A Cross-Section Analysis ofIntra-Industry Trade in the U.S.Processed Food and Beverage

Darcy A. Hartman, Dennis R. Henderson, and Ian

This paper analyzes the determinants of variation across industries in levels of intra-industry

trade (IIT) for a sample of thirty-six U.S. processed food and beverage industries in 1987,

previous studies of intra-industry trade having focussed on industry characteristics in the

manufacturing sectors. The determinants predicted by IIT theory are measures of product

differentiation, economies of scale, and imperfect competition; the results of this analysis

indicate that HT variation across the food and beverage industries is positively related to

product differentiation, economies of scope, and similarity of tariff barriers among trade

partners, but negatively related to industry concentration.

Introduction

Intra-industry trade (IIT), which is defined as theconcurrent importation and exportation of similargoods (Greenaway and Milner, 1986a), has be-come an increasingly important phenomenon in in-ternational trade. IIT was first observed in empir-ical work on the evolution of the European Com-munity (EC) by Verdoom, and Balassa (1965),and since then an extensive literature has shownevidence of IIT in the trade of developed econo-mies (e. g. Grubel and Lloyd; Aquino; and Green-away and Milner, 1984), less developed countries(e.g. Balassa, 1979; Havrylyshyn and Civan) andcentrally planned economies (e. g. Drabek andGreenaway).

In developed countries such as the U. S., UKand Canada, average levels of IIT in the manufac-turing sector, as measured by the Grubel and Lloydindex at the three-digit level of the SITC (see Sec-tion 1 for a discussion of the index), rose from0.44,0.47 and 0.28 respectively in the early 1960sto 0.64, 0.72, and 0.73 respectively in the early1980s (Greenaway and Milner, 1986b; Hart andMcDonald). Also, while it tends to be lower thanin the developed countries, IIT has also grown in

The authors, respectively, are Research Associate, Professor, and As-sociate Professor, Department of Agricultural Economics and Rural So-ciology, The Ohio State University.

The paper is based on research conducted as part of Nonh CentralRegion research project NC- 194 entitled, “The Organization and Per-formance of World Food Systems: Implications for U.S. Policies”. Theauthors would like to thank two anonymous referees for their helpfulcomments on an earlier version of this paper.

the develo~in~ countries.

Sectors

M. Sheldon

for example, the averagelevel of IIT in manufacturing grew in Mexico from0.12 in 1962 to 0.42 in 1987 (Hart and McDon-ald). Therefore, IIT is a phenomenon of growingimportance in the structure of international trade,and one that is both challenging to theorists toexplain and researchers to measure, and of rele-vance to policymakers. These aspects will be ex-amined in the subsequent sections.

(i) Theory of Intra-Industry Trade

Traditional trade theory, which predicts countrieswill specialize in the production and export ofgoods that use their abundant resources and importgoods that use their scarce resources, cannot ratio-nalize the existence of IIT. In recent years, a sub-stantial theoretical literature has emerged that at-tempts to explain IIT (see Helpman and Krugman,1985; Greenaway and Milner, 1986a; and Sheldonfor surveys). These theoretical developments havepredominantly emphasized the existence of imper-fect market structures, economies of scale, andproduct differentiation as the major factors deter-mining IIT. Perhaps the best known and most gen-eral models are those based on a structure of mo-nopolistic competition, the major contributionshaving been synthesized by Helpman and Krug-man (1985), based on the earlier work of Krugman(1979, 1980, 1981), Lancaster, and Helpman(1981).

Essentially, in this type of model, countries areassumed to share the same technology, such that in

190 October 1993 Agricultural and Resource Economics Review

each economy, a perfectly competitive sector pro-duces a homogeneous good under constant returnsto scale, and a second sector produces differenti-ated products under increasing returns. Productdifferentiation is modelled upon the premise thatconsumers in aggregate have a demand for variety,which has been characterized in either of twoways. For example, in Krugman’s work, theSpence-Dixit and Stiglitz approach to product dif-ferentiation is adopted whereby it is assumed thatconsumers derive utility from variety per se, andso consume some part of all differentiated goodsbeing offered by a particular industry. Hence,there is aggregate demand for variety, product dif-ferentiation taking the form of producing a varietynot yet in supply. Alternatively, Lancaster, andHelpman (1981) deal with demand for differenti-ated goods in the spirit of Hotelling’s analysis ofspatial location. Essentially, both approaches gen-erate the same result.

