.0 U.S. Forest Products Return on Sales The return on sales of U.S. forest products firms (USFPROS) was consdncted by calculating die weighted average return on sales of the forest products firms listed in Table 3.8. Gross revenues were used as die weights in consttucdng the dme series of USFPROS. Anything diat increases lumber revenues relative to die costs of production will increase USFPROS. Thus, USFPROS should rise widi a rise in USGNP and CUSGNP. During die period from 1951 to 1986, die correlations between USFPROS and these two demand variables were -71.5 percent and 27.4 percent, respectively. The strongly negative correlation between USFPROS and USGNP is due to die fact that USFPROS ti-ended downward at a rate of 3.2 percent per year through out the study period. On the supply side, improvements in U.S. manufacturing costs relative to diose in Canada will raise the return on sales. Consequendy, USFPROS should rise with a fall in RUSAVC, a rise in RCAVC, a fall in REXCHRT, a fall in BCLRENT or a rise in BCSTUMP. The correlation between USFPROS and RUSAVC was 29.4 percent. The correlations between USFPROS and die Canadian supply shifters were as follows: RCAVC 34.2 percent, REXCHRT -65.9 percent, BCLRENT -81.5 percent and BCSTUMP 60.6 percent. A trend variable was included in the specification to test for unexplained linear trends in the data and to be consistent widi die previous analyses. 3.9.4.1 Model Results for U.S. Forest Products Return on Sales The outcome of estimating die reduced form model of die U.S. forest products return on sales is summa rized in Table 3.13. The OLS estimation of the full models resulted in unstable parameter estimates. For example, die significance of die RUSAVC and RCAVC coefficients varied dramatically between die rent and stumpage mod els. Also, the Durbin-Watson statistics for diese initial models suggested the residuals were serially correlated thus, the models were reestimated using the AUTO routine. The results in columns three and four are based on a second-order autocorrelation modeP'*. The second-order model was chosen because of die significance of the autocorrelation parameters. The t-statistics for tiie first and second order autocorrelation parameters for die model in column three of Table 3.12 were 5.69 and -3.85 respectively. The corresponding t-statistics for the model in column four of Table 3.12 were 6.55 and -3.64. Dependent Variable: U.S. Softwood Lumber Ir Time Series: 1951 to 1986 Estimation: OLS and Second-order Autocorrela idustry Return on Sales (%) tion Models lOLS 2 OLS 1AUT0 2 AUTO 3 AUTO 4 AUTO 7 AUTO USGNP Beta t-stat AUXR2 Elasticity 4.07E-03 1.50 0.994 1.645 5.96E-03 2.22 0.994 2.41 4.75E-03 1.42 0.994 1.920 4.74E-03 1.31 0.994 1.917 4.71E-03 1.47 0.994 1.905 4.98E-03 1.45 0.994 2.015 5.30E-03 1.57 0.994 2.140 CUSGNP Beu t-stat AUXR2 Elasticity 18.999 2.58 0.281 3.023 13.989 1.89 0.261 2.23 14.374 2.43 0.281 2.287 12.634 2.17 0.261 2.010 14.332 2.48 0.281 2.280 12.95 2.27 0.250 2.060 12.869 2.26 0.247 2.047 REXCHRT Beu t-stat AUXR2 Elasticity -9.8470 -2.77 0.662 -1.710 -10.326 -2.83 0.667 -1.793 -8.0828 -2.19 0.662 -1.404 -5.8272 -1.48 0.667 -1.012 -8.0154 -2.49 0.628 -1.392 -6.3145 -1.90 0.640 -1.097 -7.8533 -2.42 0.628 -1.364 RUSAVC Beta t-stat AUXR2 Elasticity -1.80E-03 -0.22 0.729 -0.080 9.69E-03 1.34 0.650 0.432 -3.27E-04 -0.06 0.729 -0.015 2.03E-03 0.32 0.650 0.090 RCAVC Beta t-stat AUXR Elasticity 2.16E-02 1.55 0.722 0.721 -2.19E-03 -0.14 0.771 -0.073 3.00E-02 2.33 0.722 1.00 2.74E-02 1.90 0.771 0.915 2.98E-02 2.42 0.608 0.994 2.91E-02 2.13 0.695 0.971 3.29E-02 2.57 0.598 1.099 BCLRENT Beta t-stat AUXR2 Elasticity -0.26452 -2.89 0.729 0.056 -0.10671 -1.46 0.729 0.023 -0.10667 -1.50 0.6498 0.023 BCSTUMP Beta t-stat AUXR2 Elasticity 0.20962 2.65 0.451 0.194 0.11681 1.56 0.451 0.108 0.10999 1.58 0.451 0.102 TREND Beta t-stat AUXR2 -0.39052 -1.88 0.994 -0.58190 -2.94 0.993 -0.49617 -1.94 0.994 -0.52053 -1.90 0.993 -0.49373 -2.00 0.994 -0.53648 -2.04 0.993 -0.56581 -2.20 0.993 Constant Beta t-stat -10.002 -1.27 -4.7866 -0.58 -9.0028 -1.38 -9.9086 -1.55 -9.0236 -1.43 -9.7616 -1.59 -8.4184 -1.37 R2 DP DW HET RESET 0.888 28 1.35 ok ok 0.884 28 1.28 ok ok 0.924 28 1.74 *** na 0.924 28 1.74 ok na 0.924 29 1.74 *** na 0.924 29 1.74 *** na 0.920 30 1.66 ok na The Durbin-Watson statistics in columns diree and four of Table 3.13 indicate diat die assumption of zero first-order autocorrelation is reasonable. The pattern of significance is quite different between die OLS and die AUTO results in columns one through four. For example, die AUTO procedure resulted in small and insignificant coefficients for RUSAVC in both die rent and stumpage models. In addition, AUTO lowed die significance of BCLRENT and BCSTUMP and raised die significance of RCAVC. Also, die sign of die RCAVC coefficient is stable in columns diree and four. Multicollinearity among the regressors means that the standard errors of the pa-rameter estimates are inflated. The collinearity between USGNP and TREND is die major source of multicollinear ity. Due to die lack of significance, die USFPROS models were reestimated widiout RUSAVC in die specifica tion. Columns five and six in Table 3.13 show the effects of removing RUSAVC form die second-order rent and stumpage models. The Durbin-Watson statistics for these models continue to indicated that the assumption of zero autocorrelation is reasonable. The chi-squared goodness of fit test for normality of die residuals was significant at the 0.01 level for both the rent and stumpage. The removal of RUSAVC had no effect on the insignificance of BCLRENT and BCSTUMP, so diese variables were also removed from the specification. The results in column seven of show die effects of removing BCLRENT and BCSTUMP from die USFPROS model. The model in column seven of Table 3.13 is die preferred specification for explaining the retiim on sales of die U.S. forest products industiy. The Durbin-Watson statistics and heteroscedasticity tests imply diat the residuals are normally disfributed and uncorrelated. Furthermore, with the exception of the coefficients for USGNP and TREND, the estimation and inference problems due to multicolUnearity are only moderate. Box-Cox ttansforma-tions (e.g., log-log) did not indicate die presence of model misspecification. Also, plots of die dependent versus the independent variables did not indicate die presence of obvious stiiictural changes in die data. As widi die models in columns diree through six, die signs of die coefficients in die preferred model agree with prior expectations. That is, higher return on sales in die U.S. forest products industiy are associated with a higher income, the upside of the business cycle, a sttonger Canadian dollar (relative to the U.S. dollar) and higher Canadian average variable manufacturing costs. The broad pattern of die results indicates that forest products return on sales are primarily a function of demand conditions. When incomes and the business cycle are favourable, the net incomes of die forest industry rise. Canadian supply factors also had a significant effect on die return on sales and, as before, changes in die exchange rate were die dominant reason for shifts in Canadian lumber supply. In summary, the results in Table 3.13 indicate that the large portion of the annual variation in die annual return on sales in die U.S. forest products industiy can be explained by general economic conditions in lumber's tia-ditional end-use markets. Furthermore, as widi die analysis of U.S. lumber prices, Canada's excess supply of lum ber was also a significant determinant of U.S. return on sales. However, there is no statistical evidence to support the contention diat rents which may have flowed through to British Columbian sawmills were a major reason for an upward shift in Canadian lumber supply and, dius, for a lower lumber prices and profits in the U.S. market. 3.9.5.0 Real Standard and Poor's U.S. Forest Products Stock Price Indexes The time horizon for the analysis of RSPFPI is from 1965 to 1986 because die index started in January, 1965. The RSPFPI tracks the real stock prices of the seven large integrated forest products companies listed in Table 3.8. Thus, factors in markets other than just the softwood lumber market affected the index. Lumber and building products divisions are very important sources of revenue, however. As shown in Appendix 3.2, building products divisions accounted for 38 to 79 percent of total 1981 revenues; thus, we should expect to find a close rela tionship between RSPFPI and the level of economic activity in U.S. lumber markets. We should expect to find a close relationship between the performance of the forest products industry and other industries in the U.S. economy. This is because many of the economic factors that affect industiies in general also affect forest products firms. Thus, RSPFPI should be positively related to average movements in die stock market^^. The correlation between RSPFPI and the real Standard and Poor's stock price index for die top 400 U.S. indusûials (RSP400) was 54.2 percent. The relationship between RSPFPI and RSP400 was accounted for in two different ways. First, RSP400 was included as an explanatory variable in the reduced form models of RSPFPI. Second, die average stock market effect was subtiacted from RSPFPI before estimating die reduced form models. Adjusting die real Standard and Poor's stock price index for average movements in die U.S. stock market (RSPAFPI) was accomplished by regressing RSPFPI on RSP400 and dien subti^acting tiie product of RSP400 times its coefficient