Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Industry structure, market structure and international trade: theoretical and empirical observations White, Lynda E. 1993

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1993_spring_white_lynda.pdf [ 4.79MB ]
JSON: 831-1.0086173.json
JSON-LD: 831-1.0086173-ld.json
RDF/XML (Pretty): 831-1.0086173-rdf.xml
RDF/JSON: 831-1.0086173-rdf.json
Turtle: 831-1.0086173-turtle.txt
N-Triples: 831-1.0086173-rdf-ntriples.txt
Original Record: 831-1.0086173-source.json
Full Text

Full Text

INDUSTRY STRUCTURE, MARKET STRUCTURE AND INTERNATIONAL TRADE:THEORETICAL AND EMPIRICAL OBSERVATIONSbyLYNDA ELAINE WHITEH.B.A., The University of Western OntarioA THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OF MASTERS OF SCIENCEIN BUSINESS ADMINISTRATIONinTHE FACULTY OF COMMERCE AND BUSINESS AND ADMINISTRATION(Policy Analysis Division)We accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAApril 1993© Lynda Elaine White, 1993In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature) e r rt of rAnitivtifee 4,14^mixisibts#01The University of British ColumbiaVancouver, CanadaDate ^Arrit 1+ 1 /993DE-6 (2/88)ABS'I'RACTThis paper describes international trade activity as affected by the structure of markets andindustries. The paper builds upon recent international economic literature that modelsinternational trade within imperfect markets. Arguments for trade intervention by nationalgovernments in order to shift industry rents to the home nation are further surveyed. Similartrade models are used to explain economic growth from both a national and global standing. ACournot perception of competitive behaviour is employed within a basic model developed oftwo-country, intra-industry, oligopolistic trade of a homogeneous good. The model predicts thatlower marginal costs lead to higher output and lower relative costs lead to higher relative marketshare than of the foreign rival industry. However, firm number within the relative industry canoffset cost effects. If fixed costs are present, gains from rationalization may be realized throughtrade and result in higher industry concentration. In order to test the hypothesis that higherconcentration of industry leads to greater trade, particularly export trade, the paper appliesregression technique to measure the relationship between Canadian industry exports and a varietyof concentration variables. Results show more highly concentrated industries to have greaterexport volume. When standardized export and concentration variables are used, weak evidenceis presented that export and concentration in Canada are positively related. However, analysisof United States industries confirms that concentration explains a significant amount of industrypropensity to export. Based on testing and other empirical results, it is concluded thatrationalization of costs associated with export trade and/or ability to discriminate betweenmarkets motivates trade in concentrated industries. Under imperfect market conditions,government may want to intervene in trade by promoting higher concentration of industry if netsocial surplus gains through trade can be realized. Competition policy may promote not onlycombine activity but also cooperation between firms in order to reduce duplication ofinternational trading costs. If competition policy affects international trade partners, regulatorsof world trade activity may take interest in national policy related to industry structure andcompetition as it affects global welfare.iiTABLE OF CONTENTSAbstract^ iiTable of Contents^ iiiList of Tables viList of Figures^ viiAcknowledgements viiiDedication^ viiiChapter One^Introduction 1Chapter Two^Finding a Model for International Trade^ 3Cooperation in Oligopolies 4Non-cooperation in Oligopolies 5Monopolistic Competition 5Cournot Model of Competition^ 6Chapter Three^Imperfect Markets and Government Trade Policy^8Gains from Free Trade in Imperfect Markets 9Gains from Trade Intervention in Imperfect Markets^11Further Considerations^ 16Trade Policy in Imperfect Markets:Conflicting Results 18Chapter Four^Trade in Imperfect Markets and Growth^ 19Chapter Five^Analyzing Imperfect Markets and Trade:Introduction to the Basic Model 24Caveats of the Basic Model^ 25Chapter Six^Analysis of the Basic Model 28Effects of the Cost^ 30Effects of Scale Economies 34Effects of Firm Number 35Firm Number and Industry Seller Concentration^39Empirical Evidence of Concentration and Trade 44iiiChapter Seven Empirical Study:A. Canadian Industry Export, Industry Firm Number,Assets and Average Firm Size^ 50Data Selection^ 50Description of the Data 53Exploration of the Data 54Regression Analysis 56Part A. Conclusions^ 66B. Canadian Industry Exports and IndustryConcentration Measures 67Data Selection^ 67Description of the Data^ 69Statistical Testing 70Regression Analysis 71Part B. Conclusions 76C. Concentration and Export Observations ofUS Manufacturing Industries^ 78Data Selection 78Description of the Data 78Exploration of the Data 79Regression Analysis^ 85Chapter Eight^Empirical Study: Remarks 90Chapter Nine^Implications for Government Policy^ 91Implications from the Basic Model 92Implications from Empirical Evidence forRent Shifting^ 94Analyzing Producer and Consumer Welfare Effects^96Chapter Ten^Final Remarks 97Bibliography 100Appendix One^Sample Data Points: Canadian Industries asDescribed by Standard Industrial Classification(SIC) Codes^ 104Appendix Two^Raw and Log Data PlotsCanadian Industry Data - Export Shipments toThe World 106Appendix Three^Raw and Log Data PlotsivRaw and Log Data PlotsCanadian Industry Data - Total Value of IndustryAssetsRaw and Log Data PlotsCanadian Industry Data - Average Firm Size asMeasured in AssetsRaw and Log Data PlotsCanadian Industry Data - Number of FirmsCanadian Industry CALURA ($000's)1982 Concentration Data DescriptionCanadian Industry CALURA ($000's)1987 Concentration Data DescriptionRaw and Log Data 1982 CALURA PlotsRaw and Log Data 1987 CALURA PlotsCanadian Industry Export and ConcentrationRatio DataRaw and Log Data Plots1982 Canadian Export and Industry ConcentrationNormal Plots of Industry Export andConcentration Log DataAppendix ThreeAppendix FourAppendix FiveAppendix SixAppendix SevenAppendix EightAppendix NineAppendix TenAppendix ElevenAppendix Twelve108110112114120121123Appendix ThirteenUS Industry Export and Concentration DataDescription^ 127Appendix FourteenUS Industry Raw and Log Data Plots^ 128Appendix FifteenUS Industry Export and ConcentrationData Normal Plots^ 130Appendix SixteenResidual Plots of US Industry Exportsand Concentration^ 133LIST OF TABLESSub-industry Asset Size as a Percentage of AggregateIndustry ClassificationDescription of Canadian Export, Asset, Firm Size andFirm Number DataTwo-Sample Statistical Testing p-ValuesRegression Results of Model 1Regression Results of Model 2Regression Results of Model 3Regression Results of Model 4Regression Results of Model 5Test Statistic Results of Model 5 b-CoefficientRegression Results of Model 6USCR4 and USCR8 ANOVA Testing ResultsUS and Canadian Export Propensity ANOVA Testing ResultsUS and Canadian Industry Concentration ANOVA Testing ResultsRegression Results of Model 7Regression Results of Model 8Table 7.1Table 7.2Table 7.3Table 7.4Table 7.5Table 7.6Table 7.7Table 7.8Table 7.9Table 7.10Table 7.11Table 7.12Table 7.13Table 7.14Table 7.15535355586062657374758181838590viLIST OF FIGURESFigure 1.^Greater Industry Market Share^ 37Figure 2.^Relative Aggregate Industry Output, x=y* , n > m^38Figure 3.^Relative Aggregate Industry Output, x>y * , n=m 38Figure 4.^Relative Aggregate Industry Output, x,y * , n < m^39Figure 5.^Relative Aggregate Industry Output, x,y * , n> m 39viiACKNOWLEDGEMENTSI would like to express my sincerestthanks and appreciation to Dr.James Brander, University of BritishColumbia, for his competent andpatient guidance throughout the pasttwo years. I am sincerely grateful forhis work as referee for this thesisand as my advisor for academicwork.DEDICATIONThis paper is dedicated with all mylove to Rick.viiiIndustry Structure, Market Structure and Trade:Theoretical and Empirical ObservationsI.INTRODUCTIONInternational trade flows are a significant part of economic activity for most nations. Since tradeis an important component of national economic wealth, governments take a keen interest in theexporting and importing enterprise within national boundaries and may seek to intervene intrading sectors in order to effect gains in state wealth. Exporting industries are of particularinterest to nations seeking to increase economic benefits derived from international trade sincethey represent a means of expanding Gross Domestic Product (GDP). Although consumption asa measure of national welfare may also be raised through low priced imports, exportingindustries provide additional benefits to a nation of employment, national treasury earnings inforeign exchange and growth through expanded market activity.The means and tools used in the pursuit of economic wealth by nations as it relates to bothexport and import trade is the subject of much debate on trade policy. National trade policiesfocus on the welfare of domestic constituents. By definition, however, international trade ismultilateral and the policies implemented by one nation will have repercussions on the tradingactivity of others. Friction results when government seeks to maximize the wealth of its ownconstituents at the expense of foreigner enterprise or consumers who in turn cry out that suchpolicies make the international playing field "unfair". Indeed, international economic relations1are often characterized by trade wars and major corporate and/or state negotiations (Jacquemin,1989).Before any level of government undertakes policymaking regarding trade, a solid understandingof trade flows and their effect on both domestic and foreign welfare must first be achieved.Modelling trade then becomes an important first step toward understanding and helps describethe determinant conditions for trade to take place. If economic parameters of trading sectors areunique and separate from those of non-trading industries, then these parameters should be ofutmost significance to policy makers and trading firms alike.It is the economic importance of international trade, particularly export trade, and the role ofgovernment within the activity that provided prime motivation for this academic study. It thenattempts to focus on export trade and describes a simple relation between exports and industrystructure. The first three sections of the paper give a brief review of the theory and literaturerelating to international trade models of imperfect competition. The fourth section emphasizesthese trade models under conditions of economic growth. The following two sections describetrade in imperfect markets using a simple two-country trade model. The model parameters andimplications are explored. The seventh and eighth sections provide empirical evidence linkingindustry structure to export trade in both Canada and the United States. Conclusions from thisanalysis are that imperfect market theories are indeed compatible with empirical findings forthese two countries. Implications of the theoretical and empirical observations for national tradepolicy are given in the ninth section and final remarks are included in the tenth and final section.2ILFINDING A MODEL FOR INTERNATIONAL TRADEThe traditional model of trade within perfectly competitive markets is useful for explaining tradeactivity on the basis of country comparative advantage. The theory of comparative advantage wasfirst described by David Ricardo (1817) and states that producers within each country that arerelatively more efficient than their foreign counterparts at providing a good will export thatcommodity. The assumption that international markets are perfectly competitive places severalconditions on the Ricardian model. Included are that all producers yield sufficiently small outputas to not affect market price, each has perfect market information, firms freely enter and exit,and production takes place with constant or decreasing returns to scale. In reality, a significantamount of global trade takes place within imperfect markets in which oligopoly or even puremonopoly equilibriums persist (Dixit, 1983).Modelling trade based on comparative advantage and perfect competition fails to explain severalempirical observations. Traditional theory may, therefore, no longer be sufficient for explaininginternational trade flows. Specifically, the model cannot account for volume and compositiontrade, role and volume of intrafirm trade, foreign direct investment, and effects on social welfarefrom trade liberalization (Helpman and Krugman, 1985:2).Trade models incorporating economies of scale and imperfect competition allow greaterunderstanding of the empirical observations that makes the conventional trade model weak.Given this additional view of model parameters, increasing use is being made of industrial3organization analyses, imperfect competition and game theory to describe oligopolistic firms andgovernments competing in a strategic trade environment for a share of global economic wealth(Jacquemin, 1989).The problem of modelling imperfect competition is that there does not exist one general modelof oligopoly that can be consistently applied. As firms seek to maximize profits they maycollude, either explicitly through cartels and formal agreements or tacitly through "follow theleader" strategies, or they may forego cooperation and choose to embrace competition of variousdegrees. The conjectured responses of firms will also affect individual firm behaviour as wellas the choice of strategic variable with which to compete (Helpman and Krugman, 1985).Indeed, Jacquemin (1982) points out that there are too many ways in which imperfect marketstructure can be modelled, leading to too many different welfare implications. Although extreme,his thoughts do encompass the daunting scope of possibilities when attempting to predictinteractive firm behaviour and market equilibrium outcomes. Some discussion of relativecompetition within imperfect markets as they relate to trade may, therefore, be warranted.Cooperation in OligopoliesCooperation within oligopolistic industries can be described in one manner by the formation ofcartels. Cartels are frequently encountered in international trade and cover both primary goodssuch as rubber, sugar, bauxite, petroleum, and cement; as well as secondary goods such as airand sea transports, electrical and telephone cables, cinema films and television films. Someeconomists argue that because cartels are inherently unstable, a model of firms maintaining this4collusive form would not be a good one (Brander, 1983). Others suggest that although unstable,the fact is that cartels do exist and, especially export cartels, continue to increase in number(Jacquemin, 1989). Consequently, this form of imperfect market structure cannot be ignored inexploring the behaviour of firms within imperfect markets.Non-cooperation in OligopoliesOligopolistic competition may be strong such that monopolistic distortions are removed and priceand output levels result at or near to the perfectly competitive equilibrium. Contestable marketsmay provide one theory of this type of firm behaviour. The theory predicts that where barriersto entry currently exist, the threat of potential entry or other means of increased output bycompetitors reduces monopoly distortions and forces average cost pricing (Helpman andKrugman, 1985).Monopolistic Competition Proposed imperfect market models often presume industry production of a homogenous good orsome output that is perfectly substitutable between firms. Monopolistic competition replaces thispresumption with one of product heterogeneity. The theory argues that firms within an industrycan differentiate their products such that it is no longer a perfect substitute for existing orpotential competitors' goods. With a differentiated and therefore unique good, a firm faces adownward sloping demand curve for its product and can exert market power using price oroutput variables.5Helpman and Krugman (1985) have developed a trade theory based on monopolistic competition.Following Chamberlin (1933), Helpman and Krugman argue that in an industry facing increasingreturns to scale from specialization, or where each of a variety of industry goods is producedunder decreasing average costs, firms operating in different countries within a global industrywill end up producing a different type of the good. Firms maximize profits by choosing toproduce a specific product variety at a certain price. The existence of heterogeneous industryproduct is predicated by the assumption that diversity is demanded by all consumers. Givendecreasing cost from specialization and a preference for variety, intraindustry trade will occurbetween countries. Each country will export the good in which it's firm has specialized inproduction and import the remaining industry products demanded. The monopolistic ordifferentiated product approach can be taken further to explain not only world trade patterns butalso volume and composition of trade and welfare effects from trade (Helpman and Krugman,1985).Cournot Model of Competition A Cournot model of imperfect markets is an intuitively attractive model since equilibrium outputis measured somewhere between monopoly and perfectly competitive levels. The Cournotsolution is a Nash equilibrium in outputs. Each firm responds to a best estimate output level ofits rival(s) when choosing their own output level for maximum profit. Deriving the Cournotequilibrium offers a reflection of a rational process undergone by firms who use all relevantinformation in assessing their own and competitors' production activity.6Application of this relation has mathematically simple properties since adversary output is takenas exogenous or fixed, however, the model does contains drawbacks despite its simplicity. Inreality, firms are not restricted to competing on the basis of output quantity and may usedifferent competitive strategies incorporating price or quality adjustments that affect output and,consequently, overall market price.Determining an empirically supported and precise model for international trade and imperfectcompetition may remain a research task for many years. Despite its limitations, however, theCournot model remains the most widely used for describing firm behaviour in imperfect markets(Brander, 1986).The application of imperfect market models has taken place within empirical studies measuringvarious aspects of international trade. Imperfect market models free the investigator of relyingon traditional sources of comparative advantage and factor endowments as assumed determinantsof trade. Using imperfect market parameters, patterns of international trade activity have beenexplained by Saxonhouse (1989) and Fung (1991)); trade policy outcomes by Brown and Stern(1989), Aw (1991), Roberts and Tybout (1991) and Fung (1991), and Schembri (1989) hasemployed an imperfect market model to illustrate international pricing behaviour of firms.The above discussion certainly does not cover the entire vector of possible market structures andcompetitive relations under the rubric of imperfect competition. However, it does identifyvarious market possibilities within a range of competitive equilibriums extending from monopoly7outcomes to levels closer to that of perfect competition. It is therefore possible to conjecture thatthe scope of the set of possible international trade policy alternatives will be similarly broad andaccommodating to the characteristics of imperfect markets. Even if this is unrealistic,government should at least pay attention to the behaviour of firms within markets in order todevise reasonable trade policy and gain some accuracy in predicting outcomes fromimplementation.IILINIPERFECT MARKETS AND GOVERNMENT TRADE POLICYGovernment intervention in trade, or "strategic" trade policy is opposed, on the whole, byprofessional economists (Brander, 1985). As Brander points out, this may be due to thetraditional view of trade theory that supports international trade and investment as determinedby comparative advantage, rather than the intercedence of government. However, by examiningtrade under imperfect markets, argument for interventionist trade policy can be made.Trade in imperfect markets may give indication that economic rents are being earned.Government intervention is one means of channelling those rents to components of the nationaleconomy. Typically, policy relies on the use of price or output instruments such as taxes, tariffs,quotas, export subsidies or production subsidies. Mandated firm behaviour as reflected incompetition (Canada) or anti-trust (United States) policy may also be included in the list ofpolicy means available to government. The instrument chosen will depend on a variety ofimperfect market factors including barriers to entry, economies of scale, imperfect information8and heterogeneity of product such that trade intervention arguments may be justified solely ona case by case basis. Common to each of these cases and interventionist tactics is the nationalpurpose of shifting economic rents from foreign rivals to domestic constituents.The opportunity for rent-shifting trade policy does not imply that government intervention isrequired for nations to experience gains from trade. Non-interventionist policies may be sociallyoptimal over interventionist ones within imperfect markets. The presence of imperfect marketsalone is not a sufficient condition for government intervention. Before continuing discussion ofgovernment trade policy effecting rent-shifting, perusal of some of the theories supporting freetrade in imperfect markets may provide