Given the structure of demand, these modelsalso assume that there is free entry by firms intothe market, This generates a structure of monopo-listic competition in equilibrium, while increasingreturns limits the number of differentiated goodsthat can be produced and consumed in one countryunder autarky. Therefore, if trade can take place,and countries have similar factor endowments,each will produce its own supply of the homoge-neous good; whereas, in the differentiated goodssector, economies of scale will ensure that produc-tion of any product will be concentrated in eitherone country or the other. Assuming countries’economies are the same in terms of size, then thestructure of trade will be pure IIT, where eachcountry produces, consumes and exports part ofthe range of differentiated products and imports therest from the other country (ies).

It should also be noted that, once the differen-tiated goods sector is assumed to have a capital-intensive production technology, and differing fac-tor endowments are allowed for, in the extreme,inter-industry trade will be observed, whereby thecapital-endowed country specializes in the produc-tion and export of differentiated products. Hence,Helpman and Kmgman’s (1985) synthesis can beregarded as a proper generalization of the Heck-scher-Ohlin model, such that the theory can predictthat IIT will occur where there is product differ-entiation, economies of scale and market structuresthat are less than perfectly competitive. In addi-tion, IIT is more likely to occur between countrieswith similar factor endowments, a prediction sup-ported in recent empirical work by Helpman(1987) on aggregate trade patterns for fourteen in-dustrial countries over the period 1970-1981.

Although there are other models of IIT (e.g.Brander and Krugman’s model of reciprocaldumping, based on duopoly; and Falvey’s analysisbased on perfect competition and external econo-mies of scale), the analysis outlined above is gen-erally regarded as being the most thorough ap-proach and, as Helpman and Krugman (1989) haverecently commented, ‘‘. . . the positive theory oftrade under imperfect competition has now reacheda certain maturity and acceptance. ” (p. 2).

While the analysis of the causes of IIT hasgrown in the past fifteen years, it is only recentlythat the policy implications have been examined indepth. The literature here addresses two interre-lated aspects. First, there is the issue of the welfaregains from trade liberalization in the presence ofIIT. The analysis outlined above indicates thatthere are likely to be gains from trade over andabove those of exchange and specialization in thetraditional model. Specialization in situationswhere economies of scale exist will generate pro-duction gains, depending on the specification ofthe cost function; if trade widens consumer choice,there are gains from exchange; and the opening upof imperfectly competitive domestic markets toimports may yield further gains, 1 All of thesepoints have been substantiated in recent empiricalwork (see Richardson, 1989, for a survey). Forexample, Cox and Harris have shown in their eval-uation of the U.S./Canada free trade agreementthat the presence of economies of scale generatesconsiderably larger welfare gains than those pre-dicted by neoclassical trade theory. Similar analy-sis by Smith and Venables of the move to greatereconomic integration in the European Community(EC) indicates that in a framework of product dif-ferentiation and economies of scale, the gains fromremoving trade barriers are much larger than thosenormally associated with conventional customsunion analysis,

Second, it has been suggested that if trade is ofan intra-industry nature, then industrial adjustmentto competitive forces will be easier than if tradewere of an inter-industry nature (see Greenawayand Milner, 1986a). Although there have been fewempirical studies to support this contention, thebasis for it is that if industries are characterized byproduct differentiation, then it is easier to adjustproduct lines than it is to undertake the industrialrestructuring implied by inter-industry trade. Thissuggests that either multilateral or bilateral trade

1 As Helpman and Krugman (1985, Ch. 7) note, these gains cannot beguaranteed in general as imperfect competition does not result in anoptimum.

Hartman, Henderson, and Sheldon Intra-Industry Trade in U.S. Food Sector 191

liberalization will generate fewer distributionalchanges than in the traditional model (see Krug-man, 1981), a factor which is important for poli-cymakers to recognize.

(ii) Empirical Analysis of Intra-Indust~ Trade

Along with the theoretical analysis, there havebeen several econometric studies of the determi-nants of IIT (see Greenaway and Milner, 1986a;and Sheldon for surveys). Most of the empiricalwork, the present study included, has set out toestimate a simple, reduced-form regression modelwhere the dependent variable is an index of IIT forindustry i at time t, and the explanatory variablesare a vector of either industry and/or country char-acteristics based on the theory outlined. Most stud-ies estimate this type of regression over a cross-section of industries using either bilateral or mul-tilateral trade data, while some studies, notablyBalassa and Bauwens, have adopted a multi-industry, multi-country framework.