from RSPFPp^. Consequentiy, changes in RSPAFPI represent die above average or abnormal return to forest products stocks. Removing the average market effect did not substantially alter the pattern of RSPFPI over time. There was an 84.0 percent correlation between RSPFPI and RSPAFPI during die period from 1965 to 1986. The U.S. GNP implicit price deflator was used to convert SPFPI and SP400 into real terms (designated as RSPFPI and RSP400 respectively). We should also expect die RSPFPI to move with die market because diis index was included in Standard and Poor's composite indexes starting in 1970. The OLS results for die regression of RSPFPI on RSP400 are as follows: R2 = 0.294 RSPFPI = 38.411 + 0.90428*RSP400 (1.31) (2.89) Values in parendieses are t-statistics Like the other measures of industry performance discussed so far, RSPFPI and RSPAFPI are functions of the factors diat determine die demand and supply conditions in the U.S. and Canada. On die demand side RSPFPI and RSPAFPI should rise with die aggregate demand for lumber and, dius, witii USGNP and CUSGNP. The corre lations between RSPFPI and these two demand variables were -61.4 percent and 1.5 percent respectively. The corre sponding conelations for RSPAFPI were -28.8 percent and -23.7 percent The negative correlations between depen dant variables and USGNP are a reflection of die downward ttend in RSPm and RSPAFPI which declined at annual rates of 3.03 percent and 3.35 percent respectively. On the supply side, improvements in the relative competitiveness of U.S. forest products industiy should improve RSPFPI and RSPAFPI. Thus, we should expect to see RSPFPI and RSPAFPI rise with a fall in RUSAVC. There was a -36.2 percent correlation between RSPFPI and RUSAVC and a 6.9 percent correlation be tween RSPAFPI and RUSAVC. RSPFPI and RSPAFPI should also rise widi downward shifts in Canadian lumber supply. Consequendy, RSPFPI and RSPAFPI should rise widi a fall in REXCHRT, a rise in RCAVC, a fall in BCLRENT or a rise in BCSTUMP. The correlations between RSPFPI and die Canadian supply variables were as follows: REXCHRT -47.8 percent, RCAVC -2.8 percent, BCLRENT 8.7 percent and BCSTUMP 19.9 percent. Similarly, die correlations between RSPAFPI and die Canadian supply variables were REXCHRT -56.4 percent RCAVC -29.4 percent BCLRENT 18.0 percent and BCSTUMP 8.9 percent 3.9.5.1 Model Results for die Real Standard and Poor's U.S. Forest Products Stock Price Index Tables 3.14 and 3.15 summarize die estimation results for die models of RSPFPI and RSPAPI. The esti mation process stated by performing OLS regressions of die full models and as before, die residuals were serially cor related. The full models were estimated again; however, a third-order correction was required to correct for die auto correlation^^. The t-statistics for die autocorrelation parameters for die RSPFPI model in column diree of Table 3.13 were 5.92,-3.08 and 0.54. The corresponding t-statistics for die model in column four of Table 3.13 were 5.98, -3.24 and 0.81. The t-statistics for die autocorrelation parameters for the RSPAFPI model in column three of Table 3.14 were 5.80,-3.00 and 0.47. The corresponding t-statistics for the model in column four of Table 3.14 were 5.68, -3.00 and 0.46. Dependent Variable: Real Standard & Poor's U.S. Forest Products Stock Price Index: 1986=100 Time Series: 1965 to 1986 Estimation: OLS and Third-order Autocorreladon Models lOLS 2 OLS 1AUT0 2 AUTO 3 AUTO 4 AUTO S AUTO USGNP Beta 2.75E-02 3.43E-02 2.96E-02 2.16E-02 3.47E-02 2.66E-02 2.95E-02 t-stat 1.01 1.11 1.67 1.29 2.05 1.77 1.93 AUXR2 0.847 0.856 0.847 0.856 0.825 0.842 0.811 Elasticity 0.723 0.902 0.778 0.568 0.913 0.700 0.776 CUSGNP Beta -201.70 -175.98 113.25 147.76 t-stat -0.76 -0.55 0.96 1.27 AUXR2 0.305 0.417 0.305 0.417 Elasticity -1.722 -1.502 0.967 1.261 REXCHRT Beta -252.01 -289.86 -362.46 -305.92 -385.15 -399.67 -351.82 t-stat -2.03 -1.87 -3.89 -3.41 -4.30 -4.23 -4.43 AUXR2 0.713 0.780 0.713 0.780 0.707 0.732 0.683 Elasticity -2.408 -2.769 -3.463 -2.922 -3.679 -3.245 -3.361 RUSAVC Beta -0.47253 -0.70251 -0.31554 -0.23134 -0.30114 -0.23883 -0.25616 t-stat -1.11 1.60 -2.49 -1.76 -2.30 -1.94 -2.15 AUXR2 0.911 0.900 0.911 0.900 0.814 0.857 0.670 Elasticity -1.139 -1.694 -0.761 -0.558 -0.726 -0.576 -0.618 RCAVC Beta 1.2987 1.2734 0.22992 0.032083 t-stat 1.86 1.14 0.70 0.08 AuxR 0.837 0.923 0.837 0.923 Elasticity 2.3827 2.336 0.422 0.059 BCLRENT Beta 5.8201 -1.3710 -1.4299 t-stat 1.67 -0.84 -0.87 AUXR2 0.471 0.471 0.454 Elasticity 0.025 -0.006 -0.006 BCSTUMP Beta -0.89644 0.89625 1.0639 t-stat -0.21 0.63 0.92 AUXR2 0.818 0.818 0.110 Elasticity -0.042 0.042 0.050 RSP400 Beta 1.3676 1.0737 0.83666 0.70398 0.93050 0.8881 0.97491 t-stat 2.34 1.59 2.55 1.85 3.84 3.53 3.96 AUXR2 0.820 0.838 0.820 0.838 0.780 0.806 0.735 Elasticity 1.031 0.809 0.631 0.531 0.701 0.670 0.735 Constant Beta 254.15 256.97 293.58 243.38 458.86 411.14 419.39 t-stat 0.76 0.69 1.81 1.56 4.00 3.87 3.94 R2 0.689 0.628 0.915 0.913 0.905 0.906 0.901 DF 14 14 14 14 16 16 17 DW 1.28 0.96 1.73 1.74 1.78 1.76 1.80 HET ok ok ok ok ok ok ok RESET ok ok na na na na na The Durbin-Watson tests for the AUTO models indicated that the estimation problems due to autocorrela tion were no longer serious. All of the coefficients in Tables 3.14 and 3.15 have the expected sign. Furthermore, die pattern of significance is die same in bodi tables. That is, REXCHRT and RUSAVC are die most significant Dependent Variable: Real Standard & Poor's U.S. Adjusted Forest Products Stock Price Index: 1986=100 Time Series: 1965 to 1986 Estimation: OLS and Third-order Autocorrelation Models lOLS 2 OLS 1AUTO 2 AUTO 3 AUTO 4 AUTO 5 AUTO USGNP Beta 0.19511 0.3308 0.32639 0.26889 0.35481 0.28352 0.28091 t-stat 0.76 1.12 2.16 1.98 2.27 2.18 2.14 AUXR2 0.816 0.843 0.816 0.843 0.794 0.833 0.720 Elasticity 1.541 2.646 2.578 2.124 2.802 2.239 2.218 CUSGNP Beta -1179.6 -1691.7 997.76 1016.1 t-stat -0.47 -0.54 1.10 1.08 AUXR2 0.154 0.399 0.154 0.399 Elasticity -3.026 -4.340 2.560 2.607 REXCHRT Beta -2416.0 -3027.5 -3811.3 -3405.3 -4025.1 -3555.0 -3654.5 t-stat -1.93 -1.93 -4.12 -4.02 -4.43 -4.39 -4.54 AUXR2 0.699 0.780 0.699 0.780 0.698 0.732 0.646 Elasticity -6.94 -8.693 -10.943 -9.778 -11.56 -10.207 -10.493 RUSAVC Beta -6.4447 -7.7077 -3.3156 -2.7120 -3.1676 -2.4907 -2.6859 t-stat -1.63 -1.83 -2.58 -2.09 -2.40 -2.01 -2.20 AUXR2 0.890 0.888 0.890 0.888 0.733 0.769 0.606 Elasticity -4.670 -5.586 -2.403 -1.965 -2.295 -1.805 -1.946 RCAVC Beta 14.185 12.604 2.7026 1.8704 t-stat 1.98 1.15 0.88 0.56 AUXR 0.835 0.915 0.835 0.915 Elasticity 4.782 6.950 1.490 1.031 BCLRENT Beu 50.148 -14.110 -15.339 t-stat 1.50 -0.85 -0.92 AUXR2 0.388 0.388 0.341 Elasticity 0.064 -0.018 -0.020 BCSTUMP Beta -4.1251 5.3395 10.883 t-stat -0.11 0.42 0.97 AUXR2 0.766 0.766 0.452 Elasticity -0.059 0.076 0.155 Constant Beta 2490.7 4045.4 3108.0 2776.2 4852.4 4270.9 4515.3 t-stat 0.72 0.79 1.90 1.72 4.64 4.26 4.65 R2 0.540 0.471 0.880 0.875 0.866 0.867 0.859 DF 15 15 15 15 17 17 18 DW 1.27 0.98 1.74 1.75 1.78 1.76 1.79 HET ok ok ok ok ok ok ok RESET ok ok na na na na na variables followed by USGNP. Because die coefficients for CUSGNP and RCAVC were insignificant in bodi of die RSPFPI and RSPAFPI models, diey were removed from die analysis^^. Columns five and six in Tables 3.14 and 3.15 show die impact of dropping CUSGNP and RCAVC. In bodi tables, die significance levels of USGNP have risen while diose for RUSAVC have stayed about die same. 39 CUSGNP was removed first but RCAVC remained insignificant so RCAVC was also dropped from the analysis. Also, REXCHRT is still the most significant factor explaining stock prices and the adjusted stock prices. BCLRENT and BCSTUMP were also dropped from die analysis because of dieir continued insignificance. The results in column seven of Tables 3.14 and 3.15 indicate that these are the preferred models for explain ing RSPFPI and RSPAFPl. As widi die odier models in diese tables, the tests for heteroscedasticity do not lead to the rejection of assumption diat die residuals were normally distiibuted. Also, the Durbin-Watson tests suggested that die assumption of zero first-order autocorrelation was reasonable. The Aux R^ values indicate diat die errors of the coefficients are inflated but die multicollinearity problems are not that severe. Finally, die signs of the coeffi cients agree with economic theory. Higher stock prices (bodi adjusted and unadjusted) are associated with higher USGNP, a lower exchange rate and lower RUSAVC. In die case of unadjusted stock prices, RSPFPI also rises widi a rise in RSP400. Generally, the results presented in this section indicate that most of the change in the U.S. forest products stock price is due to changes in relative competitiveness as a result of fluctuations in the Canada-U.S. exchange rate and in U.S. average variable costs. Total income has also had a relatively significant impact on the long term prof itability of the U.S. forest products industiy. This is because higher aggregate income translates into higher con struction activity. Thus, for a given production cost, higher income leads to higher lumber prices and profits. Beyond diese determinants, it appears that factors affecting die stock market in general also affect the average stock price of U.S. forest products firms. The insignificance of CUSGNP was not that surprising give the relationship between the average index of the stock market and die business cycle. Finally, the results do not support the allegation that rents captured by British Columbian sawmills caused a significant upward shift in Canada's lumber supply. Thus, there is no signifi cant link between rents captured by British Columbian sawmills and U.S. forest products stock prices. 