some balance to the rationale for strategic intervention.Gains from Free Trade in Imperfect MarketsUninhibited trade provides domestic exposure to international competition that in turn tempersthe inefficiencies of imperfectly competitive domestic markets (Kruegar, 1978). FollowingKruegar's argument, Brander (1981) develops a two-country, two-firm model that shows tradewill take place within the oligopolistic market as long as price markups over marginal costexceed transportation costs. The international rivalry between oligopolist firms will act as acause of two-way trade. Competitive rivalry produces more efficient equilibrium price andoutput levels than under autarky.Brander's model is expanded under Brander and Krugman (1983) to consider the same duopolist,two-country world operating under conditions of imperfect competition. Specifically, firms are9able to price discriminate between the domestic and foreign market and make price and outputdecisions according to profit maximization requirements. Each country's monopolist has positiveoutput levels within each of the domestic and foreign market at the standard monopolist solutionof marginal revenue (MR) = marginal cost (MC), although foreign market output decisions aremade under perceived greater MR and higher costs due to transportation. Consequently, eachfirm's price markup over cost for the export market is lower than that of the domestic andreciprocal "dumping" occurs. Again, global gains from trade result since competition forceslower prices and greater output in each market than what would have occurred under autarky.Resultant increased market competition supports the case for non-interventionist trade policy.In imperfect markets with increasing returns to scale, potential gains from trade stem fromrationalization within the industry where export output has cost-reduction consequences. Graham(1923) suggested that to ensure rationalization gains are maintained within the country is toensure that expansion of production occurs within that country. Otherwise, increased productionoccurring outside the nation could feasibly cause relative domestic cost increases, national firmswould not be cost competitive with foreign rivals whose expanded operation caused decreasedaverage cost. Although Graham's suggestion of trade intervention in order to protect the homemarket and induce domestic gains (or at least prohibit domestic producer losses) is intuitivelyappealing, trade intervention may not be a necessary condition.Under conditions of free trade, Krugman (1992) employs a monopolistic competition model todemonstrate that there may be both domestic and global gains from trade. Benefits result from10both rationalization and world diversification of goods. Consumer surplus is realized due togreater product choice. Even if scale economies are not realized (but are not negative), consumersurplus gains alone, if they are realized in the domestic market, are a sufficient argument fornon-interventionist trade policy. Under monopolistic competition, Graham's concern for domesticmarket protection is tempered.Krugman expands his argument to explain that under scale economies and productdifferentiation, production within the home nation will take place near the largest foreign marketin order to minimize transportation costs. If domestic markets are large and there are sufficientlydissimilar tastes between countries then domestic producers will specialize in the home marketpreferred good and become a net exporter of that good. Patterns of net export trade maytherefore follow home market preferences (Krugman, 1992).Gains from Trade Intervention in Imperfect MarketsWhile laissez-faire strategies can induce gains from trade under imperfect market conditions,opportunity for governments to intervene and extract global industry rents may also be present.Following are descriptions of models of trade that yield such opportunity. The cases provideexamples of the variety of circumstances under which interventionist policies can yield optimalwelfare benefits.Krugman (1992b) builds a case for import protection based on rationalization gains only. Theprincipal of Graham's argument is somewhat reiterated here. Focusing on a duopoly industry11facing declining average costs and consisting of a domestic and foreign firm, Krugman arguesthat a protectionist import policy can shift domestic sales to the home firm and away from theforeign firm. Costs of the home firm should decrease and those of the foreign firm increase.Output to the foreign market supplied by the domestic firm will also increase and market shareof the rival will decrease due to lower relative production costs effected in the protecteddomestic market. Krugman's argument is a circular cause and effect of increasing economiescausing increasing output causing further reductions in output costs. Krugman does not provideany welfare analysis. Welfare analysis may indeed have revealed consumer surplus losses fromhigh domestic prices as a result of import protection. Losses could conceivably be large enoughthat a national negative impact from the protectionist policy results. Krugman does not,therefore, generalize the import protection approach as an applicable strategy for all industries.Rather, the argument demonstrates that such import protection policy can at least be modelledas a first step to evaluating protectionist trade policy and welfare effects under imperfect marketconditions.Aw (1991) provides empirical investigations of import protection under imperfect marketconditions. Voluntary export restraints (VERs) placed on Taiwanese and Korean footwearmanufacturers were investigated to determine if any price affects resulted within the UnitedStates (US) domestic market. Aw highlights the Krugman (1992b) caveat that domestic marketprotection within an imperfect market may harm the consumer through raised industry prices.Empirical results of Aw show that neither the foreign importers nor domestic industry engagedin non-competitive pricing behaviour. However, the reduction in foreign supply brought on by12the VERs did raise domestic industry prices above those of the period when VERs were not inplace. This price-cost margin increase was the result of market scarcity brought on by the VERs.Although no welfare analysis was explicitly undertaken, the price increase and quantity decreasewould indicate that there was some consumer surplus loss on the part of the US footwearconsumers.Brander and Spencer (1984) focus attention on the use protectionist policies; specifically importtariffs, subsidies or cartelized domestic export industries; as methods of extracting rents fromimperfectly competitive international markets. The optimal policy tools recommended willdepend on cost and demand circumstances. Import tariffs are predicted to shift rents fromimperfect foreign imports in the domestic market to the national treasury. These rents can belarge enough to offset reduced consumer surplus from lower domestic supply. Domestic exportcartels may also be used to extract these industry rents from foreign competitors. A significantpoint of the Brander and Spencer paper is that economic rents provide strong incentive fornational governments to intervene in trade and act unilaterally to secure national social benefits.Eaton and Grossman (1986) use similar argument to Brander and Spencer above but isolate thecase of non-domestic consumption within imperfect markets. Both trade policy and industrialpolicy are demonstrated to effect rent-shifting. The authors state that domestic competitionamongst home producers has a negative effect on national welfare since opportunity for thenational industry to move closer to monopolistic equilibrium levels is foregone. Either aproduction or export tax decreases domestic industry output such that the home-country exploits13monopoly power in trade. Note here that a policy allowing domestic industry firms to form acartel would also achieve the desired national welfare results.Richardson (1990) supports a case for exports subsidies to encourage trade by influencingdomestic market structure. An incumbent monopolist in the domestic market is assumed to makeexpenditures that deter the entry of rival firms. The monopolist exports only surplus productionand makes pricing and output decisions based on the domestic market alone. Under suchcircumstances, export subsidies or promotion may well be welfare improving. In effect, theyincrease the good's world price as realized by the monopolist and increase the export salesmargin. By altering the relative profitability of export markets, the monopolist may choose tocompete as a duopolist in the domestic market, conceding domestic market share and price forgains in export activity. Domestic consumer gains are realized through increased competition andproducer gains are realized by reduced cost from entry-prevention expenditures no longerincurred. Although fairly specific, the Richardson argument places important focus on thepossible inter-relation of trade policy and domestic market activity and structure.Eaton and Grossman (1986) use their previous argument for intervention by way of trade andindustrial policy to warn against the complications of domestic industry structure on policyresults. Endogenous entry and exit of firms is assumed along with the previous postulate of nodomestic consumption. If a trade or industrial policy causes profit-shifting to the domesticindustry and attracts additional domestic firm entrants, rents could be dissipated throughincreased competition that negates any national benefit. Economic rents may also be lost to14payment of additional entry fees or the increase of firms could raise the industry average costof production. Under these conditions, incremental policy deterring firm entry is required, eitherthrough export or production taxes. The authors draw opposite conclusions if there existsresidual demand within the industry. As such, the promotion of domestic firm entry, say throughexport or production subsidies, deters foreign firm entry or causes their exit and, therefore,industry benefits accrue to the greater number of national producers. The important point of thecase presented by Eaton and Grossman is that industry structure variables must be consideredbefore rent-shifting or other trade policies can be implemented. These variables could causeoptimal policies to vary between tax and subsidy under similar trading conditions.Brander and Spencer (1985) provide an interesting rationale for rent-shifting policy throughproduction subsidies under the rubric of international strategy. If a potential international marketis efficiently served by only one firm, then monopoly rents from that industry will be sought bynational governments as well as the potential domestic entrant. Government should be willingto provide a subsidy to the firm up to the amount of monopoly rents available within the industryif the subsidy guarantees domestic firm victory in securing the monopoly position. Eachdomestic firm, before any subsidy, faces potential losses from entry if the market is eventuallylost to the foreign rival. If the domestic government were to subsidize potential entry losses thenthe foreign firm would face the greater risk of market entry and may not pursue an entrystrategy. The credibility of governments providing subsidies within international markets mayconsequently deter foreign competition and yield monopoly rents to the domestic firm andnation.15Further ConsiderationsEach of the trade scenarios reviewed above gives possible rationale for laissez-faire and/or tradeinterventionist strategies. Some concerns regarding trade policy and imperfect markets, however,remain. The first consideration of the models proposed was initially raised by Dixit (1983). Ifcapital markets are efficient and shareholders are international (ie. there exists a distribution ofshareholders across countries) then rent-shifting policies would be pointless if wealth wasdistributed to shareholders. Producer rents paid to the firms owners in the form of dividendswould dissipate outside of national boundaries and the original government objectives of countrywealth maximization would be thwarted. If, however, shareholders are predominantly domesticthen the accrued industry rents should remain within the country and wealth maximization goalsshould be achieved.A second consideration is that of policy retaliation. Of the rent-shifting models described, noneconsider retaliation by foreign governments. Each assumes that the foreign government continuesto pursue a laissez-faire policy strategy while the domestic government applies interventioniststrategies. From the global perspective, rent shifting should have no impact as long as outputand price levels are unaffected overall, ie. that the domestic social gain equals foreign sociallosses. Many of the strategies discussed, however, create monopoly distortions in markets forthe benefit of domestic constituents. From a global perspective, if the loss of the foreign nationis not offset by equal or greater gains within the domestic, global welfare is reduced. If bothcountries pursue rent-shifting strategies, then both nations may be worse off than if each wereto follow a laissez-faire strategy. Unilateral rent-seeking by nations is cause for global concern16over optimal world trade policy. Such actions give multinational institutions such as the GeneralAgreement on Tariffs and Trade (GATT) a legitimate role in tempering the intervention ofgovernments within international imperfect markets.A third concern relates to the variety of specific models explaining international trade strategiesin imperfect markets. Seemingly the number of scenarios justifying policy intervention ispotentially as great as the modeller's imagination in describing a number of combinations andpermutations of industry, market and trade parameters. Helpman and Krugman (1985) are moreexplicit in their criticism and state that the body of international trade theory in general has lefta proliferation of models each with its particular assumptions whose implications for trade policyare often inconsistent. This result will be emphasized further by work of additional authorsdescribed below.Fourth, the models do not consider the distributional effects and subsequent political impact oftrade policies as they influence various sectors or groups within the national economy. Baldwin(1989), for example, discusses economic self-interest as an explanation for general trade policyor even industrial trade policy when groups support politicians whose proposed trade policy shifta greater proportion of rents to them. Another distributional effect is the cost of redistributingwealth between the national winners and losers such that a potentially pareto improving tradepolicy can be truly pareto efficient.Finally, net effects of producer and consumer welfare changes resultant of trade policies must17be considered. Trade policies directed at national producers become viable only if producersurplus gains exceed any consumer surplus losses, or net social benefit results. From a cost-benefit analysis perspective, the assumption of net welfare gains is usually sufficient.Realistically, policymakers may not weigh producer and consumer welfare equally, especiallyif costs or benefits are concentrated within a domestic constituent group, making optimal tradepolicy difficult to implement.Discussion of these more complex interactions is beyond the scope of this paper and may bebetter served if addressed as a refinement of trade policy once general effects have been isolated.General effects may also be difficult to determine, however, as outlined below.Trade Policy in Imperfect Markets: Conflicting ResultsMarkusen and Venables (1988) complete a theoretical analysis under fairly strict assumptionsthat test trade policy effects under opposing assumptions of industry structure (oligopoly versusfree entry) and market structure (segmented markets versus an integrated world market). Mixingand matching the opposing assumptions yields a quadrant of possible industry/marketcombinations (actually five propositions were made within the paper) in each of which importtariffs and export taxes are evaluated. We would expect each policy instrument to have a similareffect on domestic wellbeing. Results show, however, that within each quadrant one policy hasno or an ambiguous effect and the other yields conflicting results for total domestic welfare.Results under each of the quadrants generated are not necessarily robust and general results formarket structure are made independent of industry structure considerations and vice versa.18However, one purpose of drawing attention to the Markusen and Venables (1988) work is thatit highlights the models' dependence on specific parameter assumptions. Changing thoseassumptions can lead to conflicting policy recommendations. The work is also useful as a meansof integrating and organizing thinking about market and industry variables in light of policyevaluation.Eaton and Grossman (1986) also conduct theoretical analysis to determine optimal trade policyin oligopolistic international markets. Their results vary between a tax, subsidy or free tradedepending on the model of strategic behaviour of firms chosen. The models employed toevaluate the various policies assumed Nash-Cournot, Bertrand or consistent conjecturalbehaviour.The significance of models and model assumptions is concluded to be non-trivial whenevaluating trade policy. The lack of a robust model makes determination of government policyeffects difficult since model choice may lead to conflicting results. This difficulty underlines theimportance of understanding the market structure, industry structure and behaviour of competingfirms before trade and trade policy can be effectively illustrated.IV:TRADE IN IMPERFECT MARKETS AND GROWTHNational welfare and social benefits from trade cannot be fully discussed simply bydemonstrating gains through static, one-time increases in country wealth. Continual wealth is19also a goal of national government policy and can be achieved through trade under imperfectmarket conditions.During the five year period ending in 1989, the volume of world trade experienced a substantialincrease (United Nations Centre on Transnational Corporations, 1991). Grossman and Helpman(1991) suggest that a linkage exists between the recent historic trends of growth in globaleconomic well-being and increased contributions of technical innovation. The link between worldgrowth and technology can be explained on the grounds of rapid communication betweeninnovators in different countries facilitating the spread of new ideas. These same authors identifya number of studies that name factors correlated to growth in national output. Many are relatedto a nation's collective human knowledge such as literacy rates, the number of science andengineers in society and school enrolment rates. Others relate to trade factors, specifically,Michaely (1977) and Feder (1982) find that growth rates are greater for nations that export alarge share of their output.Traditional growth theory describes the formation of "capital" in the economy which inducesrises in labour productivity and is the major cause of economic growth (Grossman and Helpman,1991). The theory narrowly refers to capital as machinery and equipment of productiontechnology that yields constant returns to scale. Capital, however, can also be interpreted torepresent knowledge accumulation. Knowledge yields technological progress and innovation thatenables capital formation to continue and economic growth to take place. The studies notedabove support the notion that knowledge and technology can each be interpreted as growth20producing "capital". Traditional growth models have historically treated technological growthdeterminants as exogenous, or even explained national growth through technological "spillovers"from one industry or sector to another.Traditional trade models assuming perfect competition cannot very easily handle large investmentin physical capital, human capital or technology through research and development (R&D).Trade models of imperfect competition, however, can be used to model these activities. Forexample, Grossman and Helpman (1991) refer to the Helpman and Krugman (1985) trade modelof product specialization and monopolistic competition to explain the linkage between trade inimperfect markets and domestic growth. In the trade model, fixed production costs incurred arepart of total production costs. Applied as a growth model, fixed costs can take the form of up-front investment in R&D. The country making that particular expenditure in R&D will createand export unique goods resulting from the research outlay and import goods from anothercountry making different R&D expenditures resulting in different goods produced. Thespecialization assumption also allows gains from trade to be realized due to increased efficiency,specifically, lack of duplication of innovative activities results.Many of the preceding arguments for both non-interventionist and interventionist trade strategiesdescribe benefits from trade due to directed expansion of the domestic industry and the resultingone-time productivity gains from rationalization. These economies of scale specifically describeeconomies in production. At the firm level, opportunity to expand wealth through trade is limitedsince we would expect such gains to soon be exhausted in light of a large global market (Dixit,211983). Krugman (1992) moves beyond production scale using his argument of import protectionas a means of securing domestic industry growth in output. Cost decreases due to expandedoutput at the expense of reduced foreign firm sales provide incentives for R&D or growthproducing