As researchers in this field acknowledge (Green-away and Milner, 1986b), the econometric analy-sis of IIT is difficult, both from a methodologicaland a practical viewpoint. 2 Because there is nocompletely general model of IIT, it has not beenpossible to set up general hypothesis tests as hasbeen the case with the Heckscher-Ohlin model, norhas it been possible to devise means of discrimi-nating between competing hypotheses concerningmarket structure. Consequently, most researchershave resorted to ‘‘. . . ‘identifying’ (using regres-sion techniques) the cross-sectional characteristicsof their data sets, i,e. to describing a range ofsources of influence on IIT that are suggested by,or consistent with, the various models of IIT. ”(Greenaway and Milner, 1986b, p. 5). In addition,as many of the crucial explanatory variables, suchas product differentiation, are notoriously difficultto measure, researchers have had to rely on the useof proxy variables.

However, notwithstanding these problems, the

2 This type of empirical work has recently been subject to criticism byL&wner. Generally he argues that the linkage of the theory of UT to theempirical aualyses is rather casual, and is based on combining the resultsof several dLfferent theories in a single regression medel, However, thiseconometric work is still really in its early stages, and the only empiricalwork that has incorporated the type of utility function described earlier,along with economies of scale and imperfect competition, bas been thecomputable partial equilibrium work of Venables and Smith, and Smithand Venables. In addition, if Schmalensee’s position on inter-industrystudies of structure and performance is adopted, it can be argued thatwhile such studies cannot yield consistent estimates of structural param-eters, they can uncover ‘‘. stable, robust, empirical regulari-ties ,‘’ (p. IWO), rmd while there are data problems in this type ofanalysis, they are not ‘‘. so severe as to render cross-section workvalueless ,‘’ (p, 952).

results of this econometric analysis indicate fairlyrobust and consistent support for the theory as laidout earlier; proxy variables for market structure,product differentiation and economies of scale dohave some explanatory power along with othercontrol variables such as tariffs (see Greenawayand Milner, 1986a),

(iii) Intra-Industry Trade and the Food Industry

A distinguishing feature of this paper is that thefocus is on one sector, the processed food andbeverage industry, as opposed to a broader cross-section of industries in the manufacturing sector.Also, until recently, most studies have, by andlarge, ignored the processed food and beverageindustries, This is due, in part, to a perception thatthe food and beverage industries are perfectly com-petitive. On the contrary, there is evidence that thefood and beverage industries exhibit various im-perfect market structures and produce heteroge-neous goods. (For a thorough discussion of thefood and beverage industries, see Connor, et al.,and Sutton). In addition, IIT has been documentedin the processed food and beverage industries (Mc-Corriston and Sheldon; Christodoulou; Hart andMcDonald; Hirschberg, et al.). For example, Mc-Corriston and Sheldon indicate that in 1986, theEC exhibited a higher level of IIT than the U.S.across a sample of food processing industries,largely as a result of the volume of intra-EC trade.While Hirschberg, et al., in analyzing the deter-minants of IIT in food processing for a thirty coun-try sample over the period 1964-1985 found that itwas a growing phenomenon in this sector and wasa positive function of a country’s GDP per capitaand equality of GDP per capita between countries,thus providing strong support for some of the pre-dictions of the Helpman and Krugman (1985)model.

As gains from trade in the form of greater prod-uct variety, increased realization of economies ofscale, and increased competition are predicted bythe theories of IIT, a priori, itwould seem impor-tant to measure the extent of IIT in the U.S. pro-cessed food and beverage industries, and to exam-ine its causes in these industries using the type ofregression methodology outlined above. WhileHirschberg, et al., have studied the extent to whichcountry characteristics explain the level of IIT inthese industries over time, this study focuses oncharacteristics that affect the level of IIT acrossindustries for a specific point in time for the U.S,

The rest of the paper is organized as follows: inSection 1, the levels of IIT for the various food andbeverage industries are calculated and discussed.

192 October 1993 Agricultural and Resource Economics Review

Section 2 develops a simple regression model ofthe industry characteristics of IIT based on the the-ory already outlined and empirical studies for otherindustries, while in Section 3, the results of cross-section analysis are discussed. Some concludingcomments are made in Section 4.