3.9.6.0 U.S. Sawmill Production Hours U.S. sawmill production hours (millions of hours) is die last of the six measures addressed in this study diat is listed in die U.S. Tariff Act of 1930^0. The demand for labour by U.S. sawmills (USSMHRS) depends on USSMHRS represents the total number of production man-hours in all types of U.S. sawmdls. Thus, USSMHRS is die sum of production hours in the following types of mills: softwood lumber, hardwood the level of pnxluction which, as discussed above, is a function of the demand and supply conditions in the U.S. and Canada. Consequendy, we should expect to see a rise in USSMHRS widi a rise in demand and, dius, widi a rise in USGNP or CUSGNP. Over die study period, die correlations between USSMHRS and die demand shifters were 83.6 percent and 22.9 percent respectively. The negative correlation between USSMHRS and USGNP is due to the fact that USSMHRS frended downward at an annual rate of 3.2 percent. USSMHRS should also rise with factors that improve the competitiveness of U.S. sawmdls relative to Canadian sawmUls. Thus, USSMHRS should rise widi a fall in RUSAVC, a rise in RCAVC, a fall in REXCHRT, a fall in BCLRENT or a rise in BCSTUMP. The correlation between USSMHRS and RUSAVC was 21.1 percent. The correlations between USSMHRS and die Canadian supply shifters were as follows: RCAVC 18.1 percent, REXCHRT -57.2 percent, BCLIŒNT -82.1 per cent, and BCSTUMP 38.2 percent. As mentioned above, USSMHRS ti-ended downward over die stiidy period; however, die decline in die de mand for labour was not smoodi. For example, from 1951 to 1960, USSMHRS declined rapidly at a rate of 5.8 percent. From 1961 to 1986, USSMHRS declined much more slowly at a rate of only 1.7 percent per year. Much of the decline in the demand for labour in U.S. sawmills was due to a process of rationalization where the production from smaller mills was fransferred to much larger mills. The employment effects of this redistiibu-tion of U.S. production to more mechanized mills were dramatic (Lea 1962). Tables 3.15 and 3.16 show how ratio nalization has increased both die average output per mill and labour productivity over die study period. Using data for Washington and Oregon States, Table 3.16 illustiates die rapid drop in die number of sawmills and the related in crease in die ou^iut per mUl for each of die U.S. census years between 1947 and 1987. Table 3.17 shows how the average productivity of labour has increased diroughout die period from 1950 to 1985. Tables 3.16 and 3.17 also in dicate diat mills closures, output per mill and labour productivity rose during die 1950s at nearly twice die rate as during die 1960s, 1970s and early 1980s. It appears diat shifting production to fewer but larger sawmills (in con-lumber, wood chip, softwood cut stock, softwood flooring and custom saw mUls. In 1982, softwood lumber mills accounted for 55.9 percent of all U.S. sawmdls and planing mUls. Redistribution of U.S. Lumber Production: 1947 to 1987 Total Production (BBF) Oregon Wash. Number of Establishments Oregon Wash. Avg. Output per Establishment (MMBF) Oregon Wash. 1947 1954 1958 1963 1967 1972 1977 1982 1987 7.10 8.85 7.54 7.76 6.97 7.94 7.51 4.68 8.85 3.71 3.03 3.45 3.67 3.28 3.75 4.03 3.06 4.65 1,466 1,201 645 520 396 255 332 287 274 808 552 469 389 306 340 255 223 241 4.8 7.4 11.1 14.9 17.6 31.1 22.6 16.3 32.3 4.6 5.4 7.4 9.4 10.7 11.0 15.8 13.7 19.3 junction widi odier sources of technological improvement^ 1) was responsible for at least some of the decline in die use of labour in U.S. sawmills. Constantino and Haley (1989) also found diat productivity in U.S. Pacific Northwest sawmills grew faster during the late 1950s and 1960s as compared to the 1970s and early 1980s. In particular, they found that during the period from 1957 to 1971, total factor productivity in U.S. Pacific Northwest sawmills grew at an annual rate of 1.1 percent but, during die period from 1972 to 1982, die annual rate growth in total factor productivity had fallen to 0.2 percent An indicator variable and a ttend variable were used to account for the two different rates of technological change. An indicator variable (D5160) was used to divide a ttend variable (TREND) into a period from 1951 to 1960 and a period from 1961 to 1986. D5160 was given a value of one for the years 1951 to 1960 and a value of zero otherwise. An interaction variable equal to the product of D5160 time TREND (DTREND) was also created. In summary, D5160, TREND and DTREND were added to die reduced form model of USSMHRS to test for two differ ent rates of technological change in U.S. sawmills. For example, chipping headrigs were introduced in die mid 1960s and computers were incorporated into various stages of the lumber production process throughout the 1980s. U.S. Sawmill Production and Employment Production Year Employment (Thousands) Production (BBF) per Employee (MBF) 1950 431.0 30.6 71.0 1955 315.1 29.8 94.6 1960 220.4 26.7 121.1 1965 181.5 29.3 161.4 1970 155.1 27.5 177.3 1975 142.3 26.7 187.6 1980 151.8 28.2 185.8 1985 119.9 30.5 254.4 3.9.6.1 Model Results for U.S. Sawmill Production Hours Table 3.18 summarizes die OLS results firom estimating die full USSMHRS models. Because at least one of the heteroscedasticity tests were significant at the 0.01 percent level, White's heteroscedasticity-consistent esti mates of the variance-covariance matiix were used to calculate die t-statistics. The Durbin-Watson statistics in columns one and two indicate diat die assumption of zero autocorrelation cannot be rejected. All of die coefficients from the initial estimation have the correct signs except for REXCHRT, RUSAVC and RCAVC. Because the coef ficient for die Canada-U.S. exchange rate was insignificant, it was dropped from the analysis. Columns diree and four show die impact of removing REXCHRT from die rent and stiimpage models. The assumption of zero autocorrelation among die residuals appears to be reasonable. The pattern of significance in these last two models is similar to die pattern of significance in die first two models, except diat removing REXCHRT has made die coefficient of USGNP insignificant The coefficient of RUSAVC and RCAVC continue to have die wrong sign. In addition, die coefficients of BCLRENT and BCSTUMP remain insignificant. The Aux R^ values indicate die presence of a high degree of multicollinearity among the regressor variable, most of which is due to the correlation among die ttend and USGNP variables. The chi-squared tests for normally distiibuted residuals were sig nificant at die 0.01 percent level and so White's heteroscedasticity-consistent estimates of die variance-covariance ma-bix were used to calculate t-statistics. One of die Ramsey Reset tests for specification error was also significant at Dependent Variable: U.S. Sawmdl Production Hours (Millions of Hours) Time Series: 1951 to 1986 Estimation: OLS Models lOLS 2 OLS 3 OLS 4 OLS 5 OLS USGNP Beu 9.53E-02 6.23E-02 0.10358 6.25E-02 0.11668 t-stat 2.60 1.59 1.85 0.99 2.17 AUXR2 0.996 0.997 0.996 0.997 0.995 ElasUcity 0.663 0.433 0.721 0.435 0.812 CUSGNP Beta 258.28 243.75 259.78 243.75 249.66 t-stat 2.73 2.11 1.86 1.80 1.80 AUXR2 0.344 0.339 0.343 0.339 0.339 ElasUcity 0.707 0.667 0.711 0.667 0.683 REXCHRT Beta 27.097 1.5033 t-stat 0.54 0.03 AUXR2 0.685 0.717 Elasticity 0.081 0.004 RUSAVC Beu 0.64368 0.66458 0.63135 0.66406 0.66793 t-sut 5.09 6.51 3.90 4.37 4.29 AUXR2 0.789 0.771 0.781 0.766 0.766 Elasticity 0.493 0.509 0.483 0.509 0.511 RCAVC Beu -0.36323 -0.70051 -0.41304 -0.70495 -0.40010 t-stat -1.56 -2.56 -1.39 -2.01 -1.35 AraR 0.830 0.898 0.801 0.865 0.801 Elasticity -0.208 -0.402 -0.237 -0.405 -0.230 BCLRENT Beu -1.3860 -1.5293 t-stat -0.75 -0.88 AraR2 0.765 0.756 Elasticity 0.005 0.006 BCSTUMP Beu 2.9855 3.0026 t-siat 1.61 1.54 AUXR2 0.768 0.732 Elasticity 0.048 0.048 D5160 Beu 205.33 246.17 206.07 246.45 204.32 t-stat 10.04 8.57 7.29 6.36 7.28 AUXR2 0.903 0.974 0.994 0.972 0.944 TREND Beu -14.464 -11.169 -14.933 -11.173 -16.283 t-sut -4.72 -3.31 -3.12 -2.02 -3.61 AUXR2 0.997 0.997 0.996 0.997 0.996 DTOEND Beu -18.887 -23.263 -19.150 -23.301 -18.465 t-sut -6.40 -6.49 -5.88 -5.30 -5.86 AUXR2 0.903 0.955 0.899 0.948 0.893 Constant Beu -39.795 75.985 -9.5221 78.171 -20.314 t-stat -0.26 0.51 -0.07 0.51 -0.14 R2 0.986 0.986 0.986 0.986 0.985 EF 26 26 27 27 28 m 1.87 1.81 1.86 1.81 1.87 HET • •* »** *** *** *** RESET *** *** *** *** ok the 0.01 percent level, indicating diat the models may be misspecified. Because of die insignificance, die of BCLRENT and BCSTUMP, die USSMHRS model was reestimated widiout BCLRENT and BCSTUMP in die specification. Column five in Table 3.18 shows die impact of removing BCLRENT and BCSTUMP from die USSMHRS model. The chi-squared goodness of fit test failed at die 0.01 percent level. The Ramsey reset tests statistics were insignificant at the 0.01 percent level. In addition, the results of Box-Cox transformations and a re view of residual plots indicated diat die linear functional form was reasonable. The signs of tiie coefficients for RUSAVC and RCAVC are wrong widi respect to economic theory. The significant and positive coefficient for RUSAVC is likely tiie result of spurious correlation caused by die general rise in labour productivity and corresponding decline in manufacturing costs in U.S. sawmills over much of the study pe riod. The sign on die RCAVC coefficient is negative because Canadian manufacturing costs rose less quickly and fell more rapidly relative to U.S. manufacturing costs