expenditures. Greater R&D investments made by the domestic firm than the foreignrival can be amortized over higher output in both home and foreign markets. The same argumentcan be applied to industries characterized by some learning curve in production, where highercumulative output influences greater current productivity. Greater home than foreign firm salesachieved today yields greater productivity gains and growth of domestic producers. Krugman,therefore, shows that a protected domestic market may encourage welfare as well as growthopportunities.Trade itself may cause economic growth. By exchanging product, equipment or information theproducer may deliberately (or not) share innovative ideas and speed the process of informationexchange leading to higher industry growth rates. In the short term, possession of technologyand information may give firms some monopolistic advantage in the global market that leads tohigher levels of trade.The motivation of monopolistic trade is the subject of much of the current dispute over theinternational disregard for patent laws. When laws are violated, future trade can possibly bedeterred since the innovator risks losses of the proprietary right to the economic rents earnedthrough trade. Effects on further investment in R&D may be ambiguous, depending on manyfactors including the risk aversion of the innovator. Katz (1986) confirms that since the22externality of technology spillovers confers benefits on competitors, firms may have a tendencyto underinvest in R&D. The easier the dissemination of information, the greater the spilloversare likely and the larger, we would expect, the underinvestment in innovation (Baumol, 1992)and, therefore, economic growth.Baumol (1992) makes the argument that technology cartels and collusion in R&D offersincreased rents by reducing the positive externality of technology spillovers and reducing thedissemination of rents from innovation. If innovative technology is produced in internationalmarkets then spillovers may provide positive externalities to foreign competitors and opportunityto earn industry rents. Policy that encourages formation of domestic technology cartels andcollusion as described by Baumol may exemplify a form of trade intervention that shifts ormaintains rents within the domestic industry.Caves and Khalilazdeh-Shirazi (1977) reveal a number of empirical linkages between firminnovation and trading activity. Results showed the following: that proprietary information wasfound to be important to a larger proportion of exporting than import-competing businesses; thatone third more of exporters' sales were accounted for by new products introduced in the threeyears preceding the study; and, that R&D outlays, as a percentage of sales, were larger forexporting businesses. Another conclusion drawn by the authors is that trade-related business(both exporting and import-competing) are more innovative than producers not exposed to trade.Trade may therefore provide stimulation for growth through innovation, and exporting industriesin particular may be an engine for national growth through innovation.23Based on the previous theoretical and empirical literature, strong argument exists that trade maylead to increased investment and innovation that generates dynamic welfare benefits enjoyedfrom either a national or global standing. Linking the areas of trade and growth, Helpman andGrossman (1991) argue that dynamic models of growth can be made analogous to static trademodels using imperfect market assumptions, specifically with reference to trade models ofproduct specialization explaining patterns of intra-industry trade. Since externalities areseemingly an inherent attribute of innovative activities, trade within innovative industries maybe characterized as imperfect. Therefore, argument for some form of rent-shifting trade orindustrial policy exists if "spillovers" to foreign producers are to be minimized and industryrents are to be maintained by domestic firms.Market imperfections are significant variables in assessing trade elements of national welfare,growth and government policy. Industry structure is a factor of market imperfections and is alsoa critical element of model assumptions predicting and evaluating trade. The remainder of thispaper will focus on industry and market structure variables that influence trade and trade policy.V.ANALYZING IMPERFECT MARKETS AND TRADE: INTRODUCTION TO THEBASIC MODELA model of oligopolistic international trade was designed in order to better understand trade inimperfect markets and the influence of market structure on trade. The purpose of the model isto simplify conditions under which trade takes place in order to clearly illustrate some of the24industry structure influences on trade.The basic model consists of trade between two countries in each of which firms produce ahomogeneous good. Consumption takes place in the two markets, one referred to as "domestic"and the other as "foreign", such that markets clear. The domestic and foreign country togethercomprise the global industry. The model describes one period of output and consumption.There exists n domestic firms and m foreign firms within the respective countries. Each firm hasa Cournot perception and seeks to maximize profit within the model by taking rival firm outputas fixed for the period. A domestic firm produces output x for domestic consumption and output* ^y*x for delivery to the foreign market. Similarly, y and y are produced by the foreign firm fordomestic and foreign market consumption respectively. Consumers in both markets haveidentical preferences and tastes represented by identical linear inverse demand curves. Marginalcost, c, is at first assumed to be constant for all firms in the global industry. Transport costs areincurred in exporting the good from one country to the other and are identical for all firms inthe global industry. Transport costs are applied within the model in the form of a "tax" onoutput where t O.Caveats of the basic modelAssuming a one period model:The analysis looks at a one period, static model without considering reactionary effects from pastor future output by own or rival firms. If the model was more dynamic, expected actions of25competitors in one period would affect firm and therefore industry output levels in subsequentperiods. Such a model could conceivably be built starting with a Cournot perception written ingeneral functional form for a given firm, i, in some single period, n, as;=^- C i(xi),where:Jr = output function of firm i in the current period nxi = output of firm i for the period nxi,m and xk,. = output of firm i,k for all previous periods mxk = output of all other firms k where k i and where xi + xk = total industry outputfor the period nP(xi,k) = inverse demand function for the periodFurther conjecture could also be incorporated into the model in order to generate an equilibriumCournot solution over a number of periods. Without continuing to speculate on the dynamicpossibilities, the point to be offered is that dynamic models offer much more comprehensivebehavioral information not included in the basic model described l . The basic model sacrificesthese considerations in order to generate simple solutions that clearly demonstrate a specifiedrelation.The Cournot model, although a static one-period model, is rational and yields an equilibrium thatresults from each firm's best response to it's rivals' actions considering all information. It may1 see Axelrod (1984) for work on dynamic models of interactivefirm behaviour.26therefore serve as a close approximation for more dynamic industry behaviour while still servingthe purpose of allowing analysis to focus on specific variables in the environment.Assumption of a homogeneous product:Oligopolistic industries may tend to offer differentiated products in order to capitalize onmonopolistic market opportunities or increasing returns from specialization, testing thereasonableness of the assumption of homogeneous product within the basic model. It is notunrealistic, however, to observe industries where a homogenous product is produced underconditions of increasing returns to scale. Primary goods or commodities such as steel and oilmay be produced within industries having barriers to entry due to large initial capital investment,geographic restrictions, or some proprietary process technology. Therefore, gains fromrationalization may be available to such extent that oligopolies do tend to arise in homogeneousproduct industries and the model's assumption of homogenous product may be considered viable.Additional Caveats:In addition to the homogeneous and static period assumptions discussed, further weakness of themodel must be identified. First, the model does employ general functional forms and dependson specific linear functions to describe demand and firm output. Second, trade does not considerthe effects of substitute or complement goods as prices and output levels change within themarket. Third, consumers are indifferent to aspects of the good other than price. Country oforigin, quality and other effects on the purchase decision are ignored. This same argument is thebasis of the final caveat. The output decision is the only strategic variable employed within the27model and quality, pricing and other effects indirectly affecting output are assumed to remainconstant.VI. ANALYSIS OF THE BASIC MODELThe general profit maximizing conditions under which each domestic firm, i, and foreign firm,j, operates can be written as:lIi = x.13(X) + x*P(Y) - c(x+x*(1+t)) - F, and^ (1)yll. = * 43(X) + y-13(Y) - c(y +y*(1 +t)) - F, (2)where asterisks denote production for export and F represents fixed costs to the firm. Marketprice is dependent on total consumption within the domestic (X) or foreign (Y) market. Thedemand functions for each market are assumed to be linear and identical;P(X) = a - bX (3)where X = A + x i + yi * , and (4)P(Y) = a - by (5)where Y = B + x i* +^and (6)where A =xk + T y*k and B = yk + E x *k .K#Due to the assumption of identical price and cost structures across all firms within the domestic28or foreign industry, each of these industry's firms will have identical output. Therefore each ofx, x* , y and y represents "typical" firm output such that the subscripts i and j are no longerrequired for firm differentiation.Since there are n domestic firms and m foreign firms, and costs within each industry are equal,total output for the domestic market, X, and the foreign market, Y, can be written as:X = nx + my* , and^(7)Y = nx* + my. (8)From equations (7) and (8) we can rewrite equations (4) and (6) as:A = (n-1)x + (m-1)y", and^(9)B = (n-1)x" + (m-1)y.^ (10)Since profit functions for each firm are additive, segregating them into profit from foreign anddomestic activity yields two equations for each firm d and f. The first order conditions of theseprofit functions yields four symmetric equations in a, b, c, t, n, and m for each of the outputvariables x, x* , y and y* . These equations are:29x = a - bmy* - c,^(11)b(1 +n)x = a - bmy - c(1 +t),^ (12)b(1 +n)y = a - bnx* - c, and^ (13)b(1 +m)y* = a - bnx- c(1 +t).^ (14)b(1 +m)Effects of the cost functionSince equations are symmetric, results are equally applicable to either country's firms facingforeign and domestic markets. Using equations (11) and (14) to describe the relative output ofthe typical domestic foreign firm in the the domestic market, X, the relation can be written as:x = ys + ct/b.^ (15)Note that each variable c, t, and b are positive by definition. Equation (15) shows that domesticoutput will be equal to foreign exports to the market only in the absence of transportation costs(t=0). Otherwise, domestic firm output is greater than that of the foreign counterpart for t > O.30In the absence of transportation costs, domestic and foreign firms will share the market equally,each controlling 50%. When t > 0 then domestic firm market share will exceed that of theforeign firm. Therefore, the presence of transport costs creates a wedge between domestic andforeign supply to the domestic market that also increases the market share differential of foreignand domestic firms.The relationship of cost and output has interesting implications for firms in exporting industries.It is recognized that some increase in marginal costs over and above transportation may bedirectly attributable to exporting. These costs can be accommodated by the basic model and mayinclude compliance with regulation (ie. documentation), increased advertising and distributioncosts in the foreign market, higher financing charges or costs from product adaptation for salein the foreign market. By simply re-expressing t as the sum of all incremental costs due toexporting, including transportation, these costs are incorporated. By re-evaluating equation (15),these additional positive costs affecting the exported goods increase the wedge between domesticand foreign firm output.Notice that trade still takes place within the industry even though the cost disadvantage due totransportation or incremental costs exists for foreign firms in the domestic market. Brander(1981) first described this phenomena by showing that within an imperfect market with marketprice exceeding total marginal cost of both firms, including transportation, each firm has a profitmaximizing solution for the foreign market at some positive output level and intra-industry tradewill continue.31In order for firms in exporting industries to maximize their own welfare, producers facing sameproduction costs as their foreign rivals and same transport (and other export marginal) costs astheir domestic competitors must minimized these costs. Only under conditions of identical totalmarginal costs between the rival and foreign industry can the domestic industry move towardsa maximum 50% market share in the foreign market.By relaxing the cost constraint equating domestic and foreign firm production costs, equation(15) can be rewritten using different marginal cost values across the two national industries - c xrepresents a domestic firm's marginal cost and cy represents a foreign firm's marginal cost.Therefore;x r y* + (cy - cx)/b + cyt/b.^ (16)Equation (16) implies that in order for domestic and foreign firm output to be equal, costs ofdomestic production must be equal to total costs of export production, or:(17)Equation (17) shows that any affect of transportation or other incremental export costs on theforeign producer must be offset by lower production costs if equal per firm output and equalmarket share is to be maintained. As costs increase and output falls, demand in the market isincreasingly met by the lower cost/higher output country. Assuming that transport costs and32other incrementals are inevitable, one strategy that firms within an exporting industry maypursue is to offset marginal influences from exporting with lower relative marginal productioncosts. Such a strategy would also lower domestic output costs below those of the rival level.Focus on lower relative production costs, therefore, represents a means for expanding output,market share and profit in both the foreign and domestic market.The same cost-output relationship exists between firms within an industry. Looking solely atdomestic firms' export production, variant costs across domestic firms result in variant firmoutput levels. Output for firms i and k within the domestic industry is represented as follows:x = a - bA - by - ;0+0, and^(18)x*k = a - bA - by - c,k(1 +t).^ (19)Analysis of equations (18) and (19) shows that for output across firms to be equal, firm costsmust be equal or c i = ck . Lower production costs in either firm i or k bear correspondinglyhigher export output of that firm. We note too that firms facing dissimilar transportation costs,perhaps due to some advantage of proximity to the export market, may enjoy greater exportoutput than their domestic firm rivals all other things equal.Not all firms within an industry can be expected to be exactly identical, therefore, industryexports may not be evenly distributed across all firms. The relevance of this point for the33industry will be pursued further under considerations of relative firm number within eachindustry. Indeed, not all firms within the industry may export if costs relative to foreign rivalsare large and exceed foreign market price levels.Note that the analysis so far has only considered marginal cost. Fixed costs of the firm forpurposes of export or production have not been considered to affect the firms' output decisions.Fixed costs present in the model indicate that, within some range, decreasing average productioncosts and increasing returns to scale are present. Consequently, any increase in output wouldyield additional profits to the firm if expanded operations occurred within the declining costrange.Effects of Scale EconomiesThe original model's annual fixed cost component, F, may represent some initial start-up oroperating investment during the period relating to production for both the domestic and exportmarket. As stated previously, this factor implies that average costs of production are falling oversome output range even though marginal costs are constant. If average costs are falling withinthe relevant output range for all firms then industries are subject to economies of scale. Gainsfrom trade including rationalization of output within an industry were discussed in detail withinprevious sections of the paper. It would be useful here, however, to reiterate these argumentsas related to discussion of the basic model and some welfare analysis.Due to fixed costs, output increases within the domestic industry yield total producer surplus34increases. Rent-shifting policies may focus on allowing domestic output to increase at theexpense of foreign rival's such that greater gains from rationalization are realized within thedomestic industry while reduced output in the foreign industry reduces producer surplus fromhigher cost production. Marginal cost differentials may, therefore, also contribute to nationalindustries' realization of economies of scale. Consider the case of the domestic industry. Ifexport marginal costs are minimized and/or marginal production costs differentials betweendomestic and foreign firms offset export marginal costs, then domestic industries may be ableto achieve at least as great or greater market output and market share at the expense of foreignrivals within the foreign market. Any additional output may drive the domestic industryoperations further down its average cost curve than foreign firms with the same productiontechnology and investment. Therefore, marginal cost advantages may serve to reinforce gainsfrom rationalization due to trade.One effect of industry rationalization is to reduce the total number of firms competing and allowindustry concentration to grow. The issue of industry structure as described by firm number isdiscussed in the following section.Effects of firm numberThus far the analysis has assumed equal number of firms n and m within domestic and foreignindustries. The effect of firm number within an industry on domestic or foreign market outputhas not been isolated. By assuming the cost relationship of (15) holds and is strict, ie. x > y * fort > 0, the effects of firm number can be explored.35First let us assume that the number of domestic firms, n, is greater than or equal to the numberof foreign firms, m. Since x is strictly greater than y * then rvx will be strictly greater thanm . The aggregate domestic industry supply to the domestic market will be greater than thatof the aggregate foreign industry. Further, since X = i•x + my * the domestic industry'smarket share will be greater than the foreign industry market share. Assuming X is constant (ie.there are no price effects), any increase in domestic firm market supply must, by definition ofX, come at the expense of the rival foreign industry. Due to the presence of declining averagecosts, a shift in industry rents from foreign firms to domestic would take place.Only by strictly defining relative firm output and directing the relative sizes of n and m can anyclear relation between firm number and industry output be described. If n 5 m given x > y * thenthe lower foreign per firm output may be offset in the industry aggregate by a higher numberof firms such that foreign industry output is equal to or greater than that of the domesticindustry. This aggregate output ambiguity is demonstrated in the matrix shown in Figure 1. Theindustry having the greater share of the domestic market, X, under various assumptions aboutthe relative sizes of output, x and y * , and firm number, n and m, is denoted within the matrixcells.36Figure 1. Greater Industry Market Sharen> m^1 n=m I n <mx> y* domestic domestic ?x=y* domestic equal foreignx<y* ? foreign foreignwhere;domestic^= domestic output has dominant share of marketforeign^= foreign output has dominant share of marketequal = domestic and foreign output has equal share of market= ambiguous outcomeCases where cost effects may be offset in the aggregate by the number of firms in the industryare illustrated in Figures 2 through 5. Total domestic and foreign industry output to the domesticmarket is graphed by plotting individual firm output on the vertical against the rank order offirm size as measured by per firm output from highest to lowest. The total area under each curverepresents some total equilibrium output for the domestic market X. These cases are equallyapplicable for the foreign market Y using output of x * and y. If area under the aggregatedomestic industry curve is greater than that of the foreign then domestic industry market sharewill be greater.If costs are equal across the global industry then relative industry output as shown in Figures2 and 