1. Measurement of IIT

In this study, the Grubel and Lloyd index (GL) isused to measure IIT in the U.S. processed food andbeverage sectors (Grubel and Lloyd). A review ofprevious studies reveals that GL has been the pre-dominant measure used, examples being Toh,Greenaway and Milner (1984), and Balassa andBauwens. The GL index measures the absolutevalue of industry i‘s exports offset by industry i’simports, expressed as a proportion of that indus-try’s total trade:

GL, = (Xi + Mi) _ IXi _ Mil(1) ,

(Xi + Mi)

f)sGLi<l.

GLi corresponds directly to the level of IIT. Whenno trade overlap exists, GLi equals zero. If there iscomplete overlap, GLi equals unity. (See Green-away and Milner, 1986a, for a review of the mea-sure. )

Herein, measurement of GLi is based on theUnited Nations (UN) D-Series Trade Data, where,in order to match this data with industry data re-ported in SIC’s, the SITC codes were converted tofour-digit SIC codes using a concordance devel-oped by Dayton and Henderson. The industrycodes for the food and beverage industries rangefrom SIC 2011 to 2099; Table 1 contains codedescriptions.3 The four-digit classification wasused because it was necessary to minimize the pos-sibility of categorical aggregation, which is the in-appropriate grouping of trade categories for thepurposes at hand, by disaggregating as much aspossible; however, the data used for measuring in-dustry variables could not be disaggregated beyondthe four-digit level for most independent variables.

3 As one of the referees pointed out, the sample of SICS used in theanal ysis omits cectain industries such as, ice cream and frozen desserts(2024), prepared flour mixes and doughs (2045), and others in the SIC.This follows from the fact that the SITC codes do not precisely match theSIC codes, hence several SIC codes have no equivalent SITC code, andtherefore have to be excluded from the sample. In adrfitinn, matchingSIC with SITC codes is never completely precise, and so another re-searcher’s concordance might produce different values for IIT to thoseshown in Table 1. A copy of the concordance between SITC and SICcodes used by the authors is available on request.

The measurement of GLi is based on total U.S.trade with a group of thirty countries. These thirtycountries were chosen due to their consistency inreporting of trade data, In addition, they constitute92% of total world trade in processed food andbeverage products. The data were taken from thereports of the importing countries. Import data aregenerally accepted as more accurate than exportdata since countries tend to be more concernedwith imports for such purposes as the collection ofduties, etc., and since transshipment are not in-cluded.

The estimates of GLi for the food and beverageindustries in 1987 are reported in Table 1; 1987was chosen as it is the most recent year for whichCensus of Manufactures data are available. Theestimates of GLi have a sample mean of 0.329 witha variance of 0.095. The distribution is as follows:4 categories have IIT levels of 0,80-1,00, 3 cate-gories have IIT levels of 0.6W0.79, 5 categorieshave IIT levels of 0.40-0 .59,8 categories have IITlevels of 0.20-0,39, while 16 categories have IITlevels of 0.00-0.19. The industries with almostcomplete overlap are meat packing (20 11), butter(2021), breakfast cereals, and other (2099). Whilea large percentage of categories (44Yo) show al-most no IIT, the majority do have substantial tradeoverlap. These results reinforce other evidence forthe existence of IIT in the food and beverage in-dustries.

2. Determinants of Inter-Industry Variationof Intra-Industry Trade

In choosing determinants of IIT to be tested, someobvious choices are those factors presented in thetheoretical work on IIT which was outlined in thesection above, i.e., a priori, the level of IIT in anindustry will be related to product differentiation,the existence of economies of scale, and imperfectcompetition. Beyond that, reviewing previous em-pirical work in the field yields some additionalsuggestions. The variables described below arethose that were used in the final analysis. Othervariables suggested by the theory and earlier em-pirical work were tested in preliminary analysisand were discarded due to being consistently in-significant; these will be mentioned briefly at theend of this section.

(i) Product Dlfherentiation: As suggested in theopening section on the theory of IIT, product dif-ferentiation is considered by most researchers to beone of the key determinants of IIT. In addition,empirical support for this hypothesis has beenfound in several previous econometric studies,

Hartman, Henderson, and Sheldon [ntra-industry Trade in U.S. Food Sector 193

Table 1. U.S. Trade Figures, 1987 ($000)