over much of die study period. The coefficients of die remaining variables are significant and have die expected signs. Thus, USSMHRS rises with a rise in USGNP or die upside of the business cycle. Furthermore, the indicator and trend variables indi cate that the processes of rationalization and technological change explain a large proportion of the downward ttend in the demand for labour in U.S. sawmdls. Also, as hypodiesized, the results show that die demand for labour de creased more quickly during die period from 1951 to 1960 dian during die period fi-om 1961 to 1986. The OLS model in column five is the preferred model for explaining the variation in the demand for labour by U.S. sawmills despite die insignificance of RCAVC^^ Dropping RCAVC from die model estimated witii OLS resulted in significant Ramsey Reset specification error tests and Durbin-Watson statistics. Reestimation using tiie second-order Auto routine did not have an appreciable effect on die magnitude of the coefficients or on the pattern of significance. Thus, the model in column five is die preferred model explaining die variation in USSMHRS. In general, die results in Table 3.18 indicate that most of the variation in the use of labour in U.S. sawmills is due to ttends attiibutable to die process of rationalization and technological change. Factors affecting die U.S. demand for lumber also explain a significant proportion of the fluctuation in the number of production hours. As expected, when demand is high, die hours of production rise, conversely, when demand is low, die hours of pro duction fall. The spurious correlations which caused die wrong signs on the coefficients of RUSAVC and RCAVC are possibly die result of a substitution of capital for labour due to rising material costs. Rising harvesting and log ttansportation costs in the U.S., especially during die period from 1968 to 1980, pushed RUSAVC up more quickly The coefficient for RCAVC in column 5 of Table 3.17 is significant at die 0.10 level. than RCAVC*^. If the use of logs in the production process is quasi-fixed, dien the easiest way to lower costs is to control die labour and non-wood input costs. However, since it is widely believed diat wages can only be lowered by as a result of prodacted negodations, it seem plausible diat U.S. sawmill might have substituted away firom labour in die production process. Finally, Table 3.18 shows that there is no statistical evidence to support the contention rents which flowed dirough to British Columbian sawmills caused a significant increase in Canada's lumber supply which in turn, caused a decrease in die demand for labour in U.S. sawmills. In fact the results tend to suggest diat die factors affect ing die Canadian supply of lumber had very litde to do with the decline in U.S. sawmill production hours. During die period from 1968 to 1980, diere was a 88.3 percent correlation between die weighted average of real U.S. regional harvesting and log transportation costs and RUSAVC. Also over die same time period, the ratio of RUSAVC to RCAVC trended upward at an annual rate of 2.1 percent, dius U.S. manufacturing costs rose more quickly diat Canadian manufacturing costs. 3.10 Economie Significance of Injury Determinants In this section, the preferred models reviewed above are used to study die response of the injury measures to hypothetical changes in four key variables. The period from 1975 to 1986 was picked for die analysis because Canada's presence in die U.S. market was relatively low in 1975. Also, 1975 marked die start of a period in which die U.S. dollar appreciated in value against die Canadian dollar. Four counterfactual scenarios were evaluated. First, it was assumed diat the real Canada-U.S. exchange rate stayed at its 1975 value of $C 1.0490. Second, U.S. GNP was held constant at its 1975 value of $U.S. 2,990 bil lion. Third, it was assumed that the average real B.C. stumpage price was raised to SCIO.OO per cubic mette. Fourth, value of BCLRENT was held at zero. The results of these scenarios are listed in Table 3.19. The counterfactual values for 1986 were obtained by adjusting die predicted 1986 values by die product of the counterfactual change in a variable times the variable's coef ficient For example, the counterfactual predicted value for die U.S. lumber price index in 1986, assuming diat die exchange rate remained at its 1975 value, was obtained as follows. First, the exchange rate coefficient from die pre ferred model in Table 3.10 was obtained. Next, die difference between die 1986 and 1975 values of REXCHRT (Appendix 3.1) was calculated. Finally, die product of the coefficient times tiie change in REXCHRT was added to the predicted 1986 value of die U.S. lumber price index (Table 3.19) obtained from the preferred model and actual 1986 data44. A review of Table 3.19 shows diat five injury measures were affected by die exchange rate scenario. Had the U.S. dollar not appreciated against the Canadian dollar, the results indicate that the U.S. lumber price and U.S. lumber production would be higher by 17.3 percent and 4.50 billion board feet, respectively. Furthermore, the rise in tiie exchange rate from its 1975 value lowered die competitiveness of U.S. sawmills and allowed Canada's market share to rise about 4.0 percent by 1986. Finally, if die exchange rate had not risen above its 1975 level, the average The 1986 predicted value for USLMPI in Table 3.18 is 102.78. The coefficient for REXCHRT in column six of Table 3.9 is -59.035. The 1975 and 1986 values of REXCHRT obtained from Appendix 1 are 1.049 and 1.342 respectively. The counter factual 1986 value of USLMPI assuming a 1975 exchange rate was obtained as follows: 1986 counter factual value = 102.78 - 59.035(1.049-1.342) = 120.08 Table 3.19 Economie Importance of Variables Counter-Affected Actual Predicted factual Change Change Simulation Injury 1986 1986 1986 From From Description Measure Values Values Values Actual Predicted Exchange USLMPI 100.00 102.78 120.08 20.08 17.30 Rate at USLMPROD 34.20 31.44 36.48 2.28 4.50 1975 CANMSHR 30.35 32.98 28.97 -1.38 -4.01 Value USFPROS 4.59 2.71 5.01 0.42 2.30 RSPFPI 100.00 102.82 205.90 105.90 103.08 U.S. GNP USLMPI 100.00 102.78 90.15 -9.85 -12.63 at 1975 USLMPROD 34.20 31.44 15.44 -18.76 -16.00 Value USFPROS 4.59 2.71 -3.32 -7.91 -6.03 RSPFPI 100.00 102.82 69.20 -30.80 -33.62 USSMHRS 247.70 227.79 94.87 -152.83 -132.92 B.C. Stumpage USLMPI 100.00 102.78 115.52 15.52 12.74 at $10.00 USLMPROD 34.2 31.44 33.11 -1.09 1.67 Zero B.C. Rent USLMPROD 34.20 31.98 32.17 -2.03 0.19 return on sales in U.S. forest products industry would have been 2.3 percent higher and the U.S. forest products stock price index would have been 103.1 percent higher. The second scenario assumed diat U.S. GNP did not rise from its 1975 value of 2,990.0 billion U.S. dol lars. The 38.1 percent increase in real U.S. GNP has had a strong and positive impact on the U.S. lumber industry. Table 3.19 indicate diat the sharp rise in U.S. GNP since 1975 has increased bodi U.S. lumber prices and production by 12.6 percent and 16.00 bUUon board feet respectively. Also, coincident with higher lumber demand and prices, the expanding economy caused the return on sales and the stock price index in the U.S. forest products industry to rise by 6.03 percent and 33.6 percent respectively. Finally, the rise in U.S. GNP translates into 132.92 million more production hours by 1986. At an average of 1.95 diousand hours per worker per year, die rise in U.S. GNP between 1975 and 1986 is equivalent to about 68 diousand production jobs. The fourth scenario assumes diat die real (constant 1986 dollars) British Columbian stumpage price in 1986 was SCIO.OO per cubic metie. This represents a 112 percent increase in die average British Columbian stiimpage price for die period from 1975 to 1986. The results indicate diat this hypothetical stumpage prices increase would reduce Canada's excess supply of lumber, causing an increase in die U.S. price of lumber of 12.7 percent. However, die hypodiesized rise in British Columbian stumpage prices only causes U.S. production to rise by 3.5 percent or 1.67 billion board feeL In the last scenario analyzed, the amount of rent which flowed dirough to Bridsh Columbian sawmills was set to zero. The actual value of BCLRENT in 1975 was -0.32 dollars per cubic mede so diis scenario is similar to assuming that BCLRENT was kept at its 1975 value. The results show diat die value of rent collected by British Columbian sawmills caused U.S. lumber production to fall by 0.19 billion board feet or 0.6 percent. This is a small impact relative to tiiose caused by changes in die exchange rate and the level of U.S. GNP. In summary, die results of die four simulations highlight the sensitivity of die injury measures to general economic factors tiiat relate to level of lumber demand in die United States. Furthermore, the results indicate that the exchange rate has played a very important role in eroding the long-run competitiveness of the U.S. softwood lumber industiy. 