3 are simple to determine. These figures essentially depict the upper-middle and middle-left cells of the matrix of Figure 1. Figures 4 and 5 show how costs variant across the industryskew the output curve to the left and create some downward sloping curve describing aggregate37industry output for the market. The figures depict the ambiguous effects described in the lowerleft and upper right cells of the matrix in Figure 1. Integrating the area under the curves is theonly means of specifically determining relative industry output.Figure 2. Relative Aggregate Industry Output, x=y* , n > mFigure 3. Relative Aggregate Industry Output, x >y* , n=m38Figure 4. Relative Aggregate Industry Output, x,y* , n<mFigure 5. Relative Aggregate Industry Output, x,y * , n>mFirm Number and Industry Seller Concentration Using the cost-output relationships of the original model we can explore the effect of industryconcentration in the domestic market and its effects on trade within the foreign market. Recall39that the domestic market share of the domestic industry is expressed as:nx/X, where X = nx + my*and nx/X + my */X = 1Combining this linear example with the cost relationship of the basic model described in (15),foreign and domestic industries should each enjoy a 50% market share in both their own andforeign markets in the absence of transportation costs, ie. nx/X = my*/X given x =y* . Astransportation costs increase, export output falls and domestic market share increases. Given aconstant number of firms, domestic concentration levels would increase as the same number offirms supply a greater share of the domestic market.However, since the relationship is symmetric, increased transportation costs imply lower foreignmarket share for the higher concentrated domestic industry. High levels of domestic industryconcentration due to high transportation costs would therefore be associated with low levels ofexports since each country's industry supplies the majority of their own markets.Extending the argument from transportation costs to total marginal costs of export output, firmswith lower costs in either the domestic or foreign markets will achieve higher relative marketshare in both markets. Higher domestic industry concentration would therefore by associatedwith higher export levels given a fixed number of firms. The additional market shares come atthe expense of foreign rivals whose market participation decreases with higher total costs.40If a domestic industry is imperfectly competitive such that increasing returns are present, thenrationalization within the domestic industry should yield higher per firm domestic market shareand therefore higher industry concentration. Lower costs for export production yield similarresults. In conclusion, the presence of economies of scale within an industry should yield apositive relationship between domestic industry concentration and export trade. Since the modelassumes that the same number of domestic firms now have a greater global market share,welfare effects from increased producer surplus achieved through relatively greater global outputmust be weighed against any loss in consumer surplus from reduced foreign competition withinthe domestic market that may raise market prices.As concentration increases, benefits from rationalization increase until some socially optimalnumber of firms exist and gains from rationalization are exhausted, ie. the minimum efficientscale for the industry is reached. This scale is achieved considering the global industry, not justthe domestic market, such that rationalization could conceivably result in the existence of onlyone firm that supplies the global industry in order to move as far along the decreasing averagecost curve as possible. This is the case described by Brander and Spencer (1985) where nationalfirms vie to become the global monopolist. The national perspective of gains from trade in thiscase, however, must consider the monopoly distortions within the domestic market and theireffect on national welfare.The pattern of trade can also be explained by rationalization within the basic model. Thedomestic industry should be a net exporter if the extent of rationalization within allows greater41cost reductions than the foreign industry. Given equal market size and firm number beforerationalization, some difference in the extent of rationalization between similar domestic andforeign industries could be caused by unlike competition policy or anti-trust legislation betweennations. In countries with stricter legislation impeding the concentration of industries, thesepolicies may serve to raise industry costs and reduce opportunity to gain domestic producersurplus from both domestic and export markets.Dixit (1983) argues that the importance of economies of scale and scope can be expected todecrease at the level of production technology as trade expands market size. Consideringintangible assets, however, such as knowledge or specific managerial skills, economies of scalein these variables may become increasingly important at the international level (Caves, 1982).If economies of scale are present for such intangible assets then domestic industry rationalizationand consequent concentration would still provide gains from trade. If economies are presentacross national boundaries then rationalization of firms within the global industry as a wholewould provide global gains from international trade. As such, domestic concentration indicescould conceivably increase if some domestic firms exited the industry and others acquiredforeign rivals. These assumptions, however, are specific. Domestic industry concentration maynot necessarily capture the relation between international industry concentration andrationalization from trade. Also, opportunity for national policies effecting rent shifting may notbe present since the loss of producer surplus from domestic firms exiting may be less than theaddition of producer surplus from foreign acquired firms.42If domestic industry concentration is increased by way of fewer foreign imports then higherproducer surplus would still be realized by the domestic firms. Again, price effects caused bygreater market power of fewer total firms competing must be considered in light of consumersurplus changes before any net gains from concentration and trade can be determined.No doubt there also exists cases of the domestic industry experiencing increased export marketshare without any change in domestic firm number and market concentration. These cases mayreflect changes in variables considering the export market only, such as cooperative domesticfirm ventures in order to reduce export associated costs. Similarly, export subsidies providedby government would affect export trade without affecting domestic market structure as long asadditional firms did not enter the industry seeking rents from government policy.If the number of firms increase rather decline or remain constant, domestic industry foreign andhome country market share should also increases, all other things equal. In domestic marketswith increasing costs of production, the increase in number of competitors causing decreasedindustry concentration and rising foreign market share means concentration should be inverselyrelated to export activity.The above analysis demonstrates that industry concentration may be either positively ornegatively related to export trade, depending on the cost structure and entry barriers assumedfor the industry as well as the assumptions of the basic model. Study of empirical evidence mayserve to resolve or at least clarify alternative outcomes. Evidence, however, is also mixed43regarding domestic market structure and trade. Examples of empirical work are outlined below.Empirical Evidence of Concentration and TradeIndustry concentration is often used as an indicator of the magnitude of the domestic industryprice-cost margin or industry levels of profitability. Indices also represent oligopolisticinterdependency as measured by the relative output of firms competing within the industry.Concentration measures give indication of both market and industry structure.The discussion so far has focused on international markets and their oligopolistic structure yetif domestic market structure is in some way related to international trade then the importanceof domestic concentration in trading industries is emphasized and may help explain both tradeactivity and patterns of trade.Basically, three linkages between trade movement and industry concentration are cited. As notedabove, one means of increasing domestic industry concentration is through the contraction in thenumber of firms given a level of demand. One cause of contraction could be the exit or mergerof firms due to industry rationalization. Another could be caused by barriers to entry where thelack of increase in firm number given increasing demand would results in high industryconcentration. If export trade takes place under conditions of increasing returns or high barriersto entry, then we would expect to find a positive relation between exports and domestic industryconcentration. If trade instead takes place under conditions of increased competition within theglobal market from a greater number of firms, either at home or abroad, then export activity44should be negatively related to domestic industry concentration. Finally, trade may be motivatedby increased rents. If firms within the industry can price discriminate between foreign anddomestic markets such that rents are earned in both destinations, a concentrated or oligopolisticindustry would be able to earn increased rents from export. In this case, export activity andindustry profitability, indicated by industry concentration, should also be positively related.Research relating to each of these three hypotheses is discussed below.Concentration and Economies of Scale:Bernstein and Mohnen (1991) apply a model of industry concentration and export production tothree Canadian manufacturing industries and show that where economies of scale exist, exportingyields firm benefits. Welfare improvements realized by the firms studied accrued from lowercosts since cost complementarities existed between domestic and export output. Also, Canadianindustries studied exhibited higher price-cost margins in the domestic market than foreignmarkets. This may have indicated that the Canadian industries studied faced flattened foreignmarket curves due to greater international competition, while greater market power was enjoyedwithin domestic markets. The existence of rationalization gains from trade lends credence to theargument that domestic concentration is positively linked to international trade. Existence offoreign and domestic price-cost differentials offers indirect proof that oligopolies can pricediscriminate and earn additional rents in foreign markets and are thereby motivated to trade.Caves and Khalilzadeh-Shirazi (1977) predicted that domestic concentration of exportingindustries would be higher than other industries. The authors base their hypothesis on results45characterizing exporting sectors by higher levels of capital intensity and labour skill than non-trading industries. These characteristics would dispose exporting firms toward greater firm size.Larger firms within industries are correlated to higher concentrated industries given a level ofdemand. Firm size considerations include not only scale economies in asset levels but may alsoinclude costs of information, advertising and distribution, and risk-bearing specific to exporting.Empirical research supports the hypothesis that relative exports and firm size are positivelyrelated (see Bain (1966), Saving (1961), and Eastman and Stykolt (1967)). Evidence also existsfor various countries that exports increase more than proportionately with firm size indicatingfurther economies of scale realized through trade (see Rapp (1976), Glesjer (1980) andAuquier (1980)). Finally, Caves and Kahlilzadeh-Shirazi (1977) show that export orientedindustries in the US have higher barriers to entry that non-export oriented industries, providingfurther indirect evidence that domestic market structure is linked to international trade,specifically export trade.Caves (1981) tested a variety of variables affecting intra-industry bilateral trade flows. Relatingto concentration, effects of scale economies and joint marketing functions (arguably this is anindirect means of expressing "fewer firms" since cooperative firm behaviour can rationalizeoutput) were analyzed for their effects on international trade patterns between the US and othereconomies. Caves hypothesized and showed that the more extensive scale economies, the lesswas the amount of intra-industry trade. This result was achieved since extensive scale confinedproduction of a good to a few locations, limiting the amount of inter-country trade. We wouldexpect the volume of export trade, however, to be significant. Increasing exports as scale effects46within the home country firm are realized results in benefits to the producer nation.Another interesting result of the Caves study was the statistically significant positive effect ofjointness of production on intra-industry trade. This was effected mainly by affiliated companiesoperating multinationally and sharing integrated distribution and marketing systems. Trade flowswould, therefore, include goods for resale to fill out the affiliate's product line.Finally, Robert and Tybout (1991) dispute the positive effects of trade liberalization on industryefficiency with research of trade impacts on manufacturing plants in Columbia and Chile.Although theory would predict that increased import and export opportunities would bring aboutindustry efficiency in the form of fewer plants operating lower down their declining average costcurves, in the Robert and Tybout study plant size (firm size) actually decreased due to greaterimport competition.Concentration and Industry Competition:Caves (1974) argues that a domestic oligopolist who cannot differentiate between domestic andexport markets and faces increasing marginal costs will be constrained by the export market tooutput and pricing levels closer to those of perfect competition. Increased competition is broughton by international trade and the greater number of firms in both domestic and foreign markets.As such, profits generated by the global industry should be driven closer to zero. If profitmargins move to zero, industry concentration as a measure of price-cost differentials should alsodecrease. For this reason, Pagoulatos and Sorensen (1976), in their study of European Common47Market countries, predicted that trade in domestic oligopolistic industries would show a negativerelation to industry profitability. On the other hand, if oligopolistic firms can differentiate theirproduct on world markets then rents captured by the firms would be indicated by a positiverelation between concentration and trade activity. From the latter argument we would expect tosee a positive relation between industry concentration and export activity. Pagoulatos andSorensen (1976) tested concentration and trade measures and found that exports as a percentageof shipments was negatively (but insignificantly) related to firm price-cost margins in eight ofthe ten European countries studied. These results support the first hypothesis of Caves thatinternational competition drives industry rents to zero due to flattened international demandcurves.However, Caves and Kahlilzadeh-Shirazi (1977) show that in the UK, exports as a percentageof industry output were found to have a positive relation to the industry price-cost margins, acounter result to the Pagoulatos and Sorensen information. The UK results lend support to thehypothesis that oligopolistic exporters can earn industry rents with the ability to marketdiscriminate.Jacquemin and Cardon deLichtbuer (1973) showed that industry concentration increasedfollowing the increased exporting opportunities provided by the formation of the EuropeanEconomic Community (EEC). The difference in these results may be due to the differentmeasures employed from the Pagoulatos and Sorensen study.48In summary, industry concentration as a measure of domestic industry structure may be anindicator of international trade activity. Domestic concentration may be positively linked tointernational trade under conditions of increasing returns to various scales of assets, informationor other trading costs, although evidence exists to show this result is not robust. Pricediscrimination abilities of domestic oligopolies in foreign markets may also positively linkconcentrated domestic structure and trade. Concentration may be negatively related to trade ifprice discrimination is not possible and international market demand is flatter than domesticdemand under autarky. The link of profit margin reduction and trade is made using industryconcentration as a proxy measure only. No change in domestic industry firm number needactually take place. However, concentration levels, therefore industry structure, would be dilutedfrom increased foreign competition.In order to focus specifically on the link between industry structure and trade activity, anexamination of Canadian and US industry structure and trade data was undertaken. The studyis modest yet specific in its objectives. It attempts only to determine if a relation betweendomestic industry structure and export trade exists, and if so, the directional characteristic of therelation. Since only exports are studied under the rubric of trade, results offer only partialexplanation of trading patterns.49VII.EMPIRICAL STUDYStatistical analysis was used to determine if aspects of industry structure explain export trade.It is recognized that other variables may determine the pattern and volumes of internationaltrade, however, these aspects were not covered within the analysis. Of particular note, thetraditional explanation of international trade based on the relative factor endownments of nationsis not explicitly included within any testing. Instead, any effect of factor endownments or othervariables is assumed to show up in the error term of regression models tested. The followingempirical analysis abstracts from the traditional theory to focus only on parameters describingthe industry in order to determine if these variables explain export trade.All statistical testing completed and described within the paper was generated using MinitabStatistical Software (1991). All regression analysis estimates model intercept and coefficientparameters using ordinary least squares (OLS) regression techniques.A. CANADIAN INDUSTRY EXPORTS, INDUSTRY FIRM NUMBER, ASSETS ANDAVERAGE FIRM SIZEData SelectionThe unit of measurement for the analysis is one manufacturing industry as classified by theStandard Industrial Classifications (SIC) code of Statistics Canada. Industry units are based oneither two or three digit SIC classifications, since data availability was a limiting factor of the50study. Export measures are not recorded by Statistics Canada under SIC codes but rather usingthe Harmonized System of classification. Exports are therefore estimates for each of the threedigit manufacturing industries and on occasion were estimated at the two-digit industryclassification level. Further discussion of Export data is provided within this section.Three samples were generated representing each of the years 1977, 1982 and 1987. Variousyears were studied in order to investigate any changes during the 15 year period described andto determine if results were robust such that consistency across years was evidenced. Sample sizein each year is 27 industries.Characteristics of the data measured are meant to include both trade and industry structure.Measured were annual industry Export sales to the world (Ex), industry size in Assets (A),Number of firms in the industry (N) and average industry Firm Size (S), used as a proxymeasure of industry concentration, where S =A/N. Each characteristic, except for firm Number,is measured in absolute nominal Canadian dollar values.Exports to the world measure the outward volume of trade activity for a particular industry.Employment of Import variables within the analysis may have given indication of intra-industrytrade patterns related to domestic industry structure versus the strictly export orientation takenin this study. Consequently, any policy implications born out of this analysis would be directlyaimed at exports to effect national wealth and welfare improvements.51Industry size in Assets was measured not only to generate average firm size data but also todetermine if export volumes were necessarily linked to the size of a domestic industry as awhole. Average Firm Size measures to some extent the degree of concentration within anindustry, with larger sized firms representing a more concentrated structure. Firm Number isthe final variable included in the analysis both on its own and as a determinant of Firm Size. Byincluding this variable, changes in Firm Size and industry concentration could be explainedeither through an expansion or shrinkage of the industry asset base or by changes in theaggregate number of firms. Caution is warranted here, however. Since nominal Asset valueswere used, changes in industry and firm size includes the effect of inflation.The 27 manufacturing industries analyzed are listed in Appendix 1. Some liberties inmeasurement of industry data were taken due to the problem of Export classifications mentionedabove. In order to acquire a sizeable enough sample to provide meaning to statistical results,some industries have been "double counted", however, instances of double counting are few.These instances occur in only two industries, Chemicals and Machinery, where sub-industriesrepresent the