SIC Description M, xi GLi

2011 Meat Packing 2563884 19704402013 Sausages 50239 3590882015 Poukt’y & Eggs 238969 47182021 Butter 3822 38452022 Cheese 23117 3753092023 Canned Milk Products 606 200102026 Fluid Milk 2719 41792032 Canned Foods 11860 311822033 Canned Fruits & Vegetables 177169 5517352034 Dehydrated Food 463138 904272035 Pickles, Sauces, etc. o 508652037 Frozen Fruits & Vegetables 101156 1634692041 Grain Mill Products 283205 138742043 Breakfast Cereals 27078 236392044 Milled Rice 142923 7072046 Corn Milling 681896 64052048 Feed Products 273483 918552051 Bread & Pastries o 3161182063 Beet Sugar 1525 82066 Chocolate Products 83772 3886132068 Nuts & Seeds 661881 1210992074 CottonseedOil 17859 34542075 SoybeanOil 726069 418192076 VegetableOil 45033 2351462077 AnimalOil 324896 392572079 Shortening,Margarine 108900 335822082 Beer 41249 8659192084 Wine 49985 9401212085 Liquor 101318 12896612086 Soft Drinks 268 562692091 Canned,Cured Fish 240084 4745392092 FrozenFish 13188662095

3144408RoastedCoffee 81298

2097 Ice706226

0 442482098 Pasta 5826 683372099 Other 284075 317273

Xi = exportsof industryi, M, = importsof brdustryi, GLi = Grubeland Lloydindexfor industryi.

0.8690.2450.0390.9970.1160.0590.7880.5510.4860.3270.0000.7650.0930.9320.0100.0190.5030,0000.0100.3550.3090.3240.1090.3210.2160.4710,0910.1010.1460.0090.6720.5910.2060.0000.1570.945

e.g., Pagoulatos and Sorenson, Greenaway andMdner (1984), and Balassa and Bauwens. Whilethe theoretical work on HT indicates that a struc-tural model would require the estimation of a spe-cific utility function in order to model the demandfor variety, this has proved difficult to do empiri-cally. Hence, this study follows previous empiricalwork on IIT by constructing a proxy index of prod-uct differentiation. Specifically, in this study, theadvertising/sales ratio (AS) was used to character-ize the degree of product differentiation4; this mea-sure is commonly used for this purpose in indus-trial organization research, and has been adoptedin previous empirical work on IIT by Caves,

Greenaway and Milner (1984), Tharakan, andBalassa (1986). A priori, the more money spent onadvertising in an industry, the more differentiatedare the products in that industry, and, hence, thegreater the likelihood of IIT.5 The advertising datawere taken from the Food Marketing Review, andsales data were taken from the U.S. Census ofManufactures. The major problem with the datafor this measure is that the advertising data werenot reported by SIC codes so that there may besome errors in matching the advertising and salesdata.

(ii) Concentration: Most of the theoretical re-search on IIT indicates a role for market structurein understanding IIT. However, as with product

4 An alternative measure for product differentiation, the Hut%auerindex, was tested but found to be insignificant. While this measure basbeen used in several previous studies (e g. Toh; and Bakrssa and Bau-wens), it has been criticized for measuring technological differentiationor differences in inputs.

s In the Spence-Dixit and Stiglitz approach to product differentiation,firms make fixed, recurrent outlays on advertising in order to establish avariety in tbe market.

194 October 1993

differentiation, this tends to be quite difficult tocapture in empirical work, in particular discrimi-nating between competing models based on differ-ing market structures. Nevertheless, several empir-ical studies have used an index of market structure,such as seller concentration, as an explanatoryvariable in analyzing IIT, various theories beingput forward to support inclusion of such a variable.First, if economies of scale exist in an industry, thenumber of firms in the industry is limited, whichmeans that concentration in that industry is likelyto be relatively high. It is generally believed that ifconcentration is high, there is a lack of productvariety; lack of product variety leads to productstandardization in the industry, so IIT should beinversely related to concentration (see Krugman,1979; Lancaster, 1980; Helpman, 1981). Empiri-cal support for this comes from Toh, Balassa(1986), and Balassa and Bauwens.

It has also been argued that concentration, as anindicator of market power, may be associated withreduced emphasis on either exports or imports,which would result in lower levels of IIT (Glejser,Jacquemin and Petit; Lyons). Market power maylimit exports as profits earned on home marketsales act as a disincentive to expending effort onforeign sales. To the extent that market power re-sults from entry barriers, such barriers may dis-courage imports; to the extent that market power isassociated with collusion, home firms may coop-erate to produce at a level that prevents new firmsfrom entering, thus limiting imports. Alterna-tively, as discussed by Brander and Krugman,Toh, and Fung, if high concentration is indicativeof oligopolistic market structures, there may bereciprocal dumping by home and foreign firms,which would generate observed HT. In an effort toprevent new firms from entering, oligopolists willcreate a surplus in the home market and dispose ofthis surplus by dumping it on the foreign market.This being the case, IIT would be positively re-lated to concentration. In order to measure sellerconcentration in the U. S, food and beverage in-dustries, the Herfindahl index (HI) was used, thedata coming directly from the Census of Manufac-tures, HI is measured by squaring the market sharefor each of the top fifty companies in an industryand summing,