3.11 Discussion and Conclusions The reduced form models for die six measures of injury were all estimated using Version 6.2 of SHAZAM (White et al. 1990). The approach to estimating die models started widi the development of an economic framework that recognized die relationships between die U.S. and Canadian lumber markets and included a proxy for the alleged timber subsidy received by British Columbian sawmills. This lead to Expression 26, a general reduced form model of the six injury measures. Next, the main factors affecting the supply and demand in the U.S. and Canada were specified. This resulted in Expression 27 which linked changes in the six injury measures to changes in factors af fecting tiie excess demand and supply of lumber in die United States. Demand shifts were ttaced to changes in U.S. GNP and the business cycle. Supply shifts were ti^aced to changes in die real Canada-U.S. exchange rate, modifica tions in U.S. and Canadian manufacturing costs and fluctuations in the level of infra marginal rents captured by British Columbian sawmills and British Columbian stumpage prices. The six reduced form models were dien empir ically estimated using data for die years 1951 to 1986. Initially, die parameters for the six models were estimated using ordinary least squares, but serial correlation among die residuals lead to die choice of an autocorrelation routine. The analysis of residuals for heteroscedasticity and model misspecification indicated that linear functional forms were reasonable. Regression analysis based on Box-Cox transformations of tiie dependant and independent variables did not suggest that another functional form (e.g., log-log or semi log) would have been more appropriate. The estimation and inference problems due to multi collinearity among die independent variables were not unusual for empirical stiidies of diis type. However, some of the standard errors of die parameter estimates were inflated by factors of five or more (i.e.. Aux R-^ values of 0.80 or more) and variance inflation of diis magnitude is generally considered to be severe. This stiidy found that the real Canada-U.S. exchange rate was an important determinant of the injury suf fered by U.S. lumber producers during die period from 1951 to 1986. Five of die six preferred injury models con tained significant parameter estimates for the real Canada-U.S. exchange rate. In the counterfactual scenario diat held the exchange rate at its 1975 level, die largest impact was on die U.S. softwood lumber price which rose 16.8 per cent by 1986, relative to tiie base case prediction. This scenario also resulted in a 14.1 percent increase in U.S. lumber production by 1986. Holding die exchange rate at its 1975 level also resulted in a 2.3 percent increase in die retiim on sales and a 100.2 percent increase in the stock price index. Thus, die exchange rate appears to be a very important determinant of profitabiUty. The results of the scenario that held U.S. GNP at its 1975 level indicate that income is also an important determinant of U.S. lumber prices and production. The hypothesized drop in U.S. GNP to its 1975 value caused the price of lumber and U.S. production to fall by 12.3 percent and 51.7 percent respectively. Furthermore, this scenario resulted in 58.4 percent drop in die demand for U.S. sawmill labour. The preferred model for Canadian market share does not contain U.S. GNP explicidy^^. However, die direction of die effect on Canadian market share caused by a drop in GNP is known. A lower level of U.S. GNP will lower the demand for housing and, dius, die square footage of U.S. residential constiuction will drop. This study found a positive relationship between the square footage of U.S. residential constiuction and Canadian market share consequendy, a drop in U.S. GNP lowers Canada's market share. One of die primary contentions of U.S. lumber producers in Canada-U.S. softwood lumber dispute was diat stumpage prices in Canada and in B.C. in particular were too low (i.e., below fair market value) and diat this consti tuted a subsidy which caused a large distortion in Canadian lumber production. From a statistical perspective, the U.S. allegation was diat British Columbian production decisions were distorted by the timber rents captured by sawmills and that die resulting increase in lumber shipments to die U.S. injured U.S. producers. To test diis allegation, die average real value of rents flowing through to British Columbian sawmills (i.e., the proxy measure of the alleged subsidy) was entered into each of the six reduced form models. Widi die ex ception of die model for U.S. lumber production, diis study found diat die alleged subsidy had very litde impact on the long-run performance of die U.S. lumber sector. In die case of U.S. lumber production, a simulated reduction in the level of captured rents to zero increased U.S. lumber output by only 1.4 percent. Thus, die results of this analy sis do not support die U.S. allegations. That is, timber rents diat may have flowed dirough to British Columbian sawmills do not appear to have been a major source of injury. As shown in Table 3.11, the preferred model specifies Canadian market share as a function of die following variables: USFOOT, REXCHRT, RUSAVC, RCAVC and TREND. U.S. GNP and die business cycle effect were replaced by die square footage of new U.S. residential construction because of multicollinearity problems. However in the section above oudining the specification of Expression 20, the following model was reported: USFOOT = - 4.0414 + 0.92523E-03*USGNP -i- 4.5691*CUSGNP - 0.11819*USINT Thus, die square footage of new U.S. residential construction is positively related to U.S. GNP and to die business cycle. The sensitivity of the six injury measures to the British Columbian stumpage price was also assessed in this study. Of die six injury measures, only die models for die U.S. lumber price and U.S. production contain sig nificant British Columbian stumpage price coefficients. In response to a simulated increase in the British Columbian stumpage price to $C 10.00 per cubic mette, it was found that the U.S. lumbw price and U.S. produc tion rose by 14.6 percent and 5.6 percent respectively. Like the exchange rate scenario, the U.S. price of lumber is affected more dramatically by a shift in Canada's excess supply of lumber than by changes in die level U.S. lumber production. In general, the results of diis study tend to support Canada's counter claim in die lumber dispute diat it was changes in U.S. domestic lumber demand and factor markets tiiat were die major cause of die suffering in the U.S. lumber industty. Over the time horizon in this study, the conditions faced by U.S. producers changed dramatically. The real appreciation of the U.S. dollar against die Canadian dollar, rising demand for saw timber from plywood and odier panel products, increased competition firom substitute products in ttaditional end-use markets and higher average variable manufacturing costs in die U.S. dian in Canada over most of die study period put U.S. producers in dire straits. The notion of injury is inherendy a problem of comparing healdiy and injured states using an appropriate measure. By adopting die measures of injury reviewed in this study, the comparison of healthy and injured states be comes an empirical exercise diat can be addressed using relatively simple reduced form models and standard economet ric techniques. Although there will inevitably be situations where data availability or structural complexity will limit the usefulness of reduced form models in injury determinations, there are quite likely a great many cases where this approach would ensure or improve the objectivity of the various processes that are currendy being used to decide if an unfair trading practice has caused injury to domestic producers. Adams, D.M., and R.W. Haynes. 1985. Changing Perspectives on die Oudook for Timber in die United States. Journal of Forestiy 83: 32-35. Adams, D.M., and R.W. Haynes. 1980. The 1980 Timber Assessment Market Model: Stiucture, Projections and Policy Simulations. Forest Science Monograph No. 24. 64 p. Adams, D.M., B.A. McCarl and L.H. Homayounfarrokh. 1986. The Role of Exchange Rates in Canadian-United States Lumber Trade. Forest Science 32(4): 973-988. Adams, D.M., K.C. Jackson and R.W. Haynes. 1987. Production, Consumption and Prices of Softwood Products in North America: Regional Time Series data, 1950 to 1985. Resource Bulletin PNW-RB-151, USDA Forest Service, Pordand, Oregon. Adams, E.G. and J. Blackwell. 1974. An Econometiic Model of die United States Forest Products Industiy. Forest Science 19: 82-96. Alberta Department of Forestiy, Lands and WUdlife. Various Years. Annual Report. Alberta Department of Forestiy, Lands and Wildlife, Edmonton, Alberta. Austin, J.W. and D JR. Darr.1975. The Jones Act and die Softwood Lumber Industry. Journal of Forestiy: 644-648. Balassa, B. 1989. Subsidies and Countervailing Measures: Economic Considerations. Journal of World Trade 23(2): 63-79. Baldwin, RE. 1989. The Political Economy of Trade PoUcy. Journal of Economic Perspectives 3(4): 119-135. Bank of Canada. 1988. Selected Canadian and International Interest Rates Including Bond Yields and Interest Arbitrage. Bank of Canada, Department of Monetary and Financial Analysis, Ottawa, Ontario. 67 pp. Belsley, D.A., E. Kuh and R.E. Welsch. 1980. Regression Diagnostics: Identifying Influential Data Sources of Collinearity. John WUey and Sons, New York. Bhagwati, J. 1988. Protectionism. MIT Press, Cambridge, Mass. 147 pp. Boyd, R. and K. KrutUla. 1987. The Welfare Implications of U.S. Trade Restiictions Against die Canadian Softwood Lumber Industiy: A Spatial Equilibrium Analysis. Canadian Journal of Economics 20(1): 17-35. Brander J. and B. Spencer. 1981. Tariffs and die Extraction of Foreign Monopoly Rents Under Potential Entry. Canadian Journal of Economics 14: 371-389. Brander J. and B. Spencer. 1985. Export Subsidies and International Market Share Rivalry. Journal of International Economics 18: 82-100. Brander, J. A. 1988. Government PoUcy Toward Business. Butterwordis, Toronto, Ontario. 