three-digit industry level and included within the sample are the noted two-digitindustries noted. Aggregate industry statistics were determined to hold significant informationover and above the sub-industry data. It was, therefore, expected that double counting wouldhave a minimal impact on the overall analytical results. The relative asset measures of sub-industries double counted are presented in Table 7.1 below.52Table 7 1 Sub-industry Asset Size as a Percentage of Aggregate Industry Classification1977 1982 1987Farm Machinery (as % of Machinery) 32.7% 45.8% 42.2%Fertilizers (as % of Chemicals) 3.5 2.2 6.6Pharmaceutical (as % of Chemicals) 9.5 9.6 12.7Source: Statistics Canada, various yearsDescription of the DataThe distribution of industry raw data is extremely skewed for each of Exports, Assets, Numberand Size in each of the three years studied. The data was transformed to log values in order tonormalize the distribution and provide meaning to statistical testing. Both raw data and thetransformed log data plots of distributions are shown in Appendices 2 through 5.Once transformed, the log data showed fairly normal distribution as indicated by the plotsthrough Appendices 2 through 5 and descriptive information of the log data listed in Table 7.2.Variance within samples is low and similar between years.Table 7.2 Description of Canadian Export, Asset, Firm Size and Firm Number DataLog Data ($000s)1977 N MEAN MEDIAN TRMEAN STDEV SEMEANexports 27 13.098 13.093 13.105 1.697 0.327assets 27 14.434 14.200 14.431 1.145 0.220number 27 5.666 5.969 5.680 1.347 0.259size 27 8.768 8.754 8.710 1.413 0.27253Table 7.2 Description of Canadian Export, Asset, Firm Size and Firm Number Data(cont'd)1982exports 27 13.827 14.015 13.850 1.633 0.314assets 27 14.977 14.779 14.982 1.214 0.234number 27 5.848 5.991 5.872 1.365 0.263size 27 9.129 9.038 9.073 1.447 0.2781987exports 27 14.094 14.582 14.101 1.758 0.338assets 27 15.422 15.275 15.427 1.141 0.220number 27 5.913 6.172 5.938 1.363 0.262size 27 9.509 9.560 9.468 1.497 0.288Exploration of the DataTwosample-T Tests:Cross-year data of industry variables was gathered in order to explore the relationship betweentrade and industry structure over a number of periods of time. Before this specific relationshipwas tested, exploration of the industry variables themselves was completed. Data was measuredto determine if any indication of change in means during the ten year period between 1977 and1987 was present based on two-period comparisons of the sample means and variance.Twosample t-tests were performed to test the null hypothesis that the means of the populationsin two different years were equal. The null hypothesis is written as;Eg. Ho: muEx82 muEx87Each of Exports, Assets, Number and Size and were tested across each of the five year periodsextending, from 1977-82 and 1982-87, and the ten year period extending from 1977-87. Table547.3 provides resulting p-values from twosample testing.Table 7.3 Two-Sample Statistical Testing p-ValuesInterval 77-82 82-87 77-87Exports .11 .57 .04Assets .097 .17 .0025Size .36 .35 .067Number .62 .86 .51Testing leads to the conclusion that at the 95% confidence level a change in industry Assets andannual Exports has occurred in the ten year period between 1977 and 1987. Although statisticallyweak for the second five year period, industry Assets seem to have changed in each of the fiveyear periods as well. One-sided testing for an increase in Assets and Exports during the ten yearperiod was statistically significant at the 95% confidence level (p=.0013 and p=.02respectively). A change in average firm size over the same period is significant at the 90%confidence level where an increase has taken place (p=.033). The null hypothesis of a changein number of firms across all industries could not be rejected within the two-five year periodsor throughout the ten year period ending in 1987 due to statistical insignificance.Since export and asset change is apparent over the longer ten year period it may be reasonableto hypothesize that there were no shocks to the Canadian economy during this time, at least nonewith lasting effects, that would have caused rapid growth or contraction in domestic andinternational trade and thus confounded further statistical testing of export and industryparameters. Also, a consistent number of firms across industries does not give any evidence ofshocks to affect industry structure on average, although changes within an individual industrymay have taken place. Finally, the increase in assets and no change in firm number is consistent55with the increase in firm Size noted for the sample. Nominal asset growth within all industrieshas seemingly not come from the addition of enterprises. Since nominal values were employed,real expansion of firms was not measured. If real expansion did take place, however, the aboveanalysis may provide indirect evidence of increased industry productivity from rationalization,higher capital intensities required by improved technology or increased investment activitystimulated by other macro-economic variables in the environment.ANOVA Testing:Further testing of industry characteristics during the ten-year period was carried out throughAnalysis of Variance testing (ANOVA). ANOVA testing for each of the three sample yearsshowed that the null hypothesis of all means equal could only be rejected at the lower 90%confidence level for Exports (p=.089), at the 95% level for Assets (p=.01) and could not berejected for firm Size (p=.179). In summary, the data is consistent. Asset change over the tenyear period is apparent, however, firm number was not. The expected results of some changein firm size were not confirmed by all three years compared using ANOVA testing, however,two-year comparisons did show the expected change in firm size. Note again that changes maybe inflationary only.Regression Analysis Final statistical testing was undertaken in order to measure the relationships between exportingand domestic industry variables. Assets, Size and Number were regressed on the dependentvariable, Exports, for each of 1972, 1977 and 1982 in order to determine if these factors56explained export activity.A number of models were proposed in order to test the data thoroughly. Within the modelsconjectured, we would expect to see Exports increasing with firm Size and industry size asmeasured by Assets. Empirical evidence described previously indicated several studies whereindustry exports were positively related to firm size. These results gives indirect evidence ofrationalization occurring within the industry as international trade increases. Since Assets are anindicator of output potential, we would also expect to see Exports increase with industry Assets.Therefore, within the regression models Asset and Size coefficients are expected to be positive.Coefficients for firm Number could be positive if increased participation by enterprises increasesindustry export output, or negative if rationalization within the industry occurs under exporttrade and firms exit.Model 1 The first model testing for both Assets and average firm Size as explanatory variables of tradeis described below. The model becomes a simple linear relation when log data is employedwhich previous data exploration showed to have a better, normal distribution for use in statisticaltesting.Exports = a•Assetsb -Sizec •errorOnly the Size coefficient, c, was statistically significant at the 5% testing level in each of the57three years tested showing exports increase as industry concentration increases. Although theAsset coefficient, b, and the constant were not significant, the sign of b was correct in two ofthe three years. Results are summarized in Table 7.4.Table 7.4 Regression Results of Model 11977constant/coefficient a b c R2(adjusted)value 2.36 0.1657 0.9706 49.2%t-stat -0.79 0.88 4.16p-value 0.439 0.39 0.001982constant/coefficient a b c R2(adjusted)value -0.523 0.0853 0.9061 46.7%t-stat -0.18 0.46 4.11p-value 0.858 0.648 0.001987constant/coefficient a b c R2(adjusted)value -2.745 -0.609 1.1294 46.1%t-stat -0.80 -0.31 4.43p-value 0.431 0.757 0.00Analysis of the residuals indicated non-normal distribution and slight heteroscedasticity withslight tails or step formations within in the standardized normal distribution plot. In conclusion,58the model does not explain much of the data and violates the linear assumptions of the regressionmodel. Since Size is derived from Asset data, some interaction between variables may bepresent. Interaction may also explain insignificant results for Model 1 yet significant coefficientsfor models regressing Assets and Size separately (see Models 2 and 3). Results for 1987 areunexpected, showing a negative coefficient relating industry Assets and Exports, although thisrelationship is insignificant.ModelSince both industry and firm size together did not explain export activity well, each factor wasregressed separately against the independent variable, Exports. The second model relates onlyAssets to Exports and is written below. Recall that actual log data is used within each model.Exports = a•Assetsb •errorRegression analysis results for all three years of Model 2 are given in Table 7.5. The secondmodel yields important information. A consistently positive sign for the Asset coefficient resultsin each year and is highly significant. Industry assets therefore provide at least some explanationof industry exports. The magnitude of the coefficient is also interesting to note since it isapproximately unity in each of the three years tested. Since the model uses log data, coefficientvalues represent elasticity measures of export activity given industry Asset size. The unitycoefficient indicates that a change in Assets would yield a proportionate change in Exports, orthat the magnitude of Exports movements is perfectly elastic with respect to movements in59industry Assets.Table 7.5 Regression Results of Model 21977constant/coefficient a b R2(adjusted)value -2.27 1.07 49.7%t-stat -0.76 5.16p-value 0.45 0.001982constant/coefficient a b R2(adjusted)value -0.48 0.96 48.4%t-stat -0.17 5.04p-value 0.87 0.001987constant/coefficient a b R2(adjusted)value -2.72 1.09 48.1%t-stat -0.81 5.01p-value 0.43 0.00Since the sample under study consists of manufacturing industries it would not be unreasonableto assume that many of the Assets measured represent capital investment. Under this assumption,the model indicates that exporting industries are capital intensive which supports previousconclusions of Caves and Khalilzadeh-Shirazi (1977). Industries characterized by asset intensityshould also face declining average costs in increased production and, therefore, be subject toeconomies of scale. Economies of scale assumes that firms do not operate at full capacityutilization all of the time. If an industry is characterized by increasing returns then exports60represent a means of realizing scale economies and lower average cost through increasedindustry output. Industry assets positively correlated with industry exports lends indirect evidencethat imperfect trade models are compatible with empirical evidence of industry trade.Conversely, the greater value of assets correlated to industry exports may simply give indicationof greater industry output to the global market which is supported by a larger aggregate assetbase. Asset intensity is better described with reference to the number of firms employingindustry assets for use in export. Firm Size is a better measure of asset intensity and results forthis variable are given below.Model 3 The third model employs firm Size as the sole independent variable and Exports as thedependent and is written below.Exports = a•Sizeb errorFirm Size differs from industry Assets in that firm concentration of assets within the industryis captured within the Size measurement. Considering the rationalization arguments alreadymade, we would expect larger firms to have larger capital investment and possibly larger fixedoperating costs. Given these imperfect market conditions, trade would provide opportunity torealize increasing returns to scale, therefore we would expect firm Size and Exports to bepositively correlated.61The model is another attempt to improve precision over Models 1 and 2 explaining exports withAssets and both industry Assets and firm Size. Results show that Model 3 is fairly accuratealthough the R2 is relatively low compared to Model 2. Other relative measures of significanceare that the constant and the b-coefficient are both statistically significant at the 95% confidencelevel for 1977 and 1982. Measurement for the final year, 1987, is nearly significant at the 90%level for both coefficient and constant. Regression results are listed in Table 7.6.Table 7.6 Regression Results of Model 31977constant/coefficient a b R2(adjusted)value 8.46 0.53 16.1%t-stat 4.42 2.45p-yalue 0.00 0.021982constant/coefficient a b R2(adjusted)value 9.69 0.45 12.8%t-stat 5.07 2.20p-value 0.00 0.041987constant/coefficient a b R2(adjusted)value 10.63 0.36 6%t-stat 4.95 1.63p-value 0.00 0.12The data exhibited three outliers for industry cases of Tobacco, Petroleum and Coal and TotalMining in each of the years studied. By weighting the data (these three cases for were given 0weight), the precision of the model was improved overall. Most notably, the b-coefficient value62became statistically significant at the 95% confidence level for 1987 (p=.02). Revised modelparameters are listed below.Year a b1977 6.59 0.761982 7.14 0.771987 8.17 0.64Results from Model 3 show the Size coefficients to be of the expected sign. One notable pieceof information from Model 3 different from Model 2 is the magnitude of the coefficient.Whereas Model 2's Asset coefficients were close to unity, Size coefficients are roughly half theorder of magnitude or, in the weighted case, roughly three-quarters of those of Model 2. Thisindicates that Exports change less than proportionately with changes in firm Size. These resultsare in line with previously empirical study already noted, but disagree with those studiesshowing exports to increase more than proportionately with increases in firm size. Thisrelationship may, therefore, vary across countries.The coefficient magnitude indicates that industry Exports are less sensitive to the Size of firmswithin the industry than to the size of the industry itself. The lower responsiveness of Exportsto firm Size may be due to variability within each industry of the number of firms exporting.If not all firms within an industry export, yet all firms increase in size, then a less thanproportionate change in exports would result. Asset changes at the industry level are fullycaptured and compared with exports of the industry. However, asset changes at the firm levelare diluted over the total number of industry participants and are therefore less responsive to63industry export activity.Model 4The positive relation between firm Size and Exports gives indirect evidence of rationalizationunder conditions of export trade. If firm size can be used as a proxy measure of industryconcentration, then arguments for increased trade from increased industry concentration can beposed. Indeed, this is the central theme that will be studied in further models. Increased firmSize as a measure of industry concentration would also indicate greater oligopolisticinterdependencies amongst fewer, larger firms. Since one of the basic components of industryconcentration is the number of firms, Model 4 below was constructed in an attempt to illustratethe relation between industry firm number and export activity.Exports = a•umberb -errorIf industry rationalization indeed takes place under trade, we would expect Exports to have nopositive relation or to be negatively related to firm number. If instead Exports are fuelled by agreater number of industry participants, the model coefficient is predicted to be positive.The results of the regression analysis are shown below in Table 7.7.64Table 7.7 Regression Results of Model 41977constant/coefficient a b R2(adjusted)value 12.03 0.188 0%t-statistic 8.3 0.75p-value 0.00 0.461982constant/coefficient a b R2(adjusted)value 12.39 0.245 0.4%t-statistic 8.82 1.05p-value 0.00 0.311987constant/coefficient a b R2(adjusted)value 12.17 .325 2.6%t-statistic 12.17 .325p-value 0.00 0.21The b-coefficients of firm Number have a positive sign in each of the three sample yearsanalyzed. This result lends weak support to the hypothesis that Exports are fuelled by thenumber of industry participants, each producing similar levels of output. Pursuing thishypothesis, analysis would suggest that a significant number of firms within the industry exportsuch that industry export activity would be sensitive to the number of firms within. The lowmagnitude of the model coefficient contradicts this hypothesis.If firm Size can be presumed to be representative of concentration, then Models 3 and 4 provideconflicting information regarding industry concentration and trade. Model 4 suggests thatincreased firm number, resulting in dilution of industry concentration, is positively related toexport trade. This dilemma was discussed within the basic theoretical model when ambiguous65effects of firm number were described under conditions of varying output across firms. Inconclusion, however, none of the Model 4 coefficients are significant, suggesting that anyexplanation of industry export activity described by firm number alone is weak.Part A. ConclusionFrom the industry data analyzed and regressed against industry Exports there exists cross-yearevidence that trade activity is explained by both the size of an industry and the size of firmswithin. These size considerations support previous empirical results. Exports increase with bothindustry and firm capital intensity and can be concluded to take place in the presence ofproduction or other returns to scale. Firms may, therefore, be motivated to trade underconditions of industry rationalization. Rationalization within an industry should in turn causegreater concentration, creating a link between industry structure and export trade.The previous regression analyses do not give any indication of market power derived fromindustry structure. The ability of firms to price-discriminate within foreign or domestic marketsis not captured within the above analysis. As well, firm Size is a very narrow descriptor ofindustry structure. The next section of statistical analysis will look specifically at Canadianmanufacturing industries' industry structure and market power as measured by industryconcentration.66B. CANADIAN INDUSTRY EXPORTS AND INDUSTRY CONCENTRATIONMEASURESIn order to better explore the relationship between industry structure, market structure and exportactivity, concentration data for the 27 industries was gathered and regressed against industryexport information for the years 1982 and 1987. Again, analysis was limited by data availability,however, concentration measures are consistent with other independent variables measured sinceeach corresponds to SIC three-digit codes.Data SelectionA number of measures of industry concentration were made available for employment withinregression models in order to test the robustness of the hypothesis that industry structure isrelated to export trade. Two sources of data were used to measure industry concentration. Thefirst data source is the Statistics Canada (1985) publication of Industrial Organization andConcentration (IOC) in the Manufacturing, Mining and Logging Industries, available for 1982.The IOC measures concentration as shipments of the leading four and eight enterprises as apercentage of total industry shipments. Industry shipments are classified under the 1980 SICfour-digit industry codes.Concentration Ratio is a proxy measure of both industry structure and the degree of marketpower measured through market share of the largest producer enterprises. Both the four firm(CR4) and eight firm (CR8) concentration ratios were incorporated into the analysis. Also67measuring concentration within the same data source is the Herfindahl index. The Herfindahlindex uses a similar means of calculating degree of concentration based on market share,however, indices are calculated by summing the squares of the market share of each industryenterprise. Indices tend to be smaller than concentration ratios for any more than one firmsupplying 100% of the market due to squaring. For example, a monopolist industry would showa Herfindahl index of 1, the maximum scale number of the index, and an industry of two firmswith equal market share would show an index of (1/2) 2 + (1/2)2 = 0.5. A similar figure wouldresult within an m-firm industry dominated by two firms, where m > 2. Although the Herfindahlmeasures contain more information regarding the market structure of an industry, each of theHerfindahl and Concentration Ratio measures indicate a more highly concentrated industry witha relatively higher index measure.Industry Export data estimated did not match industry concentration data within the IOC sourcedue to differences in classification and aggregation of industry information. IOC sources gaveConcentration Ratio and Herfindahl measures for the four-digit SIC coded industries and exportdata was estimated on a three-digit basis. To make the sources comparable, a weighted averageconcentration ratio was calculated of the four-digit industries within the three-digit code basedon the values of shipments. Where concentration data was missing due to confidentiality, theconcentration ratio for that four-digit industry was assumed to be that of the weighted averageof the other sub-industry classifications. Only in two of the 23 cases used was it necessary toestimate this figure, therefore, overall impact on the analysis should be minimal.68The Export data of the