(iii) Economies of Scope: Caves has analyzedthe possible impact of economies of scope on IIT.Economies of scope occur when a firm’s averagecosts fall if it produces more than one product.Thus, HT could increase as joint production pos-sibilities increase because firms trade similar prod-ucts. In addition, following the literature on multi-product cost functions (e.g. Baumol et al.), econ-

Agricuhural and Resource Economics Review

omies of scope can have similar effects on marketstructure to those of economies of scale, i.e. witheconomies of scope, only limited numbers of firmsare required to produce a range of varieties thuslimiting the number of producers in an industryunder autarky. 6 It should be noted, however, thatCaves found no empirical evidence to support thishypothesis.

For this study, a variation of Caves’ measurewas used. Specialization (SPi) equals the ratio ofthe shipment of primary products of industry imade by plants classified in that industry to totalshipments by those plants; this ratio is reported inthe U.S. Census of Manufactures. It is hypothe-sized that IIT in the U.S. food and beverage sec-tors is positively related to (1-SPi) = PSi. PSi isthe ratio of the shipment of products of other in-dustries made by plants classified in a specific in-dustry to total shipments by those plants.

(iv) Similarity of Tariff Rates: As well as theabove variables, which are based on the theory ofIIT, most empirical studies incorporate institu-tional-type variables. In particular, it is generallyhypothesized that IIT decreases with an increase intariff rate dbersion. which is the difference be-.tween domestic and foreign tariff rates. Specifi-cally, it is argued that high relative tariffs in onecountry(ies) are a deterrent to IIT, whereas similartariffs on similar products are less of a deterrent toIIT, as long as the tariffs are not prohibitive. Al-though no consistently strong indication of a pos-itive or negative effect of tariffs has been found inprevious studies, e.g., Caves, Toh, Balassa andBauwens, Hirschberg et al., and Pagoulatos andSorenson did find support for this hypothesis.Given the level of protection for the food and bev-erage sectors, it was felt that some form of tariffdispersion measurement was needed as a controlvariable in the analysis.

Unfortunately, comprehensive tariff data aresparse, and recent rates were simply unobtainablefor foreign countries, the measure ultimately usedbeing based on two sets of data.7 The first comesfrom the U,S. International Trade Commission’sPublication 737. which contains measures of U.S.and foreign trade-weighted tariff averages for1970. The second also comes from the Intern-ationalTrade Commission and consists of collectedduties divided by the cost of imports including in-surance and freight (c. i.f, ). The measures were

6 Smith and Venables have explicitly incorporated both scale andscope economies in their analysis of EC integration,

7 While tariff data exists for broadly defined agricultural commodi-ties, data are not available in the necessary SIC classification for the foodand beverage sector in foreign countries,

Hartrnan, Henderson, and Sheldon [ntra-industry Trade in U.S. Food Sector 195

based on information coded by SIC. To calculate aforeign weighted tariff for 1987, the difference be-tween the U.S. tariff rates for 1987 and 1970 wasdetermined for each SIC code; these changes werethen assumed to be the same in percentage termsfor the foreign weighted tariff rates based on therationale that GATT negotiations in the past fifteenyears have tended to involve mutual reductions intariffs.

The measure of similarity of tariff barriers (S)used in the following:

(2) s, =T~s + T~OR– ITT – T~]

(T~s + T~OR)

O=Si<l

where the superscript US refers to U. S, tariffs andFOR to foreign tariffs in industry i. The greater thedegree of tariff dispersion between U.S. and for-eign tariffs, the closer Si will be to zero. A priori,IIT should be positively related to Si, as it wouldbe an indication that countries with similar tariffrates are protecting similar industries; hence, IITwould be likely to occur among these countries.

(v) Other Variables: As stated in the introduc-tion to this section, other independent variableswere tested in preliminary analyses but were foundconsistently to be insignificant. First, a strong de-parture from the theory outlined earlier was theinsignificance of economies of scale as measuredby minimum efficient scale (MES). However,given the small variation of MES among these in-dustries, its insignificance was not surprising.