320 pp. Brealey, R., S. Myers, G. Sick and R. Whaley. 1986. Principles of Corparate Finance: First Canadian Edition. McGraw-Hill Ryerson Ltd.. Toronto. 886 pp. Breton, A. 1985. Supplementary Statements, pp.486-526. in. Royal Commission on die Economic Union and Development Prospects for Canada. Volume Three. 1985. Minister of Supply and Services, Ottawa, Ontario. 699 pp. British Columbia Ministry of Forests. Various Years. Annual Report. Ministry of Forests. Victoria, B.C. Buongiomo, J. and H.C. Lu. Effects of Cost Demand and Labor Productivity on the Prices of Forest Products in die United States, 1958-1984. Forest Science 35(2): 349-363. Buongiomo, J., J.P. Chavas and J. Uusivouri. 1988. Exchange Rates, Canadian Lumber Imports and United States Prices: A Time Series Analysis. Canadian Joumal of Forest Research 18: 1587-1594. Button, K.R. 1989. The United States-Canada Free Trade Agreement An Overview and an Assessment for die U.S. Non-ferrous Metals and Forest Products Industries. Law and Policy in Intemadonal Business : 765-793. Canadian And-dumping Tribunal. 1983. Annual Report: 1983. Revenue Canada, And-dumping Tribunal, Ottawa. Canadian Intemadonal Trade Tribunal. 1987. Annual Report: 1987. Department of Finance, International Trade Tribunal, Ottawa, Ontario. Cass, R.A. 1989. Economics in die Administtation of U.S. Intemadonal Trade Law. Ontario Centte for Intemadonal Business Working Paper No. 16, July 1989. Ontario Centre for International Business, University of Toronto, Ontario. 34 pp. Chen, NJ., G.C.W. Ames and A.L. Hammett. 1988. Implications of a Tariff on Imported Canadian Softwood Lumber. Canadian Joumal of Agriculttttal Economics 36: 69-81. Coalition for Fair Lumber Imports. 1986. In the Matter of Certain Softwood Lumber Products from Canada: Petition for the Imposition of Countervailing Duties Pursuant to die Tariff Act of 1930, as Amended. Coalition for Fair Lumber Imports, Washington, D.C Constantino, L J. and D. Haley. 1989. A Comparative Analysis of Sawmilling Productivity on die British Columbia Coast and in die U.S. Douglas-fir Region: 1957-1982. Forest Products Joumal 39(4): 57-61. Coughlin,C.C., J.V. Terza and N.A. Khalifah. 1989. The Determinants of Escape Clause Petitions. The Review of Economics and Statistics 71(2): 341-347. DaiT, D.R. 1975. Softwood Log Exports and die Value and Employment Issues. Research Paper No. PNW 200. USDA Forest Service, Pacific Northwest Range and Experiment Station, Pordand Oregon. 13 pp. Darr, D.R. 1977. Floating Exchange Rates and Log Export Policy. Joumal of Forestiy 75(2): 88-90. Darr, D.R., R.W. Haynes and D.M. Adams. 1980. The Impact of die Export and Import of Raw Logs on Domestic Timber Supplies and Prices. Research Paper No. PNW 277. USDA Forest Service, Pacific Northwest Range and Experiment Station, Portland Oregon. 38 pp. Deardorff, A.V. and R.M. Stem. 1989. Current Issues in Trade Policy. ppl5-76 in. Stem, R.M. ed. 1989. U.S. Trade Policies in a Changing World Economy. MIT Press, Cambridge, Massachusetts. 437 pp. Dombusch, R. and J.A. Frankel. 1989. Macroeconomics and Protection, pp 77-144 in. Stem, R.M. ed. 1989. U.S. Trade Policies in a Changing World Economy. MIT Press, Cambridge, Massachusetts. 437 pp. Dunn, R.M. Jr. 1974. Canada's Experience Widi Fixed and Flexible Exchange Rates in a North American Capital Market, pp. 465-479 in. Keirstead, B.S. J.F. Earl, J.R.G. Brander and CM. Waddell. 1974. Economics Canada: Selected Readings. Macmillan of Canada, Economic Council of Canada. 1988a. Back to Basics: Twenty-Fifdi Annual Review, 1988. Minister of Supply and Services, Ottawa, Ontario. 82 pp. Economic Council of Canada. 1988b. Adjusttnent Policies for Trade-Sensitive Industties. Minister of Supply and Services, Ottawa, Ontario. 194 pp. Economic Council of Canada. 1987. Reaching Outward: Twenty-Fourdi Annual Review, 1987. Minister of Supply and Services, Ottawa, Ontario. 82 pp. Economic Report of die President Transmitted to Congress Febuary 1991 Together Widie die The Annual Report of die CouncU of Economic Advisors. 1991. U.S. Govemment Printing Office, Washington, D.C. 411 pp. Epstein, E.M. 1969. The Corporation in American Politics. Prentice-Hall, Englewood CUffs, NJ. Federal Reserve Bank of San Francisco. 1962. Mondily Review, March 1962. Federal Reserve Bank of San Francisco, San Francisco, CaUfomia. Feigenbaum, S., H.O. Ortiz and T.D. Willett. 1985. Protection Pressures and Aggregate Economic Conditions: Comment on Takacs. Economic Inquiry Vol. 23: 175-182. Feinberg, R.M. 1989. Exchange Rates and Unfair Trade. The Review of Economics and Statistics 71(4): 704-707. Feinberg, R.M. and B.T. Hirsch. 1989. International Journal of Industrial Organization 7: 325-340. Finger, J.M. 1981. The Industty-Counuy Incidence of "Less-dian-Fair Value" Cases in U.S. Import Trade. Quarterly Review of Economics and Business 21(2): 260-279. Finger, J.M., H.K. Hall and D.R. Nelson. 1982. The Political Economy of Administered Protection. The American Economic Review 72(3): 452-466. Finger J.M. and J. Nogués. 1987. International control of Subsidies and Countervading Duties. The World Bank Economic Review 1(4): 707-725. Forestiy Canada. 1990. Forestiy Facts: 1990. Ministiy of Supply and Services, Ottawa, Ontario. Forestiy Canada. Various Years. Selected Forestry Statistics for Canada. Forestiy Canada, Economics and Statistics Directorate, Ottawa, Ontario. Gaims, CH. 1982. Provincial Charges for Timber in Canada. Industiial Forest Service Ltd. Prince George, B.C. Giese, J.L. 1986. The Special Import Measures Act: Balancing die Interests of Foreign Exporters and Canadian Industiies. Journal of World Trade Law 21(3): 9-25. Gouvernement du Quebec Ministère de l'Energie et des Ressources. Various Years. Rapport Annuel. Ministère de l'Energie et des Ressources. Quebec, Quebec. Grdli£. 1988. Macro-economic Determinants of Trade Protection. The World Economy 11(3): 313-326. Grossman, G.M. 1986. Imports as a Cause of Injury: The Case of the U.S. Steel Industry. Journal of International Economics 20: 201-223. Gunton, T. and J. Richards. 1987. pp 1-57 in. Resource Rents and Public Policy in Western Canada. Institute for Research in PubUc Policy. Halifax, N.S. 261 pp. Haid, D. The Canada-U.S. Lumber Dispute: A Welfare Impact Analysis. Unpublished Paper. Forest Economics and Statistics Directorate, Forestiy Canada, Ottawa, Ontario. 19 pp. Hartman, D.A., W.A. Addnson, B.S. Bryant and R.O. Woodfm. 1981. Conversion Factors for die Pacific Northwest Forest Industry: Converting Forest Growth to Forest Products. Institute of Forest Products, College of Forest Resources, University of Washingto Herander, M.G. and J.B. Schwartz. 1984. An Empirical Test of die Impact of die Threat of U.S. Trade Policy: The Case of Antidumping Duties. Southem Economic Journal 51:59-79. Hibbs, D.A. Jr. 1987. The American Pohtical Economy: Macroeconomics and Electoral Politics. Harvard University Press, Cambridge, Massachusetts. 404 pp. Hufbauer, G.C. and J. S. Erb. 1984. Subsidies in International Trade. Institute for Intemational Economics, Washington, D.C. Distributed by MIT Press, Cambridge, Massachusetts. 283 pp. Intemational Monetary Fund. 1988. Intemational Financial Statistics Yearbook, 1988. IMF, Washington, D.C. International Woodworkers of America. 1985. Productivity and Unit Production Costs in die Softwood Lumber Industiies of die United States and Canada, 1977 to 1984. International Woodworkers of America, Pordand, Oregon and Vancouver, B.C. Jansen, G.W.V. 1984. Canad-United States Trade Relations: The Lessons of Softwood Lumber Countervail Case. Executive Bulletin. The Conference Board of Canada. Ottawa, Ontario. 21 pp. Johnson, J. 1984. Econometiic Mediods, 3rd ed. McGraw-Hill Book Company, New York, N.Y. 568 pp. Judge, G.G., R.C. Hill, W.E. Griffidis, H. Ludcepohl and T.C. Lee. 1988. 2nd. ed. Introduction to die Theory and Practice of Econometiics. John WUey and Sons, New York. 1024 pp. Kaempfer, W.A. and T.D. WUlett. 1989. Combining Rent-seeking and Public Choice Theory in die Analysis of Tariffs Versus Quotas. Public Choice 63:79-86. Kalt, J.P. 1987. The Political Economy of Protectionism: Tariffs and Retaliation in die Timber Industry. Discussion Paper Series. Energy and Environmental PoUcy Center, John F. Kennedy School of Govemment, Harvard University, Cambridge, Massachusetts. 33 pp. Kelly, K. 1988. The Analysis of Causality in Escape Clause Cases. The Joumal of Industiial Economics 28: 187-207. Kennedy, KC. 1989. The Canadian and U.S. Responses to Subsidization of International Trade: Toward a Harmonized Countervailing Duty Legal Regime. Law and Policy in International Business: 683-764. Kerr, W.A. 1987. The Recent Finding of die Canadian Import Tribunal Regarding Beef Originating in die European Economic Community. Joumal of World Trade Law 21(5): 55-65. Kerr, W.A. and S.E. Cullen. 1989. International Competitiveness and Resource Rents - Insights from die Canada-U.S. Softwood Lumber Dispute. Worid Competition 12(3): 25-39. Kmeger, A. 1980. Protectionist Pressures, Imports and Employment in the United States. Scandinavian Joumal of Economics 82: 133-146. Lea, S. 1962. The U.S. Softwood Lumber Situation in a Canadian-American Perspective. Canadian-American Committee, National Planning Association (U.S .A.) and the Private Planning Association of Canada, Washington, D.C. 52 pp. Lenway, Stefanie. 1983. The Impact of American Business on Intemadonal Trade Policy. Research in Corporate Social Performance and Policy, JAI Press No. 5: 27-58. Levi, M.D., 1975. World-wide Effects and Import Elasticities. Joumal of International Economics 6: 203-214. Lindblom C. 1977. Politics and Markets: The World's Political and Economic Systems. Basic Books, New York. Magee, S.P. and L. Young. 