previous study was re-expressed as a percentage of total 1982 industryshipments as provided within the IOC information. The resulting export propensity variable isthe dependent variable used in the regression analysis of CR4, CR8 and Herfindahl models tofollow. Export propensity provides a standardized industry export variable for comparison tostandardized industry concentration data.The second class of industry concentration measure was sourced from Statistics Canada'sCorporate and Labour Union Returns Act (CALURA) information. The top four and eight firms'industry sales and asset measurements for each of 1982 and 1987 were used as an alternativemeasures of industry concentration (CALURA 4, CALURA 8) over 13 three-digit SICindustries.Description of the DataThe 1982 and 1987 CALURA data for assets (CALURA 4) and sales (CALURA 8) is describedin Appendices 6 and 7 respectively. Dotplots of the 1982 and 1987 raw and log data areprovided in Appendices 8 and 9 respectively. Within six of the eight samples, only 12 datapoints (industries) are used. Variation within 1987 samples is high compared to that of the 1982samples where distribution of assets is much tighter, although a few outliers exist. The standarderror of the CEx sample is particularly large, indicating that the mean is not a good estimatorof the population. All raw data for CALURA sales and asset information is skewed to the left.Although log transformations did not yield exactly normal distributions, log forms weredetermined to give better results than other transformations attempted. Statistical test results may69therefore be viewed with some caution due to the small sample size and poor normal distributionthat violates test assumptions.Export Propensity (CEx), Concentration Ratio (CR4, CR8) and Herfindahl (H) is described inAppendix 10 and distributional plots are given in Appendix 11. Comparison of the standarderrors and standard deviations of the raw sample data gives interesting results. There exists muchgreater variation within the CR8 sample than the CR4. Variation within the export propensityis very large as noted by the considerable standard error. The CR4 sample has a relatively tightdistribution compared to other concentration and export propensity samples. Log plots showfairly normal distributions for all CEx and concentration indices. A better presentation ofdistribution is given by the log normal plots for each of the CR4, CR8 and Herfindahl data inAppendix 12. All log sample plots have curves and data exhibits heteroscedasticity. Althoughdistributions are not perfectly normal, log transformations yielded better results than othertransformations attempted. Due to distributional problems, statistical results must be treated withcaution.Statistical TestingWe can hypothesize that increasingly concentrated industries would have a greater propensityto export if trade took place under conditions of increasing returns. A positive relation betweenexport propensity and industry concentration may also support related theories of internationalfirm specialization in a limited variety of industry goods. This cannot be fully tested here,however, since import data from intra-industry trade is not present. Intra-industry trade70measurements and domestic structure may be an interesting consideration for further study.A final hypothesis supported by positive export propensity and industry concentration relationis that oligopolistic industries may be able to price discriminate between domestic and foreignmarkets, earning rents in each. Using concentration as an indicator of the price-cost margin ormarket power within the domestic industry, ability to earn export market rents providesmotivation for trade. We would, therefore, expect a positive relation between export propensityand concentration of domestic industries.A negative relation between concentration and trade may show that domestic oligopolisticindustries cannot price discriminate between domestic and foreign markets and face flatter globaldemand curves. It may also indicate an increasing collective industry inability to participatewithin the more competitive global industry. This inability may be linked to domestic industrystructure. Barriers to industry entry may have caused inefficiencies within industries weakeninga firm's ability to compete globally. Finally, industry export propensity may depend on thenumber of firms exporting. Under this condition, a negative concentration/export relation wouldalso result since firm number would dilute industry concentration.Regression AnalysisModel 5Regression analysis testing the relationship between industry concentration and export activitywas undertaken by first testing CALURA data within the following model:71Exports = “CALURA Industry Concentration Measure*)b •errorwhere;*Industry Concentration Measure is indicated by:CALURA 4 and 8 Sales^ - 1982, 1987CALURA 4 and 8 Assets - 1982, 1987Given previous results of Models 3 showing a positive relation between industry exportquantities and industry firm size, we may speculate that since industries characterized by largerfirms tend to have greater exports, industries' largest firms will also show a positive relation toindustry exports.CALURA Sales and Assets measures showed both a positive and statistically significant relationto industry exports. One insight provided by both the CALURA Sales and Asset models is thatthe larger industry firms drive export activity of that industry. This would have to be confirmedby further study regressing separate samples of small and large CALURA firm size on exportactivity separately to identify any difference in their relation to exports. However, the magnitudeof the Model 5 coefficient may give indirect evidence of this result. Recall that overall averagefirm Size regression coefficients from Model 3 in previous analysis ranged from approximatelyone-half to three-quarters depending on the weighting of the data. CALURA coefficients areeach greater than one and highly significant. The magnitude of the Model 5 coefficient indicatesthat industry exports are more highly sensitive to large firms' characteristics of size than overallindustry measures. This may show that an industry's largest firms do drive exports and exportsare subsequently more responsive to changes within these firms than changes measured acrossthe industry as a whole. Regression results are given in the Table 7.8 below.72Table 7.8 Regression Results of Model 5b^p-value^t-statistic^R2Total Dollars^CALURA SALES 4-1982^1.36^.001^4.46^61.1%8-1982 1.42 .000 4.71 65.94-1987^1.25^.001^4.38^59.58-1987 1.26 .002 4.23 60.8CALURA ASSETS 4-1982^1.22^.005^4.99^71.28-1982 1.25 .000 5.37 71.74-1987^.795^.028^2.57^33.78-1987 .860 .023 2.69 36.2One area for further exploration of the CALURA data is to test the null hypothesis that nodifference exists between the four firm and eight firm model in describing industry exports. Thismay lend proof to the theory that a relatively few number of firms drive industry exports, suchthat the largest four firms would have a stronger relation to exports than the largest eight, or thatthe next largest four firms make no difference in explaining export activity. In order to test thisrelation, the null hypothesis was written as the b-coefficient estimated by the four firmregression model equal to the hypothesized value determined to be that of the b-coefficientestimated by the eight firm model. This hypothesis is written as;H0 :b(four firm) = b(eight firm),and the test statistic is calculated as;73(estimate) - (hypothesized value)(estimated standard deviation of the estimate).Results of the comparative tests are given in Table 7.9 below. For the four firm and eight firmCALURA Assets and Sales in each of 1982 and 1987 the null hypothesis tested could not berejected at the 90% or 95% level of confidence. The regression model estimate of the magnitudeof b-coefficient is therefore the same whether the top four or eight firms are used to explainexport activity for the industry.Table 7.9 Test Statistic Results of Model 5 b-Coefficient4& 8 Firms^t-stat t-stat t-stat Reject HoModel Comparison calculations 90% confidence 95% confidence1982CALURA -Sales^-0.197 -1.28 -1.65 yes-Assets^-0.133 -1.28 -1.65 yes1987CALURA -Sales^-0.034 -1.28 -1.65 yes-Assets^-0.210 -1.28 -1.65 yesModel 6In order to explore in greater depth the relation between an industry's propensity to export andits relative degree of concentration, the four and eight firm concentration ratio as well asHerfindahl measures were regressed against relative industry exports using the following model(data represents measures for 1982 only):74Export Shipments/Total Industry Shipments = a•(ConcentrationRatio Measurer •errorwhere;* Industry Concentration Ratio as indicated by:- CR4- CR8- HerfindahlResults of the regression analysis are given in Table 7.10 below.Table 7.10 Regression Results of Model 6Ratio b p-value t-statistic R2CR4 2.81 0.404 0.85 1.1%CR8 0.697 0.420 0.82 0%Herfindahl 0.256 0.615 0.51 0%Model 6 regression results give only weak indirect evidence that industry structure and exportpropensity are positively related. Statistical results may be poor in the model due to a numberof reasons. The export propensity information used is of poor quality. Recall that exportpropensity values use data derived from two different sources. Industry Export Shipments werebased on the Harmonized System of tariff classifications from Statistics Canada and thenestimated at the three-digit SIC code level. Total Industry Shipments data was taken from actualexport shipments in 1982 measured at the three-digit SIC code level. This difference in datasource and classification yielded somewhat inaccurate estimates for export propensity used.Other reasons for insignificant model results may be due to the interaction of domestic and75foreign industry factors not controlled for within Model 6. Concentration indicators do notaccount for foreign imports into the national market that dilute domestic firm market power.Definition of an industry by the SIC codes at the three-digit level may be too broad and basedon disparate products. Cross-industry participation by firms may also not give accurate indicationof a firm's market power, since the same firm may exert influence in some markets but not inothers. Multi-collinearity may also be present within the model, affecting significance. Finally,the size of an economy in which the industries operate may exert considerable influence overconcentration, where smaller economies are more highly concentrated than larger economieswith similar technology or similar sized firms.CR4 and CR8 models show a large difference in magnitude of the concentration coefficient.Testing for this difference, however, shows that the null hypothesis that the model's CR4 andCR8 coefficients are different cannot be rejected. Given that the model is subject to large errorsit is largely insensitive to differences in the order of magnitude of the coefficient betweenmodels. It may be inaccurate to infer any systematic cause of this result. The conclusion thatCR4 measures are more sensitive to industry export propensity than CR8 measures cannot bedrawn. Similar results and conclusions can be made for the Herfindahl model coefficientmeasure.Part B. ConclusionThe regression results of Models 5 corroborate with previous results showing a positive relationbetween the size of firms within an industry and that industry's export level. It can be tentatively76concluded that an industry's largest firms drive export volumes. Moving to standardized industryconcentration and export data, Model 6 gave only weak indirect evidence that increasinglyconcentrated industries tend to have higher export propensities, or export relatively more of thetotal industry output. Testing of industry structure and export activity in general provides indirectevidence in support of the theory that imperfect market structure is not only compatible with butmay provide motivation for export trade. The distribution of market power within an industrydoes not provide systematic evidence that links it to industry export activity and any indicationof a relationship is weak.Concentration and firm size data as evidence of Canadian industry structure may be biased sincethe Canadian economy is relatively small compared to other industrialized nations. The smallerthe boundaries imposed upon an area of economic activity, the greater the concentration will bewithin that boundary and the greater the volume of trade measured between the bounded areaand outside economies. It would be useful, therefore, to analyze the market structure and exportactivity relation within a large economy. The following section of the paper explores data fromUnited States (US) manufacturing industries for signs of a relation between domestic marketstructure and international trade.77C.CONCENTRATION AND EXPORT OBSERVATIONS OF US MANUFACTURINGINDUSTRIESData SelectionData was sourced from the 1987 Census of Manufacturers (the Census) measuring four firm andeight firm Industry Concentration Ratios as well as Exports from Manufacturing Establishments.The unit of measure is one industry as described by the US Standard Industrial ClassificationCode (SIC) of 1987 at the three digit level. Total sample size is 140 industries.Industry Concentration Ratios provided by the Census at the four digit level and were averaged,weighted according to the total value of industry shipments, in order to arrive at a three-digitindustry concentration measure for comparison to Canadian information. Industry ConcentrationRatios measure the sum of the share of industry shipments of the four (US CR4) and eight (USCR8) largest enterprises within.Direct Exports from Manufacturing Established were expressed as a percentage of Total IndustryShipments (USEx) providing an industry measure of export propensity. This relative indicatoris required in order standardize data across industries and to compare relative export data withrelative market values expressed by concentration measures.Description of the DataAppendix 13 describes the raw and log US data samples of USEx, USCR4 and USCR8. Included78are descriptions of the sample means and standard deviations. Variance within the export sampleis high although the mean is a good estimator as shown by a low standard error value. Raw datadistributions of concentration ratio samples are slightly skewed to the left. The USEx sample isextremely skewed in the same direction. All Census samples were transformed to logs in orderto normalize the data and for ease of comparison to previous Canadian study undertaken.Appendix 14 provides plots of the raw and log data for visual display of sample distribution.After normalizing concentration data, skewedness is still present and oriented towards the right.Appendix 15 provides Normal Plots of log and raw data comparing the changes in distributionbrought on by transformations. For the CR4 data the log normal plot shows a very tightdistribution around the 45 degree line indicating a normal dispersion. Distribution of the CR8and USEx data exhibits slight tails and subsequent non-normalcy. Logs may haveovercompensated for the skewedness of the original data, however, distribution of the samplesis not considered to be a deterrent to statistical analysis.Exploration of the Data Analysis of Variance:One issue explored within the US data concerned the similarity of samples of the four firm andeight firm concentration ratios. Analysis of variance (ANOVA) of the CR4 and CR8 data testedthe null hypothesis that sample means were equal, indicated as follows;H:0 mu(US CR4) = mu(USCR8)79This testing was done in order to determine if four firm concentration index was a significantlydifferent measure than the eight firm measure of industry structure. Results were expected togive some insight into the overall structure of US industry and to determine if the extent ofconcentration varied between these two measurement levels. For example, if a US industry wascharacterized by four or less firms enjoying significant market share then when aggregating overthe larger number of eight firms we would expect the eight firm concentration measure toimpose little difference on the CR4 industry concentration measure. Given roughly equal firmmarket share within an industry, we would expect greater aggregation at the eight firm level toyield greater measures of industry concentration than at the four firm level. An indicator of thissame issue was given when testing the difference in b-coefficients of Model 5 using four firmand eight firm CALURA sales and asset levels. Within Canadian industry, concentrationcoefficients did not change between the four firm and eight firm model. ANOVA testing appliesthis same query to the US data.Results of ANOVA testing for the US industry data are statistically significant (p slightly >0.000), therefore, concentration ratios are different at the four firm and eight firm levels. Table7.11 below gives statistical results for the ANOVA test. Market share of the more highlyaggregated eight firms is concluded to be greater than of the top four enterprises. This may givesome indication of low concentration levels of US industry. The domestic US market may besizeable enough to allow plant or firm production efficiency while still allowing a high numberof firms to participate within an industry. It is this difference which makes US and Canadiancomparisons interesting, since domestic market size in the US case may not constrain the80relationship between domestic market structure and international trade.Table 7.11 USCR4 and USCR8 ANOVA Testing ResultsANALYSIS OF VARIANCESOURCE DF SS MS F pFACTOR 1 6.720 6.720 32.58 0.000ERROR 276 56.932 0.206TOTAL 277 63.652INDIVIDUAL 95 PCT CI'S FOR MEANBASED ON POOLED STDEVLEVEL N MEAN STDEV -+^1gUSCR4 140 3.5121 0.4854^(----*----)1gUSCR8 138 3.8231 0.4202-+^+^+^+POOLED STDEV = 0.4542^3.45^3.60^3.75^3.90Comparison of US and Canadian Data: Export PropensityANOVA testing was continued in order to examine the null hypothesis that US and Canadianpropensity to export (CEx) for 1987 are the same. Results yielded were statistically significant(p slightly > 0.000) and allowed the null hypothesis to be rejected. Canadian and USmanufacturing industries, therefore, have different export propensities. Table 7.12 providesstatistical and graphical results of the ANOVA test run.Table 7.12 US and Canadian Export Propensity ANOVA Testing ResultsSOURCE DF SS^MS^F^pFACTOR^1^74.77^74.77 49.03^0.000ERROR^158 240.98^1.53TOTAL^159 315.75INDIVIDUAL 95 PCT CI'S FOR MEANBASED ON POOLED STDEVLEVEL^N MEAN STDEV --+^+^+^+----lgUSEx^135 1.485^1.220 (--*--)1gCEx87^25 3.367^1.314 (^ *^)-+^+^+^+ ----POOLED STDEV = 1.235^1.40^2.10^2.80^3.5081As a more intuitive indicator than logs, the average US industry exports approximately 7.6%of total output according to the 1987 Census. This figure is extremely low compared to theCanadian average derived for 1982 of nearly 42%. Recall that the problems of accuracy withinthe Canadian data as well as the larger and, therefore, more accurate US sample size. Thesefactors, along with the different methods of aggregating and classifying industries between theUS and Canada, means exact comparison is impossible. However, the magnitude of the twosamples must be appreciated. This difference in export propensity gives strong indication of theeffect that size of an economy may have on trade variables. To appreciate any effect onconcentration, ANOVA testing was undertaken for relative Canadian and US concentrationmeasures and is discussed below.Comparison of US and Canadian Concentration DataThe null hypothesis was tested that US and Canadian industries have, on average, the sameconcentration levels. ANOVA testing yielded statistical evidence that this hypothesis could notbe rejected using both the CR4 (p=.144) and CR8 measures (p=.364). The 1987 USconcentration data was compared to 1982 Canadian information due to the unavailability ofCanadian information for that year. It may be reasonable to assume that industry concentrationon a national average basis would not change significantly within a five year period. If so, the1982 Canadian data may be used as proxy measures for 1987 for the purpose of comparison.Results of ANOVA analysis used to test the null hypothesis of same industry concentration is82given below in Table 7.13.Table 7.13 US and Canadian Industry Concentration ANOVA Testing ResultsFOUR FIRM CONCENTRATIONANALYSIS OF VARIANCESOURCE DF SS^MS^F^pFACTOR 1^0.442^0.442 2.16 0.144ERROR^161 32.922^0.204TOTAL^162 33.363INDIVIDUAL 95 PCT CI'S FOR MEANBASED ON POOLED STDEVLEVEL N MEAN STDEV^+^+^+^+--1gUSCR4^140 3.5121 0.4854 (^*^)1gCCR4^23 3.6616 0.0890^( *^)---- +^+^+^+ --POOLED STDEV = 0.4522^3.48^3.60^3.72^3.84EIGHT FIRM CONCENTRATIONANALYSIS OF VARIANCESOURCE DF SS^MS^F^pFACTOR^1^0.140^0.140 0.83 0.364ERROR^159 26.829^0.169TOTAL^160 26.969INDIVIDUAL 95 PCT CI'S FOR MEANBASED ON POOLED STDEVLEVEL N MEAN STDEV ^1gUSCR8^138 3.8231 0.4202^(^*^)1gCCR8^23 3.9073 0.3467 ( *^)POOLED STDEV = 0.4108 3.80 3.90 4.00Intuitively we would expect the smaller Canadian economy to have more highly concentratedindustries than within the US. This expectation is supported by the greater average concentrationmeasures using both CR4 and CR8 indicators. However, Canadian sample variance is great andstatistically significant results were not achieved. Although we cannot draw any firm conclusionthat concentration of industries differs between the two countries, comparing the magnitude of83the relative concentration measures does give weak indication that Canadian industries tend tobe more concentrated within the economy than US industries.Conclusions:US concentration measures of CR4 and CR8 are different. The additional four firms' marketshare within the CR8 calculation imply that concentration within US industries may be low suchthat a very few firms do not dominate industries. The additional CR8 concentration measureshould provide supportive evidence testing US industry structure as first measured by the CR4index and trade.After comparing the US and Canadian data it can be only tentatively concluded that USmanufacturing industries are less concentrated than their Canadian counterparts. Although morerobust measurement was not possible, Perrakis (1990) states that empirical study of Canada andthe US during the early 1980s shows that Canadian industries are indeed less competitive thantheir US counterparts as measured by industry concentration. Insignificant differences in US andCanadian concentration measures as tested above may therefore be due to poor or insufficientsample data.Considering exports, Canadian manufacturing industries tend to greater export propensity thantheir US counterparts. This may demonstrate the effect of small economy size on trading patternsand activity. In general, country comparisons are important to determine if differences withineach national market affect any possible relation between industry structure and trade.84Regression AnalysisModel 7Regression analysis was performed in order to determine if 1982 US industry structure explainedthe export propensity of industries in that year. The premise of the analysis has not changedfrom the Canadian market study and the regression model is the same as Model 6, rewritten forthe US case as Model 7 below.US Industry Exports/Total Industry Shipments(USEx) = a.