Second, two variables were tested to determineif IIT could be considered primarily a statisticalphenomenon. The first, categorical aggregation,was a measure of the number of five-digit SICcategories within each four-digit category. Thesecond, seasonality, was a dummy variable at-tempting to capture the possibility of inter-seasonaltrade within an industry, which could be misinter-preted as HT. Both were insignificant, which givesadditional support for the existence of IIT in thoseindustries.

Third, a measure of the degree of integration ofan industry into the global economy was used, de-fined as the ratio of the sum of exports and importsin industry i, normalized by domestic productionin industry i. 8The expectation for this variable was

8This index was suggested to us by one of the referees as an aher-native to the simple sum of exports and impons which we had used in anearlier version of the paper. The latter variable proved statistically sig-nificant, but this may have been due to tbe fact that the same variable isincluded in the denominator of the dependent variable GL,.

that as the index increases, it indicates greater in-ternational integration for the relevant industryand, hence, the greater the likelihood that IIT willexist in that industry. However, the estimated co-efficient on this variable proved statistically insig-nificant.

3. Empirical Methodology and Results

(i) Estimated Model: The model in the followinganalysis was estimated using ordinary least squares(OLS) based on linear specifications for a cross-section of thirty-six U.S. processed food and bev-erage industries in 1987. Other studies have uti-lized variations of OLS such as tobit (Hirschberg,et al. ) and logit (Caves). Tobit was used byHirschberg, et al., because several of the observa-tions for the dependent variable had zero values;the study herein also has zero values for the de-pendent variable, but a preliminary test using tobitdid not offer results significantly different fromOLS. Logit was used by Caves on the basis that,since the dependent variable may be doubly trun-cated (i.e,, upper and lower bounds of 1,0), re-gression analysis needs to restrict the dependentvariable so that the predicted value would adhereto the double truncation; however, there are novalues at the upper limit in this sample.

Ultimately, the equation tested was:

where all variables are defined as above, the ex-pected signs of the estimated coefficients are: a,,a3, a4 > O; a2 > or < O; and Pi is the error term.

(ii) Resul[s: Table 2 reports the results of theOLS regression analysis adjusted for heteroskedas-ticity. 9 The model was significant at the 95% con-fidence level. Approximately 26% of IIT is ex-plained by the determinants included.

Several comments can be made about the re-sults. First, the estimated coefficient of the adver-tising/sales ratio (AS) was positive as expected,and significant at the 95% confidence level. Thisindicates that product differentiation does influ-ence the amount of IIT in the U.S. food and bev-erage sectors. One important note, however, is thatthis variable is heavily influenced by the breakfastcereals industry which has by far the highest ASratio and has a GL measurement of 0,932, almostpure IIT. In fact, when this observation is dropped,AS becomes insignificant; all other independent

9 No multicollinearhy was found for this set of independentvariables,

196 Oc~ober 1993 Agricultural and Resource Economics Review

Table 2. OLS Results

Variable Estimated Coefficient t-ratio R2 Adjusted R2 F

Advertising/Sales 3.3421 1.9025$ 0.3462 0.2619 4.1041Herfindahl Index –0.17165 10-3 –2.8017*Economies of Scope 0.18642 10-’ 2.1660$Tariff Similarity 0.21253 1.9473+

$.Significant at the 95% level.*Significant at the 99% level.

variables are unaffected by the removal of this ob-servation. Second, the estimated coefficient of theHerfindahl index (HI) was negative and significantat the 99$Z0confidence level. While this could beinterpreted in terms of scale economies or productstandardization, the lack of statistical significanceassociated with MES and the undeniable influenceof one observation on the significance of AS lendscredence to the alternative interpretation based onmarket power, i.e., market power discourages IIT.Third, the estimated coefficient of economies ofscope (PS) was significant at the 95% confidencelevel, and was positive as predicted. If industrieshave the ability to produce multiple, differentiatedgoods due to economies of scope, IIT will likelyoccur. Fourth, the estimated coefficient of the sim-ilarity of tariff barriers (S) was positive as pre-dicted, and significant at the 95% confidencelevel. If HT exists in industries where the U.S. andits trading partners have similar tariffs, it is anindication that these countries are protecting sim-ilar industries and are likely to have similar tastesencouraging the existence of these industries.

4. Summary and Conclusions

A large body of research in international trade hasuncovered simultaneous imports and exports ofsimilar goods, While previous empirical studies ofIIT have focused on manufactures, few studieshave concentrated on the U,S. processed food andbeverage sectors, and those that have did not ana-lyze industry characteristics that might explain in-ter-industry variation in IIT. Hence, the aim of thisresearch has been to determine the extent of IIT inthe U.S. processed food and beverage sectors andto find industry determinants of observed llT.