1989. Endogenous Protection in die United States, 1900-1984. pp. 145-206 in. Stem, R.M. ed. 1989. U.S. Trade Policies in a Changing Worid Economy. MIT Press, Cambridge, Massachusetts. 437 pp. McCarl, B.A. and R.W. Haynes. 1985. Exchange Rates Influence Softwood Lumber Trade. Joumal of Forestiy 83: 368-370. McKenzie, R.B. 1987. The Fairness of Markets: A Search for Justice in a Free Society. Lexington Books, Lexington, Massachusetts. 235 pp. McKinnon, R.I. 1978. Money in Intemational Exchange: The Convertible Currency System. Oxford University Press. New York, New York. McKinnon, R.I. 1978. Money in Intemational Exchange: The Convertible Currency System. Oxford University Press. New York, New York. Mishan, EJ. 1988.4th ed. Cost Benefit Analysis. Unwin Hyman, London. 461 pp. Montgomery, D.C. and E.A. Peck. 1982. Introduction to Linear Regression Analysis. John Wiley and Sons, New York. 504 pp. Moody. Various Issues. Moody's Industiial Manual: American and Foreign. Morgan Guarantee Trust Company. Various Issues . World Financial Maricets . Morgan Guarantee Trust Company, New Yoric. Morkre, ME. and H.E. Krudi. 1989. Determining Whedier Dumped or Subsidized Imports Injure Domestic Industiies: Intemational Trade Commission Approach. Contemporary PoUcy Issues 7(3): 78-95. New Brunswick Ministiy of Natoral Resources. Various Years. Annual Report of die Ministiy of Natural Resources, Ministiy of Natural Resources, Fredericton, N.B. Nicolaides, P. 1987. How Fair is Fair Trade? Joumal of World Trade Law 21(4): 147-162. OECD. 1990. Main Economic Uidicators: Historical Statistics 1969-1988. OECD, Department of Economics and Statistics, Paris. OECD. National Accounu Detailed Tables Volume II 1976-1988. OECD, Economics and Statistics, Paris. OECD. OECD Economic Oudook. Various Issues. OECD Publications Service Paris. OECD. OECD Economic Oudook. Various Issues. OECD Publications Service Paris. 01son>I. 1982. The Rise and Decline of Nations: Economic Growdi, Stagflation and Social Regidities. Yale University Press, New Haven, Conn. Olson, M. 1965. The Logic of Collective Action. Public Goods and the Theory of Groups. Harvard University Press, Cambridge, MA. Ontario Ministiy of Natural Resources. Various Years. Annual Report of the Minister of Natural Resources. Ministry of Natural Resources, Sault Ste. Marie, Ontario. Pahneter, N.D. 1986. Injury Determinations in Antidumping and CountervaiUng Duty Cases - A Commentary on U.S. Practice. Journal of Worid Trade Law 21(1): 7-45. Paper Tree Letter. Various Issues. Paper Tree Letter Independent Analysis of Forest Products Economics. Paper Tree Letter, Vancouver, B.C. Percy, M.B. 1986. Forest Management and Economic Growdi in British Columbia. A Study Prepared for die Economic Councd of Canada. Minister of Supply and Services, Ottawa, Canada. 88 pp. Percy, M.B. and C. Yoder. 1987. The Softwood Lumber Dispute and Canada-U.S. Trade in Nattual Resources. InstitiJte for Research in Public PoUcy, HaUfax, Nova Scotia. 171. Phelps, E.S. PoUtical Economy: An Introductory Text. W.W. Norton and Company, New York. 618 pp. Phelps, R.B. and H. Spelter. 1984. Changes in Post War U.S. Lumber Consumption Pattems. Forest Products Joumal 34(2): 35-41. Pindyck, R.S. and J J. Rotemberg. 1987. Are Unports to Blame? Attribution of Injury Under die 1974 Trade Act. Joumal of Law and Economics 30:101- 122. Ramsey, J.B. 1969. Tests for Specification Error in Qassical Linear Least-squares Regression Analysis. Joumal of die Royal Statistical Society, Series B, Part 2: 350-371. Random Lengths. Various Issues. Random Lengths yearbook. Random Lengths.Publications Inc., Eugene, Oregon. Rinfret-Boston Associates. 1974. Prices and Producdon: An Economic Analysis of die Production, Marketing and Pricing Behavior of Softwood Lumber and Plywood Products. Rinfret-Boston Associates, New York, New York. 65 pp. Roberts, D.G. 1988. The Impact of Exchange Rate Changes on die Canadian Forest Products Sector. Working Paper Series. Economics Branch, Forestiy Canada, Ottawa. Robinson, Vi. 1974. An Econometric Model of Softwood Lumber and Stumpage Markets. Forest Science 20: 171-179. Rossides, E.T. 1986. U.S. Import Trade Regulation. The Bureau of National Affairs, Washington, D.C. 728 pp. Rugman, A.M. and S.D. Porteous. 1989. Canadian and U.S. Unfair Trade Laws: A Comparison of Their Legal and Administrative Structures. Working Paper. Ontario Centre for Intemational Business Research Programme J^aculty of Management, University of Toronto. Salvatore, D. 1987a. Import Penetiation, Exchange Rates, and Protectionism in die United States. Joumal of Policy Modeling 9: 125-141. Salvatore, D. 1987b. ed. The New Protectionist Threat to World Welfare. Nordi-Holland, New Yoric. 581 pp. Schattschneidet, E.E. 1963. PoliticsJ>ressures and the Tariff. Archon Books, Hamden, Connecticut 301 pp. Sharma, M.L. 1986. The Economic Impact of Tariff and Quota Restrictions by the United States on Imported Canadian Lumber. M.ScF. Thesis. University of Alberta, Department of Rural Economy, Edmonton, Alberta. 208 pp. Sherman, D. 1962. Canadian Views on Canadian Shipment of Lumber to Atlantic Seaboard Markets. The Lumberman and Wood Industiies, May: p. 9. Sherman, D. 1962. Lumber Industiy Takes Positive Steps to Offset Import Competitive Imbalance. The Lumberman and Wood Industries, May: 50-53. Slayton, P. 1979. The Antidumping Tribunal: A Study of Administrative Procedure in die Antidumping Tribunal. Law Reform Commission of Canada, Ottawa, Ont. Ill p. Spelter, H. 1985. A Product Diffusion Approach to Modelling Softwood Lumber Demand. Joumal of Forest Science 31: 685-700. Stanbury, W.T. 1986. Business Govemment Relations in Canada: Grappling With Leviadian. Mediuen Publications, Toronto, Ontario. 678 pp. Standard and Poor. 1988. Standard and Poor's Analyst's Handbook, Official Series: 1988 Annual Edition. Statistics Canada. 1959. Operations in die Woods, Revised Estimates of Forest Production, 1950-1953 - Final Estimates, 1954-1955. Catalogue No. 25-501. Statistics Canada, Ottawa, Ontario. Statistics Canada. 1968. Canadian Forestry Statistics Revised 1964: Reference Paper. Catalogue No. 35-503. Statistics Canada, Ottawa, Ontario. Statistics Canada. 1988. Quarteriy Economic Summary. Catalogue No. 13-007E. Ottawa, Ontario. Statistics Canada. Various Years. Catalogue No. 13-201. Statistics Canada. Various Years. Catalogue No. 13-531. Statistics Canada. Various Years. Logging Industiy. Catalogue No. 25-201. Statistics Canada, Ottawa, Ontario. Statistics Canada. Various Years. Sawmills, Planing Mills and Shingle Mills. Catalogue No. 35-204. Statistics Canada, Ottawa, Ontario. Stem, P. and A. Wechsler. 1984. Escape Clause Relief and Recessions: An Econometric and Legal Look at Section 201. U.S. Intemational Trade Commission, Washington, D.C. Swanson, Nil. and R. Jacques. 1985. Changing U.S. Softwood Lumber Market Shares: An Econometiic Analysis. UnpubUshed Paper. Economics Branch, Canadian Forestry Service, Ottawa. Takacs, W. 1981. Pressures for Protectionism: An Empirical Analysis. Economic Inquiry 19: 687- 693. Takacs, W. 1985. More on Protectionist Pressures and Aggregate Economic Conditions: A Reply. Economic Inquiry 23: 183-184. Uhler, R. 1987. Canadian PubUc Timber Pricing in die Great Subsidy Debate. Working paper No. 100. Forest Economics and PoUcy Analysis Research Unit, University of British Columbia, Vancouver B.C. 39 pp. United States Congress. 1983. Wood Use U.S. Competitiveness and Technology. United States Congress, Office of Technology Assessment, Washington, D.C. 202 pp. United States Department of Agriculhire. 1990. An Analysis of die Timber Situation in die United States: Part 1: The Current Resource and Use Sitiiation. Draft Edition. United States Department of Agricultiu-e, United States Forest Service, Washington, D.C. United States Department of Agriculture. 1988. U.S. Timber Production, Trade, Consumption and Price Statistics 1950-86. Misc. Publication No. 1460. United States Department of Agricultiire, United States Forest Service, Washington, D.C. 81 pp. United States Department of Agricultiire. 1982. An Analysis of die Timber Situation in die United States: 1952-2030. Forest Resource Report No. 23. USDA Forest Service, Washington, D.C. 499 pp. United States Department of Commerce. 1989. Statistical Abstarct of die United States. 1989. 109di Edition. U.S. Department of Commerce, Bureau of die Census. Washington, D.C. 956 pp. United States Department of Commerce. Various Years. Annual Survey of Manufactures. United States Department of Commerce, Bureau of the Census, Washington, D.C. United States Department of Commerce. Various Years. Annual Survey of Manufactures. United States Department of Commerce, Bureau of die Census, Washington, D.C. United States Department of Commerce. Various Years. Business Statistics. United States Department of Commerce, Bureau of Economic Analysis, Washington, D.C. United States Department of Commerce. Various Years. Census of Manufactiires, Logging Camps, Sawmills and Planing Mills, Industry Series. United States Department of Commerce, Bureau of the Census, Washington, D.C. United States Department of Commerce. Various Years. Handbook of CycUcal Indicators. United States Department of Commerce, Bureau of Economic Analysis, Washington, D.C. United States Department of Labor. Various Years. Producer Prices and Price Indexes. United States Department of Labor, Bureau of Labor Statistics, Washington, D.C. United States Intemational Trade Administration. 