(US Industry ConcentrationRatio*)b errorwhere;* US Industry Concentration Ratio is:USCR4USCR8Results of the model analysis are given in Table 3.4.Table 7.14 Regression Results of Model 7Concentration Ratio Coefficient/Constant p-value t-statistic R2USCR4 b =^.741 .001 3.55 8%a =^-1.12 .132 -1.52USCR8 b = .869 .001 3.57 8%a =^-1.86 .05 -1.98The USCR4 model constant is insignificant based on poor R2 and an insignificant constant value,however, the p-value for the concentration coefficient is of the expected sign and is significantat the 5% testing level. The USCR8 data fits the model slightly better as shown by both a85significant constant value and b-coefficient value at the 5% testing level. Normal plots of themodel residuals are given in Appendix 16. Although some heteroscedasticity exists as noted byslight curves in the plots, each illustrates a fairly normal distribution for both the USCR4 andUSCR8 models. The model is concluded to be a good descriptor of the export and concentrationdata.The magnitude of the concentration coefficient is less than one. The model predicts a smallerproportional increase in US export propensity for a given increase in US industry concentration.This elasticity measure is interesting to note as an indicator of export sensitivity to domesticindustry structure. Further study could explore various country elasticity measures to determineif these elasticity measures generated for the US industries are relatively high or low.Compared to Canada, US industries show lower sensitivity to export propensity. Canadianelasticity measures were greater than unity when testing proxy measures of industryconcentration, however, these measures are not equal since standardized or relative export andconcentration measures were not used. Concentration ratio elasticity of Canadian exports couldnot be determined with any precision and model parameters indicated both large (greater thanthree) and small magnitudes for the concentration coefficients. Since the significant Canadianand US models measure different data, direct comparisons are difficult.The final test conducted within Model 7 was to determine if the USCR4 model coefficientdiffered from the USCR8 model. A difference would indicate varying sensitivity of the eight86firm concentration measure to the four firm concentration measure. If US industry exports aredriven by a very few large firms, then we would expect the b-coefficient of the four firm modelto be greater than that of the eight firm since the four firm measure would be more highlysensitive to export activity. If the top eight firms participate relatively equally in export trade,both models will be equally sensitive and the b-coefficients should be the same.Similar to the Canadian analysis undertaken with Model 5, the hypothesized coefficient valueas determined from the CR8 model was used to calculate a test statistic which indicated adifference between the USCR4 and USCR8 model coefficients. Testing revealed no differencebetween the USCR4 and USCR8 models, therefore the top four firms do not have a differentpropensity to export than the top eight firms within an industry.Conclusions:The strong significance of the US concentration coefficients supports the hypothesis thatdomestic industry structure is related to international, specifically export, trade. Since a relationis evident it may be useful to re-explore various hypotheses explaining imperfect marketstructure and export trade within the context of the US economy.Since manufacturing industries were tested, industries may be subject to production economiesof scale which are realized under conditions of increasing industry concentration when firms facelarger global markets. However, the US domestic market is large, possibly large enough to allowfirms to reach their minimum efficient scale (MES) without export production. Also, the export87propensity of industries was found to be fairly low, at least compared to the Canadian economy.One indication that the industry MES may be reached by a number of firms within the domesticeconomy was the difference between four firm and eight firm concentration measures, where anadditional four firms over the USCR4 measure significantly changed the aggregate market sharemeasure. If rationalization gains had been exhausted with only four firms, eight firms wouldeither not exist or would include marginal producers whose production would not significantlyaffect industry concentration. As industry output increased, concentration would decrease sinceadditional firms must enter the market to expand deliveries. The opposite was demonstrated withModel 7, export propensity increased with higher concentration. Therefore, either opportunityfor gains from rationalization are still present within US oligopolistic industries through exporttrade, or possibly non-production economies of information, technology, managerial skill orother trading costs are realized through trade.The above argument must be made with careful qualification. Export propensity, not exportvolumes, are increasing with industry concentration. This means that higher concentratedindustries simply direct a greater proportion of their output to foreign markets versus domestic.It is intuitively appealing to think of firms first supplying domestic demand and operating onsome residual demand curve within foreign markets. Conceivably, however, a single USenterprise could have no sales to the domestic market yet exhibit a strong correlation betweenexport propensity and market dominance. Also, export activity does not necessarily mean thatindustries are net exporters, only that firms are engaged in international trade. Intra-industrytrade could conceivably be on such a scale that the US industry would be a net importer.88The hypothesis of intra-industry trade due to firm specialization put forth by Krugman (1992)may be supported within the regression evidence. As rationalization and consequentconcentration increase, firms export greater amounts to global consumers while maintainingsome constant supply to domestic markets as dictated by domestic preference, therefore, bothconcentration and export propensity move together.The final hypothesis supported by US Model results is that oligopolistic industries can price-discriminate in foreign markets and earn export rents which provide incentive to engage inexport trade. This hypothesis does not necessarily explain the degree of industry concentrationaffecting trade, other than to further a hypothesis that the greater the domestic industry marketpower, the greater the ability to price discriminate in foreign markets.Mode 8Insignificant results were generated by testing the relation of industry concentration with totalabsolute industry exports (US Totex) shown as Model 8 below. Although comparing a relativeand absolute measure is difficult, a relation would be explainable if, for example, largerindustries with greater export volumes tend toward higher concentration levels. The model,although insignificant, did show the expected sign for the b-coefficient.US Total Industry Exports = a.(Industry Concentration Ratio*)b •errorwhere;*Industry Concentration Ratio is measured by;USCR4USCR889Statistical results of the testing are given in Table 7.15 below.Table 7.15  Regression Results of Model 8.159 .306 1.03 3.1%1.74 .002 3.16.246 .054 1.36 2.7%1.36 .177 1.94It is difficult to predict the relation between industry structure and export activity from the modelsince it relies on specific export volumes correlating to specific concentration values. Therefore,specific industry structure, as measured by concentration, would predict a specific dollar volumeof export activity versus a relative measure of export propensity. If the model could explainexport volumes through concentration measures then we may expect industries with very largeoutput to be subject to very large economies of scale and as well to be very highly concentrated.Obviously industries are variant in terms of volume output and concentration. One industry maybe represented by a domestic monopolist, however, the size of the industry and therefore themonopolist's output could be relatively small. It would be reasonable to assume that for theseor similar reasons concentration does not explain export volume and Model 8 is insignificant.VIII.EMPIRICAL STUDY: REMARKSConclusions drawn from the variety of statistical testing are that measures of industry structureexplain export industry trade. Other aspects of export trade such as factor endownments orcompetitive advantage, employed within traditional trade models, are not explicitly includedConcentration Ratio^Coefficient/Constant^p-value^t-statistic^R2USCR4^b =a =USCR8^b =a =90within the model and instead have been assumed to be influence the error term of regressionmodels.In Canada, export volumes increase not only with firm and industry size but are also positivelyrelated to the size of the largest firms within the industry. This evidence supports the hypothesisthat increased concentration leads to increased export output. Export intensity was also shownto increase with the degree of industry concentration, however, this evidence is weak in theCanadian case. By testing the larger US manufacturing economy, evidence was found that apositive relationship exists between industry concentration and export levels.Conclusions drawn from empirical data support the theory that export trade takes place withinimperfect markets, where market imperfections are characterized by large firm size and industrystructure is characterized by high industry concentration. By understanding this relationshipbetween market structure and trade, government policy directed at trading industries may bebetter focused on strategies seeking to maximize national social welfare.ILIMPLICATIONS FOR GOVERNMENT POLICYGovernment trade or industrial policy must be made with cognizance of conditions under whichinternational trade takes place. Through increased understanding of the conditions under whichtrade takes place, welfare effects of policies can be determined and wealth maximizing policiesdesigned. The empirical study described within this paper shows that trade takes place within91imperfect markets. Both market and domestic industry structure define two of the parametersof trade.Implications from the Basic ModelTrade and Industry Costs:The basic theoretical model presented of trade under oligopolistic market conditions showed thatany total marginal cost advantage of domestic producers over their foreign rivals yielded higherequilibrium market share for the domestic industry than the foreign in any given market. Themodel predicted that increases in foreign costs relative to those of domestic producers causesshrinkage in foreign output and domestic firm output increases.As domestic output rises relative to foreign, rents are shifted to the home-country nation.Therefore, any policy directed at decreasing domestic firm marginal costs increases producerprofits through output expansion. Example policies may be production subsidies or directsubsidization of marginal export costs. Any policy directed at increasing foreign firm costs couldbe implemented for the domestic market only. An import tariff or tax would shrink foreignsupply to the domestic market and increase domestic supply with expansion causing increasedproducer profits and tariff revenue improving the wealth of the national treasury.If markets are imperfect due to high fixed or large investment costs then policies causingincreased output within these industries will yield gains from rationalization. Such policies maybe politically driven, such as the North American Free Trade Agreement which seeks to provide92secured access to larger markets for Canada. Production subsidies would also result in increasedoutput. Since any subsidy represents a transfer from government to producers, national welfaregains would only result if rationalization gains exceeded the amount of the subsidy. Similarly,domestic market protection through import tariffs or taxes would decrease foreign market shareof the domestic industry and allow higher domestic output which in turn would allow efficiencygains. If domestic and export market output costs are complementary, these lower costs couldalso yield producer surplus increases from expanded export market shares.Scale efficiencies effected by increased output from trade may include rationalization of tangiblefixed production costs or less tangible costs of Research and Development (R&D), marketinginvestment in domestic and international distribution networks, management skill in foreignmarkets, and/or indivisible information costs of exporting. Concentration in exporting industriesmay be the result of firms seeking to rationalize theses indivisible costs incurred by trade. Thisprediction is supported particularly well by the empirical results provided for US industrieswhere higher degrees of concentration correlated with higher export propensity. Policy effectson industry concentration will be discussed in greater detail later.Trade and Firm Number:The policy question considering industry firm number can be made dichotomous in order tofurther analyze welfare effects from trade from the perspective of producers - should domesticindustry expansion in both domestic and foreign markets be achieved through a large numberof firms producing relatively little output, or a small number of firms delivering a larger amount93of the good, each considering their own profit maximization?Government should be indifferent to the high number (of firms), low (per firm) output versusthe low number, high output alternative if the marginal and average cost of production isconstant across all firms in the exporting industry. Under each alternative, aggregate producersurplus for the industry would be the same and divided amongst some number of firms, n. If,however, firms face economies of scale then expansion not only yields increased producerwelfare given some market price but also gains from rationalization. Under increasing returns,government policy would prefer trade expansion to be carried out by a low number of firms withhigh output.Implications from Empirical Evidence for Rent ShiftingThe conduct of trade within imperfect markets as shown by theoretical and empirical studymeans opportunity exists for shifting economic rents from foreign to domestic producers. Withinimperfect markets, increased domestic output to the global market (either domestic or foreignimport-competing) implies increased share in industry rents and decreased market share offoreign competitors.Under conditions of imperfect competition, policy directed at the concentration of domesticindustries may represent a viable means of maximizing national welfare through increasedproducer surplus. Specifically, policy to decrease domestic industry competition may reduce theaverage costs of exporting industries and improve the ability of firms within to compete94internationally and accrue greater economic profit from international markets than foreigncompetitors.Industry concentration is affected in one manner by national competition policy relating tomerger and acquisition activity. If exporting firms seek to merge then profit motivations mustbe present such that the combined firm profit is at least as great as the aggregate profit of firmsoperating at arms length from one another. By internalizing market transactions within a largerfirm, synergies or larger profits are realized. Internalized transactions may yield economies ofscale in production activities as well as those specifically relating to export such as informationcosts, marketing, distribution and finance. Cost reductions relative to foreign rivals yield greaterdomestic output and greater global market share at the expense of foreign rivals. If productiontechnologies and costs are relatively constant across international markets, more liberalcompetition policy within the domestic regime may yield net increases in national welfare.Greater domestic industry rationalization and consequent concentration could allow governmentsto shift rents within international markets where foreign country policies are stricter. Asconcentration and export market share increases, profits are earned at the expense of foreignrivals.Additionally, concentrated domestic industries may also be able to exert market power in foreigncountries and increase national producer surplus through the acquisition of international marketrents. Policies increasing domestic industry concentration, therefore, need not be motivated onthe basis of cost rationalization only.95Strategic alliances within imperfect markets represent an alternative means to increasing domesticconcentration. Alliances seek to exact a pareto optimal solution from the producer standing ofcombined project (or alliance firm) profits greater than aggregated individual project profits.Alliances are not a permanent firm merger and as such do not technically affect industryconcentration measures. Effectively, however, concentration is increased. If profit increases arethe result of increased global market share for the combined firms' project then policies topromote strategic alliances will cause rent shifting and improve national domestic welfare.Rents may also be shifted by causing industry expansion within the domestic market. As industrysize increases, exports increase and therefore greater global rents are earned. Expansion mayyield gains from rationalization as well and give domestic firms cost advantages over foreignrivals. Protection of home markets through import tariffs, production subsidies to exporters orexport subsidies applied directly would each work to reduce firm costs and shift rents todomestic producers from foreign market participants.Analyzing Producer and Consumer Welfare EffectsThe above implications for rent shifting and other trade policies implicitly assume that priceseffects are not large enough to effect reductions in industry producer surplus. Also, changes inconsumer surplus are not considered. Possibly no change in consumer surplus would result ifproduction simply shifted from foreign to domestic suppliers and no industry output or pricechange was effected. However, deadweight losses may be incurred if higher domestic marketshare causes increased market power such that domestic market prices rise and output is96reduced. These monopolistic distortions cause producer surplus that may outweigh resultantconsumer losses. Analysis must, therefore, consider each of the producer and surplus effectsfrom trade and industrial policy and focus on maximizing net social welfare. Any losses inconsumer surplus as a result of trade policy promoting producer interests must be outweighedby increases in producer benefit.One means of effecting greater oligopolistic interdependencies and greater producer surplus fromworld markets without necessarily reducing consumer welfare may be the promotion of inter-firm non-price cooperation. Cooperation represents a viable means of reducing industry costsand directly affecting output variables without increasing domestic market power as long as firmsdo not restrict supply in an effort to raise prices. Policy directed at non-price cooperation maybe the promotion of shared industry investment such as R&D, marketing investment or sharedindustry finance. Indeed, the imperfect trade model employed within this analysis could beapplied to industry rationalization not through firm exit or merger but through reduced averagecosts by reduced duplicate industry investment.X.FINAL REMARKSThe existence of imperfect international markets represent opportunity for national governmentsto intervene in international trade with policies that allow improved national social benefit fromeconomic rents shifted from foreign constituents. A variety of government policy implicationsfor trade in imperfect markets and industries have been discussed, however, a few general97themes can be summarized. Note beforehand that all policy implications do not consider foreigngovernment retaliation. If all governments gave advantage to domestic constituents at some costto foreigners, markets would be distorted and global welfare losses may easily result. Rationalfor international agreements on the conduct of governments effecting trade policy, therefore,exists.The first general theme of the paper is that industry costs lower than foreign rivals causes higherglobal output and higher global market share. Policies working to lower costs relative to foreignindustry rivals, therefore, accrue a greater majority of global industry rents to the nation.Where firms face increasing returns to scale, lower costs can be influenced by expanding bothdomestic and export output within the domestic industry. Efficiency gains as well as expandedproducer surplus may result. Policies affecting higher industry concentration may also allow forefficiency gains from trade. Industry non-price cooperation between firms to reduce export andother costs and increase international competitiveness may be an alternative policy option inorder to realize gains from trade without increasing domestic industry concentration and marketpower resulting in market inefficiencies. Benefits from concentration need not be restricted toproduction rationalization. Concentrated domestic industries may exert market power in foreignnations and extract industry rents from additional export sales under conditions of high foreignmarket price-cost margins.A variety of trade policy tools may be applied to industry depending on the various assumptions98regarding industry structure and oligopolistic interdependencies. For this reason, trade policymust be designed and applied with cognizance of the particular aspects motivating trade withineach industry. For example, trade policy directed at increasing industry concentration orrestricting industry imports may have negative welfare implications if industries do not faceincreasing returns to scale since resultant domestic market power may harm domestic consumers.Impacts of trade policy must be measured from both producer and consumer standing. Tradepolicies generally applied may not necessarily improve net social welfare.Finally, domestic structure and export trade are positively related. Under conditions of imperfectmarkets, domestic competition policy may affect international trade. Policy affecting merger andacquisition activity or promotion of inter-firm cooperation falls under the rubric of competitionpolicy. International agreements may, therefore, take interest in these strategies currentlyimplemented within the domestic policy regime. By moving domestic competition policy to theinternational realm, world utility should not be as easily jeopardized by domestic rent-seekingactivities.99BIBLIOGRAPHYAuquier, A., Sizes of Firms. Exporting Behaviour and the Structure of French Industry, TheJournal of Industrial Economics, Vol.28, December, 1980 in Jacquemin, Alexis,Imperfect Market Structure and International Trade - Some Recent Research, Kyklos,Vol.35, 1982:75-93Axelrod, Robert. The Evolution of Cooperation, New York: Basic Books, 1984Bain, J.S., International Differences in Industrial Structure, New Haven: Yale University Press,1966Baldwin, Robert E., The Political Economy of Trade Policy, Journal of Economic Perspectives,Vol.3, No.4, Fall 1989:119-135Baumol, William J., Horizontal Collusion and Innovation, The Economic Journal, January192:129-147Bernstein, Jeffrey I. and Pierre Mohnen, Price-cost Margins. Exports and Productivity Growth,Canadian Journal of Economics,Vol.34, No.3, August 1991Brander, James and Paul Krugman, A Reciprocal Dumping Model of International Trade,•^Journal of International Economics 15, 1983:313-321Brander, James, Intra-Industry trade in identical commodities, Journal of InternationalEconomics 11, 1981:1-14, in Brander, James and Paul Krugman, A Re ci zp_sgWJaupgnModel of International Trade, Journal of International Economics 15,1983:313-321Brander, James and Barbara J. Spencer, Trade Warfare: Tariffs and Cartels, Journal ofInternational Economics 16, 1984:227-242Brander, James and Barbara J. Spencer, Export Subsidies and International Market ShareRivalry, Journal of International Economics, 1985Brown, Drusilla A. and Robert M. Stern, US-Canada Bilateral Tariff Elimination: The Role ofProduct Differentiation and Market Structure, in Robert C. Feenstra (ed.), Trade Policiesfor International Competitiveness, The University of Chicago Press, 1989Canadian Socio-Economic Information Management System (CANSIM) - Statistics Canada,Balance Sheet Data and Income and Retained Earnings Statistics: All Industries, variousmatricies 5100-5308, Supply and Services Canada, 1992CANSIM - Statistics Canada, Merchandise Trade with All Countries, Summary of InternationalTrade. v. 65.001, Supply and Services Canada, 1992100Caves, R E International Trade. International Investment and Imperfect Markets, InternationalFinance Section Special Papers in International Economics, No. 10, Princeton University,Princeton, 1974Caves, R.E. and J. Kahlilzadeh-Shirazi, International Trade and Industrial Organization: SomeStatistical Evidence, in A.P. Jacquemin and H.W. deJong, (eds.), Welfare Aspects ofIndustrial Markets.  Stenfert Kroese B.V., 1977Caves, R.E., International Cartels and Monopolies in International Trade, in (eds.) R.Dornbusch and J. Frenkel International Economic Policy, Baltimore, John Hopkins Press,1979Caves, R.E., Intra-Industry Trade and Market Structure in the Industrial Countries, OxfordEconomic Papers, Vol.33, 1981:202-223Chamberlin, Edward H. The Theory of Monopolistic Competition, Cambridge, Mass.: HarvardUniversity Press, 1933 in Caves, R.E. Intra-Industry Trade and Market Structure inthelndustrial Countries, Oxford Economic Papers, Vol.33, 181Dixit, Arinash, International Trade Policy for Oligopolistic Industries, Conference Papers,Supplement to the Economic Journal, Vol.94, 1983Eastman, H.C. and S. Stykolt, The Tariff and Competition in Canada, Macmillan, 1967Eaton, Jonathan and Gene M. Grossman, Optimal Trade and Industrial Policy Under Oligopoly,in Gene M. Grossman (ed), Imperfect Competition and International Trade, MIT Press,1992Feder, Gershon, On Exports and Economic Growth, Journal of Development Economics 12,1982:59-73Fung, K.C., Characteristics of Japanese Industrial Groups and their Potential Impact on US-Japanese Trade, in Robert E. Baldwin (ed.), Empirical Studies of Commercial Policy,University of Chicago Press, 1991Glesjer, H., A. Jacquemin, and Petit J., Exports in an Imperfect Competition Framework: AnAnalysis of 1446 Exporters, Quarterly Journal of Economics, Vol.94, May, 1980 inJacquemin, Alexis, Imperfect Market Structure and International Trade - Some RecentResearch, Kyklos, Vol.35, 1982:75-93Graham, Frank, Some Aspects of Protection Further Considered", Quarterly Journal ofEconomics 37, 1923:199-237101Grossman, Gene M. and Elhanan Helpman, Innovation and Growth in the Global Economy,MIT Press, Cambridge, 1991Helpman, E. and P. Krugman, Market Structure and Foreign Trade,Cambridge, MA, MITPress, 1985Jacqemin, A.P. and M. Cardon de Lichtbuer, Size Structure. Stability and Performance of theLargest British and EEC Firms, European Economic Review, December, 1973:393-408Jacquemin, Alexis, T. Nambu and I. Dewez, A Dynamic Analysis of Export Cartels: TheJapanese Case, The Economic Journal, Vol. 91, September 1981:685-696Jacquemin, Alexis, Imperfect Market Structure and International Trade - Some Recent Research,Kyklos, Vol.35, 1982:75-93Katz, B. and Phillips, A., The Computer Industry, in William J. Baumol, Horizontal Collusionand Innovation, The Economic Journal, january 192:129-147Krugman, P.(ed), Strategic Trade Policy and the New International Economics, Cambridge,MA, MIT Press, Cambridge 1986Krugman, P., Import Protection as Export Promotion: International Competition in the Presenceof Oligopoly and Economies of Scale, in Gene M. Grossman (ed), Imperfect Competitionand International Trade, MIT Press, 1992Krugman, P., Scale Economies. Product Differentiation and the Pattern of Trade, in Gene M.Grossman (ed), Imperfect Competition and International Trade, MIT Press, 1992Markusen, James R. and Anthony J. Venables, Trade Policy with Increasing Returns andImperfect Competition: Contradictory Results from Competing Assumptions, in Gene M.Grossman (ed), Imperfect Competition and International Trade, MIT Press, 1992Michaely, Michael, Exports and Growth: An Empirical Investigation, Journal of DevelopmentEconomics 40, 1977:49-53Minitab Incorporated, Minitab Statsitical Softward. Release 8.2, Microsoft Corporation,Pennsylvania, Copyright 1991.Pagoulatos, E. and R. Sorensen, Foreign Trade. Concentration and Profitability, EuropeanEconomic Review 8, North-Holland Publishing, 1976:255-267Perrakis, Stylianos, Canadian Industrial Organization, Prentice Hall, Scarborough, Canada,1990102Pryor, F.L., An International Comparison of Concentration Ratios,  Review of Economics andStatistics, 54, 130-140, 1972Rapp, W. Firm and Japan's Export Structure,  in H. Patrick (Ed.), Japanese Industrialization andits Social Consequences, Berkeley: University of California Press, 1976, in Jacquemin,Alexis, Imperfect Market Structure and International Trade - Some Recent Research,Kyklos, Vol.35, 1982:75-93Ricardo, David, On the Principles of Political Economy and Taxation, London, John Murray, 1817Richardson, Martin, Antitrust Effects of Export Promotion,  The International Trade Journal,Volume IV, No.14, Summer, 1990:391-400Roberts, Mark J. and James R. Tybout, Size Rationalization and Trade Exposure in DevelopingCountries, in Robert E. Baldwin (ed.), Empirical Studies of Commercial Policy,  NationalBureau of Economic Research, The University of Chicago Press, 1991Saxonhouse, Gary R., Differentiated Products. Economies of Scale and Access to the JapaneseMarket, in Robert C. Feenstra (ed.), Trade Policies for International Competitiveness,The University of Chicago Press, 1989Schembri, Lawrence, Export Prices and Exchange Rates: An Industry Approach, in Robert C.Feenstra (ed.), Trade Policies for International Competitiveness,  The University ofChicago Press, 1989Statistics Canada, Industrial Organization and Concentration in the Manufacturing. Mining andLogging Industries. 1982, Catalogue 31-402P, Industry Division, Supply and ServicesCanada, 1985Statistics Canada, CALURA Corporations, v.61-210, Supply and Services Canada, 1982 and1987United Nations Centre on Transnational Corporations, World Inve1991:stm nt ^The Triad of Foreign Direct Investment, New York: UnitedNations Centre on Transnational Corporations, New York, 1991, in John C. Ries,Regional Studies in International Business Education,  Journal of International Business,Vol. 21, No.1-2, 1992United States Department of Commerce, Bureau of the Census: 1987 Census of Manufacturers,Superintendent of Documents, US GPO, 1992103APPENDIX 1Sample Data PointsCanadian Industries as Described byStandard Industrial Classification (SIC) CodesChemicals and Allied Products1. Chemical and Allied Industries2. Fertilizers3. Pharmaceutical ProductsCommunication Equipment4. Total Communication EquipmentElectrical Products5. Radio and TV ReceiversFood Industry6. Dairy Products7. Meat Products8. Fish Products9. Grain Mills10. Total BeveragesLeather Products11. Total Leather ProductsMachinery12. Total Machinery13. Farm MachineryMetal Industry14. Primary Metal Products15. Total Metal MiningNonmetallic Mineral Industry16. Total Nonmetallic Mineral Products17. Petroleum and Coal ProductsPaper and Allied Industries18. Total Paper and Allied IndustriesPublishing and Printing Industries19.^Total Publishing and Printing Industries104APPENDIX 1(cont'd)Sample Data PointsCanadian Industries as Described byStandard Industrial Classification (SIC) CodesRubber Products20. Total Rubber ProductsTextile Industry21. Total Textile and Knitting MillsTobacco Industry22. Total Tobacco ProductsTransport Industry23. Aircraft and Parts24. Motor Vehicles and Parts25. Miscellaneous Transport EquipmentWood Industry26. Sawmills and Planing Mills27. Veneer and PlywoodAPPENDIX 2Raw and Log Data PlotsCanadian Industry Data - Export Shipments to the World ($000s)PLOT 1Raw Data - Exports 1977:^.• •• • •• • •+^+^+0^2000K^4000K ^+^+^+^exp776000K^8000K^10000KLog Data - Exports 1977: • • •• • :^.^. .:.+^+^+ +^+ +^9.6^10.8^12.0 13.2^14.4 15.6PLOT 2Raw Data - Exports 1982• • • •^•• •+^+^+^+^+^+^exp820 3000K^6000K 9000K 12000K 15000KLog Data - Exports 1982•10.8^12.0^13.2^14.4^15.6^16.8106APPENDIX 2 (cont'd)Raw and Log Data PlotsCanadian Industry Data - Export Shipments to the World ($000s)PLOT 3Raw Data - Exports 1987••• 60^•000 00 00 00^•^• .+^+^+^+^+^+^exp870^6000K 12000K 18000K 24000K 30000KLog Data - Exports 1987•• •^ ••^••^• ••^.. • • :^* : OOO • •^•^-F^+^+^+^+^+^10.8^12.0^13.2^14.4^15.6^16.8107APPENDIX 3Raw Data and Log Data PlotsCanadian Industry Data - Total Value of Industry Assets ($000s)PLOT 1Raw Data - Assets 1977• • • • •^•OOOOO O • • •^• •^+^ass770 2500K 5000K 7500K 10000K 12500KLog Data - Assets 1977• :• : • • ••• • •^•^•••^•^•^+^+^lgass7712.80^13.60^14.40^15.20^16.00^16.80PLOT 2Raw Data - Assets 1982^+^ass820^6000K 12000K 18000K 24000K 30000KLog Data - Assets 1982• •^•^• •• •^•^•^OOOOO • • • • • • • •^• •^• • •^+ l^gass8212.0^13.0^14.0^15.0^16.0^17.0108APPENDIX 3 (cont'd)Raw Data and Log Data PlotsCanadian Industry Data - Total Value of Industry Assets ($000s)PLOT 3Raw Data - Assets 1987• •• • • •^+^ass870^7000K 14000K 21000K 28000K 35000KLog Data - Assets 1987• •^•• • •^•^• • • •^• •^• • • • • •^+-1gass8713.60^14.40^15.20^16.00^16.80^17.60APPENDIX 4Raw Data and Log Data PlotsCanadian Industry Data - Average Firm Size Measured in Assets ($000s)PLOT 1Raw Data - Size 1977^+^siz770^50000 100000 150000 200000 250000Log Data - Size 1977• • •• •^OOOOO • • • • •^• •^lgsiz826.0^7.2^8.4^9.6^10.8^12.0PLOT 2Raw Data - Size 1982• •• •• •• • •• • •• • •• • •^•••••^• .+^+^ +^^+^ + + ^siz820^70000 140000^210000 280000 350000Log Data - Size 1982• • • • •• •^• • •^• • •^• • • • • •^+ -lgsiz827.2^8.4 9.6^10.8^12.0 13.2110APPENDIX 4 (cont'd)Raw Data and Log Data PlotsCanadian Industry Data - Average Firm Size Measured in Assets ($000s)PLOT 3Raw Data - Size 1987.^.+^+^+^+^+^+^siz870 70000^140000 210000 280000 350000Log Data - Size 1987•^••^•^•^ +-lgsiz877.2^8.4^9.6^10.8^12.0^13.2111APPENDIX 5Raw Data and Log Data PlotsCanadian Industry Data - Number of FirmsPLOT 1Raw Data - Number 1977••^+^+^num770^700^1400^2100^2800^3500Log Data - Number 1977•• •^•^••^••••^• • • •• •^• • •  •-+^+^+^+^+^+^lgnum772.4^3.6^4.8^6.0^7.2^8.4PLOT 2Raw Data - Number 1982^+^num820^1000^2000^3000^4000^5000Log Data - Number 1982• •^•• •••^•• • • •^• • ••^••• •• •^+^lgnum822.4^3.6^4.8^6.0^7.2^8.4APPENDIX 5 (cont'd)Raw Data and Log Data PlotsCanadian Industry Data - Number of FirmsPLOT 3Raw Data - Number 1987••••^••• ••••• •••••• ••• •• •^+^num870^1200^2400^3600^4800^6000Log Data - Number 1987.^ ••^••-+^+^+^+^+^+^lgnum872.4^3.6^4.8^6.0^7.2^8.4APPENDIX 6Canadian Industry CALURA ($000's)1982 Concentration Data DescriptionRAW CALURA DATA *N MEAN MEDIAN TRMEAN STDEV SEMEANsales4 13 3330692 1885000 2328545 4550184 1261994assets4 12 4392833 1563500 3705900 4970416 1434835sales8 12 4880334 2533500 3437300 6406566 1849416assets8 12 6535334 2647500 5241400 7832752 2261121LOG CALURA DATAN . MEAN MEDIAN TRMEAN STDEV SEMEANsales4 13 14.460 14.449 14.452 1.081 0.300assets4 12 14.664 14.262 14.763 1.294 0.374sales8 12 14.870 14.745 14.862 1.052 0.304assets8 12 15.042 14.760 15.081 1.247 0.360sales4^= total annual sales of the 4 top industry enterprisessales8 =^,, 8^IIassets4^= total assets of the 4 top industry enterprisesassets8^=^„^8^,,114APPENDIX 7Canadian Industry CALURA ($000s)1987 Concentration Data DescriptionRAW DATAN MEAN MEDIAN TRMEAN STDEV EWANsales4 13 4635000 2632000 3423000 5989000 1661000assets4 12 6700083 2806500 5230900 8534234 2463621sales8 12 6569000 3986000 5171000 7449000 2150000assets8 12 8569917 4638500 7027801 9834992 2839118LOG DATAN MEAN MEDIAN TRMEAN STDEV EWANlgsal4 13 7.826 7.875 7.836 1.195 0.331lgass4 12 8.023 7.928 8.110 1.463 0.422lgsal8 12 8.244 8.285 8.278 1.174 0.339lgass8 12 8.372 8.441 8.477 1.391 0.402sales4^= total annual sales of the 4 top industry enterprisessales8 =^II^ II^8^III^ IIassets4^= total assets of the 4 top industry enterprisesassets8^=^*^*^8^*^*115APPENDIX 8Raw Data and Log Data 1982 CALURA Plots ($000s)PLOT 1Raw Data - Sales4^+^sales40^3500K 7000K 10500K 14000K 17500KLog Data - Sales4^+-lgsal412.80^13.60^14.40^15.20^16.00^16.80PLOT 2Raw Data - Assets4•••^+^assets40^3000K 6000K 9000K 12000K 15000KLog Data - Assets 4^lgass412.0^13.0^14.0^15.0^16.0^17.0116APPENDIX 8(cont'd)Raw Data and Log Data 1982 CALURA Plots (cont'd)PLOT 3Raw Data - Sales8^+^sales80 5000K 10000K 150000K 20000K 25000KLog Data - Sales8.^•^••^••••-+^+^+^+^+^+^lgsal812.80^13.60^14.40^15.20^16.00^16.80PLOT 4Raw Data - Assets8• •• • •^•^•^•^+^assets80^5000K 10000K 15000K 20000K 25000KLog Data - Assets8^+^lgass812.0^13.0^14.0^15.0^16.0^17.0117APPENDIX 9Raw Data and Log Data 1987 CALURA Plots ($000s)PLOT 1Raw Data - Sales4• •^•• •^••• •^•^• . .+^+ ^+ +^ +^+ sale40^5000K 10000K 15000K 20000K 25000KLog Data - Sales4. • • • • • •+^+ ^+ +^ + + lgsal45.0^6.0 7.0 8.0 9.0 10.0PLOT 2Raw Data - Sales8• ••••^•• •+^+^+^+^+^+^sale80^5000K 10000K 15000K 20000K 25000KLog Data - Sales8^. ^ .^: • ••^••+^+^+^+^+^+^1gsal85.0^6.0^7.0^8.0^9.0^10.0APPENDIX 9 (cont'd)Raw Data and Log Data 1987 CALURA Plots ($000s)PLOT 3Raw Data - Assets4• ••• •••^•^+^ass4870^6000K 12000K 18000K 24000K 30000KLog Data - Assets4.• • •^•^+^lgass45.0^6.0^7.0^8.0^9.0^10.0PLOT 4Raw Data - Assets8• • :• •• •^•^+^ass80^6000K 12000K 18000K 24000K 30000KLog Data - Assets8• ••^• • •^•^••^+^lgass85.0^6.0^7.0^8.0^9.0^10.0APPENDIX 10Canadian Industry Export and Concentration Ratio DataRAW DATAN MEAN MEDIAN TRMEAN STDEV SEIVEANCEx 1 24 41.80 32.00 38.18 39.08 7.98cr42 23 39.07 39.04 39.04 3.46 0.72cr82 23 52.72 49.60 51.74 18.66 3.89H2 23 11.29 9.73 10.53 7.48 1.56LOG DATAN MEAN MEDIAN TRMEAN STDEV WEANClgEx 24 3.147 3.465 3.208 1.302 0.266logcr4 23 3.6616 3.6646 3.6619 0.0890 0.0186logcr8 23 3.9073 3.9040 3.9044 0.3467 0.0723logH 23 2.250 2.275 2.231 0.587 0.122CEx = Industry Export Shipments/Total Industry ShipmentsCR4 = Top 4 Enterprises' Shipments/Total Industry ShipmentsCR8 =" 8^"H^= Herfindahl Index Measure (sum of the squares of top 4 enterprises' marketshare)12120APPENDIX 11Raw Data and Log Data Plots1982 Canadian Export and Industry ConcentrationPLOT 1Raw Data - CR4• •• •^•^•• • • • ••^•• ••-+^+^+^+^+^+ -cr432.5^35.0^37.5^40.0^42.5^45.0Log Data - CR4• •^•^• •^• •^• : • • ••^•^+-logcr43.500^3.570^3.640^3.710^3.780^3.850PLOT 2Raw Data - CR8••^• • •^:•• • •^+ cr830^45^60^75^90^105Log Data - CR8. .• • •^•^•• •^• •^•• ••-+^+^+^+^+^+^logcr83.25^3.50^3.75^4.00^4.25^4.50121APPENDIX 11(cont'd)Raw Data and Log Data Plots1982 Canadian Export and Industry ConcentrationPLOT 3Raw Data - Herfindahl• • • •• • • • • • • • • •••^•^+-H6.0^12.0^18.0^24.0^30.0^36.0Log Data - Herfmdahl• • : • • • •^• •^• ^+^logH1.00^1.50^2.00^2.50^3.00^3.50PLOT 4Raw Data - CEx•• •^+^CEx0^30^60^90^120^150Log Data - CEx• •^•• •^• • •^• • • •^.• • • • • • ••--+^+^+^+^+^+ -1gCex0.0^1.0^2.0^3.0^4.0^5.0122APPENDIX 12Normal Plots of Industry Export and Concentration Log Data123APPENDIX 12(cont'd)Normal Plots of Industry Export and Concentration Log Data124APPENDIX 12(cont'd)Normal Plots of Industry Export and Concentration Log DataPLOT 3APPENDIX 12(cont'd)Normal Plots of Industry Export and Concentration Log DataPLOT 4APPENDIX 13US Industry Export and Concentration Data DescriptionRAW DATAN MEAN MEDIAN TRMEAN STDEV SEMEANUSEx 135 7.642 5.483 6.792 7.552 0.650uscr4 140 37.39 34.05 36.41 17.30 1.46uscr8 138 49.53 47.60 49.05 18.67 1.59LOG DATAN MEAN MEDIAN TRMEAN STDEV SEMEAN1gUSEx 135 1.485 1.702 1.556 1.220 0.105lguscr4 140 3.5121 3.5278 3.5269 0.4854 0.0410lguscr8 138 3.8231 3.8628 3.8427 0.4202 0.0358127APPENDIX 14US Industry Raw and Log Data PlotsPLOT 1Raw Data - USCR4• • •^•• ••• • •^uscr42 Points missing or out of rangeLog Data - USCR4 • • •• • • • • • • •-+^+^+^+^+^+^lguscr42 Points missing or out of rangePLOT 2Raw Data - USCR8•• •• •• ••^uscr80^20^40^60^80^100128APPENDIX 14(cont'd)US Industry Raw and Log Data PlotsPLOT 2(cont'd)Log Data - USCR8 OOOOO • • • •• • • ....... • • • •^• •-+^+^+^+^+^+^lguscr82.00^2.50^3.00^3.50^4.00^4.50PLOT 3Raw Data - USEX5 Points missing or out of range•• • •^•• • •• • • •^• • •****** • • •OOOOOO • • •OOOOOO • • • •OOOOOOOOOO • •^• • • • •^• • • •-+^+^+^+^+^+^USEx0.0^8.0^16.0^24.0^32.0^40.0Log Data - USEx5 Points missing or out of range• • • •OOOOOO • • • •• •••---+^+^+^+^+^+- -1gUSEx-2.4^-1.2^0.0^1.2^2.4^3.6129APPENDIX 15US Industry Export and Concentration Data Normal PlotsNormal Plot - Raw Data USCR490+* 2uscr4 -^ 2***22*60+^ 22*24*244236432566230+^*6455633443*3223** * 2*****0+^+^+^normalN=140-3.0^-2.0^-1.0^0.0^1.0^2.0•130APPENDIX 15(cont'd)Normal Plot - Raw Data USCR890+uscr8 -*2****222243*^60+^ 3542256*464644376*253*30+^4222223* *2*** *^+^+^+^+^+^+^normalN=140-3.0^-2.0^-1.0^0.0^1.0^2.0N* = 2Normal Plot - Log USCR8lgcr8 -4.20+3.50+44542466522765435*2452*2** *3*34433423*222*2.80+^* ** *+^+^+^+^+^+^normalN=140-3.0^-2.0^-1.0^0.0^1.0^2.0N* = 2131APPENDIX 15(cont'd)Normal Plot - Raw Data USEx45+ *USEx -30+ **2*222*^15+^ 33*334454265655*235455550+ * * * **2*2223332^+^normalN=140-2.0^-1.0^0.0^1.0^2.0N* = 5Normal Plot - Log Data USEx1gUSEx -^* * * **3222*2*2.5+ 253432*655552356454552430.0+^223332*2** * *-2.5+ *^+^+^+^+^+^normaN=140-2.0^-1.0^0.0^1.0^2.0N* = 5132APPENDIX 16Residual Plots of US Industry Exports and ConcentrationNormal Plot - Residuals of US CR4 Regression Model 72.0+Res USCR4 -^ 333222*225343555520.0+^2565344553243**33**2*-2.0+^**** ****^Nscore-2.0^-1.0^0.0^1.0^2.0Normal Plot - Residuals of US CR8 Regression Model 72.0+Res USCR8 -^ *232222244335655430.0+^3655254523342232**2-2.0+^**2*^Nscore-2.0^-1.0^0.0^1.0^2.0133


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items