1986. Preliminary Affirmative Countervailing Duty Determination: Certain Softwood Lumber Products fixim Canada. United States Department of Commerce, Intemational Trade Administration, Washington, D.C. 81 pp. United States Intemational Trade Commission. 1985. Report to die President on Investigation No. 332-210 Under Section 332 of die Tariff Act of 1930, Conditions Relating to die Importation of Softwood Lumber into die U.S. USrrC Publication 1765. United States Intemational Trade Commission, Washington D.C. United States Intemational Trade Commission. 1982. Softwood Lumber from Canada: Determination of the Commission in Investigation No. 701-TA-197 (Preliminary) Under Section 703(a) of tiie Tariff Act of 1930. USrrC Publication 1320. United States Intemational Trade Administration, Washington, D.C. United States Intemational Trade Commission. Various Years. Annual Report USITC Publication, U.S. Govemment Printing Office, Washington, D.C. United States Intemational Trade Commission. Various Years. Operation of die Trade Agreements Program. Various Years. USITC Publication, U.S. Govemment Printing Office, Washington, D.C. United States Senate. 1963. Hearings Before the Committee on Commerce on the Impact of Lumber Imports on the United States Softwood Lumber Industiy, Serial No. 20. U.S. Govt Printing Office. United States Senate. 1962. Hearings Before tiie Committee on Commerce on die Impact of Lumber Imports on die United States Softwood Lumber Industiy, Part 1. U.S. Govemment Printing Office. United States Tariff Commission. 1975. Annual Report. United States Department of die Treasury, U.S. Tariff Commission, U.S. Govemment Printing Office, Washington, D.C. Varian, H.R. 1984. Microeconomic Analysis. W.W. Norton and Company, New York, New York. 348 pp. Wares, W.A. 1977. The Theory of Dumping and American Commercial Policy. Lexington Books, Lexington, Mass. 130 p. Warren, D.D. Various Years. Production, Prices, Employment and Trade in Northwest Forest Industiies. Published Quarterly. USDA Forest Service, Pacific Northwest Range and Experiment Station, Portland Oregon. White, K.J., S.D. Wong, D. WisUer and S.A. Haun. 1990. SHAZAM: Econometiics Computer Program. User's Manual Version 6.2. McGraw-Hill, New York, New York. 352 pp. Wiseman, A.C. and R.A. Sedjo. 1981. The Effects of an Export Embargo on Related Goods: Logs and Lumber. The American Joumal of Agricultural Economics, August 1981. Wonnacott, P and J. WUliamson. 1987. The United States and Canada: The Quest for Free Trade: An Examination of Selected Issues. Policy Analysis in Intemational Economics Report No. 16. Institute for International Economics, Washington, D.C. 173 pp. Appendix 2.1 U.S. LFV Data and Predicted Values YEAR Actual ULFV Pred ULFV USREFEXR 1980=100 JEŒFEXR 1980=100 USJCI URDPFT Billion 1980 $ U.S. URIMP Billion 1980 $ U.S 1975 8 0.00 110.4 103.5 1.07 28.61272 143.5116 1976 14 21.15 114.4 102.5 1.12 42.52717 169.3071 1977 15 24.61 111.4 110.5 1.01 49.17197 193.0318 1978 23 12.80 100.3 127.6 0.79 52.90629 208.8375 1979 24 19.00 99.0 110.8 0.89 40.67612 229.3239 1980 103 78.70 100.0 100.0 1.00 21.30000 245.2600 1981 32 35.84 111.5 108.8 1.02 19.16058 238.1204 1982 181 183.36 125.7 97.1 1.29 1.79949 209.0403 1983 55 48.20 129.8 101.3 1.28 14.19142 212.9125 1984 101 115.14 139.7 102.3 1.37 30.26211 258.7212 1985 95 87.80 145.6 99.9 1.46 21.99074 266.4198 1986 95 84.81 117.0 119.8 0.98 23.14050 277.9564 1987 18 44.59 101.5 123.9 0.82 29.90518 296.1342 Appendix 2.2 Canadian LFV Data and Predicted Values Actual Pred CREXR GREXR CGRPC CMEI CIP YEAR CLFV CLFV 1980=100 1980=100 CGCI % 1980=100 1980=100 RTCMLAP 1975 8 0.00 111.9 91.9 1.2176 4.7 97.98 84.7 1.00 1976 32 24.33 124.3 93.6 1.3280 6.5 99.30 89.3 1.04 1977 60 52.62 116.4 99.6 1.1687 3.2 97.98 91.6 1.04 1978 32 35.23 103.8 102.3 1.0147 3.4 98.95 95.5 1.03 1979 26 22.07 100.3 103.6 0.9681 2.9 101.93 100.7 1.02 1980 21 21.91 100.0 100.0 1.0000 2.2 100.00 100.0 1.01 1981 24 25.96 106.5 91.0 1.1703 2.3 100.61 101.0 1.00 1982 73 70.95 109.7 91.1 1.2042 -2.6 91.24 90.2 0.98 1983 29 32.10 114.0 92.5 1.2324 3.4 88.69 95.3 1.09 1984 25 24.55 111.7 90.2 1.2384 4.6 85.89 103.7 1.12 1985 35 35.25 106.4 89.7 1.1862 5.2 87.64 108.2 1.02 1986 19 37.41 198.3 97.8 1.0051 4.4 89.48 109.6 0.99 . 1987 32 25.63 99.9 104.4 0.9569 4.5 101.23 114.9 0.93 Appendix 3.1 Data Listing for Imports as a Cause of Injury USLMPI USLMPROD CANMSHR FPROS RSPFPI RSPAFPl Real Billion Real Real YEAR 1986=100 fbm % % 1986=100 1986=100 1950 120.31 30.6 8.70 13.27 -9999 -9999 1951 124.52 29.5 6.74 12.98 -9999 -9999 1952 122.16 30.2 6.71 10.28 -9999 -9999 1953 117.85 29.6 7.62 8.97 -9999 -9999 1954 114.13 29.3 8.70 9.12 -9999 -9999 1955 117.71 29.8 9.93 11.11 -9999 -9999 1956 114.80 30.2 9.34 10.72 -9999 -9999 1957 104.11 27.1 9.06 9.17 -9999 -9999 1958 100.11 27.4 10.29 8.36 -9999 -9999 1959 106.06 30.5 10.89 8.56 -9999 -9999 1960 97.99 26.7 12.06 6.85 -9999 -9999 1961 92.00 26.1 13.37 5.64 -9999 -9999 1962 92.23 26.8 14.65 4.22 -9999 -9999 1963 93.00 27.6 15.60 5.30 -9999 -9999 1964 92.65 29.3 14.58 6.51 -9999 -9999 1965 90.16 29.3 14.53 6.95 112.69 16.30 1966 91.42 28.8 14.46 5.84 93.03 -110.81 1967 90.98 27.3 15.25 4.89 109.40 13.62 1968 104.65 29.3 16.88 5.79 166.05 537.52 1969 110.60 28.3 17.46 6.06 192.72 872.64 1970 88.14 27.5 17.79 4.43 190.09 1085.68 1971 104.35 30.0 19.76 4.67 180.98 862.58 1972 117.95 31.0 22.88 6.99 150.05 449.77 1973 141.48 31.6 22.75 8.80 143.60 481.93 1974 128.09 27.7 20.43 7.20 121.08 553.04 1975 110.61 26.7 18.30 5.13 119.13 582.86 1976 128.67 30.6 21.41 6.78 152.26 842.64 1977 144.53 32.7 24.80 6.66 127.59 643.43 1978 156.67 33.5 26.77 6.85 104.58 466.57 1979 158.19 33.3 26.01 7.59 100.71 427.21 1980 131.69 28.2 26.69 4.83 94.38 313.03 1981 119.38 25.4 28.16 3.76 84.93 248.79 1982 105.23 24.9 28.12 1.86 69.98 150.70 1983 116.44 28.9 30.57 2.12 90.40 230.35 1984 107.27 30.8 31.20 1.91 74.10 68.50 1985 101.83 30.5 33.34 2.07 75.71 -7.84 1986 100.00 34.2 30.35 4.59 100.00 100.00 Data Lisdng for Imports as a Cause of Injury Continued USSMHRS USGNP REXCHRT RUSAVC Million Billions Real 1986 Real 1986 YEAR Hours $U.S. CUSGNP $C/$U.S. $U.S./MBF 1950 751.0 1335.4 -9999 1.2286 86.73 1951 693.3 1473.6 1.103 1.1195 92.93 1952 687.7 1531.1 1.039 1.0113 94.15 1953 645.6 1592.4 1.040 1.0337 97.59 1954 558.4 1571.2 0.987 1.0223 90.47 1955 598.9 1658.5 1.056 1.0649 103.96 1956 592.7 1692.7 1.021 1.0581 103.93 1957 532.0 1720.9 1.017 1.0468 109.01 1958 407.4 1707.8 0.992 1.0661 90.61 1959 442.1 1807.4 1.058 1.0587 91.28 1960 432.5 1847.5 1.022 1.0759 98.59 1961 380.7 1895.7 1.026 1.1134 92.17 1962 378.1 1996.4 1.053 1.1833 92.10 1963 364.2 2078.4 1.041 1.1851 93.87 1964 354.2 2189.3 1.053 1.1737 93.15 1965 358.7 2316.1 1.058 1.1646 93.83 1966 346.7 2450.1 1.058 1.1515 95.06 1967 322.7 2520.0 1.029 1.1410 103.77 1968 316.2 2624.6 1.042 1.1569 105.93 1969 316.6 2688.5 1.024 1.1622 121.57 1970 306.0 2680.7 0.997 1.1375 118.45 1971 301.9 2756.8 1.028 1.1266 122.18 1972 301.9 2894,1 1.050 1.0941 156.15 1973 314.4 3044.4 1.052 1.0831 176.20 1974 328.5 3028.1 0.995 1.0096 226.94 1975 281.2 2990.0 0.987 1.0490 201.17 1976 301.6 3136.1 1.049 0.9949 222.68 1977 315.6 3282.5 1.047 1.0775 258.17 1978 304.8 3456.2 1.053 1.1704 283.02 1979 331.9 3541.7 1.025 1.1884 324.88 1980 296.8 3536.1 0.998 1.1692 363.22 1981 272.7 3604.5 1.019 1.1861 395.14 1982 219.5 3512.6 0.975 1.1961 349.17 1983 244.9 3638.1 1.036 1.1811 363.94 1984 245.2 3879.7 1.066 1.2506 350.29 1985 238.3 4014.9 1.035 1.3166 352.33 1986 247.7 4129.2 1.028 1.3420 344.88 Data Listing for Imports as a Cause of Injury Continued RCAVC BCSTUMP BCLRENT RSP400 USFOOT Real 1986 Real 1986 Real 1986 Billion YEAR $C/MBF $C/m3 $C/m3 (1986=100) ft2 1950 60.46 5.75 -4.57 26.67 2.132890 1951 70.26 7.46 -5.93 29.29 1.718450 1952 70.22 8.70 -8.37 33.47 1.786035 1953 66.92 6.32 -4.60 33.54 1.680086 1954 65.74 6.11 -4.59 41.85 1.891150 1955 66.26 8.11 -6.16 57.31 2.145485 1956 70.55 11.08 -9.68 64.46 1.848560 1957 61.68 6.74 -6.77 60.28 1.711866 1958 59.64 4.61 -5.49 63.79 1.877838 1959 58.77 5.86 -5.75 76.21 2.055872 1960 59.56 5.44 -0.95 74.90 1.715565 1961 59.66 3.94 -1.46 85.96 1.751612 1962 59.70 4.27 0.70 81.00 1.914356 1963 67.97 5.14 -0.38 90.45 2.125682 1964 72.22 6.65 -1.60 106.66 2.075821 1965 72.46 7.91 -2.76 113.09 2.065415 1966 76.29 6.57 -0.37 105.62 1.729896 1967 81.15 4.55 1.81 113.07 1.949663 1968 85.55 6.71 1.57 122.77 2.271030 1969 94.90 8.66 0.63 118.67 2.282104 1970 95.02 4.38 1.85 95.46 2.141256 1971 96.42 4.15 3.15 111.48 3.767111 1972 110.62 6.93 0.11 121.55 3.593265 1973 130.95 13.12 -4.46 104.31 3.333518 1974 151.48 7.73 -3,20 66.83 2.279989 1975 156.25 2.26 -0.32 62.51 1.951696 1976 162.00 2.24 4.52 71.82 2.578916 1977 171.08 3.31 1.42 66.54 3.289558 1978 196.62 7.67 0.41 60.17 3.337581 1979 236.58 12.57 -0.22 58.05 2.923198 1980 252.11 10.84 -2.45 59.77 2.166899 1981 268.72 3.18 -0.09 56.95 1.844315 1982 267.94 2.30 3.46 54.01 1.778640 1983 256.54 2.55 2.03 68.15 2.756250 1984 270.54 2.47 2.13 68.37 2.857488 1985 267.38 2.45 1.40 79.96 2.840746 1986 266.89 3.10 0.67 100.00 2.994623 U.S. Contribudon to Earnings from Building Products Divisions: 1981 Company Division To Sales To Operating Profit Weyerhaeuser Co. Building Products 50% 56% Lousiana-Pacific Corp. Building Products 79% 100% Champion Int. Corp. Building Products 38% 4% Georgia Pacific Coip. Building Products 56% 29% Boise Cascade Corp. Wood Products & 37% * Budding Materials * Two divisions combined posted an operating loss of 23 mdlion dollars 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0086555/manifest

Comment

Related Items