Using a cross-section of SIC’s, the extent of IITin the U.S. processed food and beverage sectorsfor 1987 was estimated using the Grttbel and Lloyd(GL) index. While previous studies (Hirschberg, etal.; Hart and McDonald) have measured HT inthese industries, neither used highly disaggregatedSIC categories. The results of the calculations sup-port the existence of IIT in the U.S. processed food

and beverage sectors. While some categories ex-hibit almost pure lIT, the majority of the categoriestend toward the lower values of the GL index;however, the variation in HT across industries wasconsidered sufficient to warrant further examina-tion.

Based on the theory of HT and previous empir-ical research, a reduced-form model explaining in-ter-industry variation in IIT was developed andtested using OLS. The results show that, for1987, cross-industry variation in IIT in the U.S.processed food and beverage sectors was posi-tively correlated to product differentiation, econo-mies of scope, and similarity of tariff barriers, andnegatively related to industry concentration, Fu-ture research in this area might focus on improvingthe measurement of explanatory variables such asproduct differentiation, and on how changes in in-dustry variables affect changes in IIT over time.Also, an empirical methodology needs to be de-veloped to allow researchers to discriminate moreprecisely between competing theories of IIT.

Finally, given that theory indicates that IIT maygenerate welfare gains over and above those fromconventional comparative advantage (greater vari-ety, greater realization of economies of scale, in-creased competition), some concluding remarkscan be made with respect to the policy implicationsof this research. Specifically, it can be argued thattrade liberalization in industries characterized byIIT is likely to generate greater gains relative tothose industries where little HT occurs. While theU.S. has several institutions in place to promoteexports, e.g. the Export Enhancement Program,imports usually have restrictions placed on themsuch as tariffs and quotas. Hence, if the benefitsfrom IIT are to be realized, then import barriersneed to be removed, e,g. the cheese import quotaregime. In this study, it was found that IIT in thefood and beverages industry is positively corre-lated to the level of similarity of foreign and do-mestic tariff rates, and thus the extent of IIT couldbe increased from both the reduction of U.S. tariffsand equalization of tariff rates between the U.S.and its trading partners. Given the existence ofboth inter- and intra-industry trade, and given the

Harirnan, Henderson, and Sheldon Inma-lndusoy Trade in U.S. Food Sector 197

gains from trade from both, if elimination of bar-riers to trade helps increase both types of trade,then there is all the more justification for reducingtrade barriers. Furthermore, analyses of the impactof trade liberalization should take into account theexistence of IIT in food processing when measur-ing the effects of structural adjustment. This seemsparticularly important in the light of the politicaldebate over the proposed North American FreeTrade Agreement (NAFTA).

References

Aquino, A. ‘‘Intra-Industry Trade and Inter-Industry Special-ization as Concurrent Sources of International Trade in

Manufactures. ” Wehwirtschaj/iches Archiv 114(1978):

275-95.

Balassa, B. Economic Development and Integration. Centro de

Estudios Monetarios Latinoamericos, Mexico ( 1965).

Balassa, B. ‘‘Intra-Industry Trade and the Integration of De-veloping Countries in the World Economy. ” Staff Work-

ing Paper 312, The World Bank (1979).

Balassa, B. “The Determinants of hrtra-Industry Specialization

in United States Trade. ” Oxford Economic Papers

38(1986):22&33.

Balassa, B., and L. Bauwens. ‘‘Intra-Industry Specialization in

a Multi-Country and Multi-Industry Framework. ” Eco-

nomic Journal 97(1987):923–39.

Baumol, W. J., E.E. Bailey, and R.D, Willig. “Weak Invisible

Hand Theorems on the Sustainability of Multiproduct Nat-

ural Monopoly. ” American Economic Review 67( 1977):

350-65.

Brander, J. A., and P. Krugman. “A Reciprocal Dumping

Model of International Trade. ” Journal of International

Economics 13(1983):3 13–21.

Caves, R.E. ‘‘Intra-Industry Trade and Market Structure in the

Industrial Countries.” Oxford Economic Papers 32(1981):

202-23.

Census of Manufactures– 1987. Bureau of the Census. U.S.

Department of Commerce. Industry Series MC87-1-20 A-I

(1990) and MC87-S-6 (1992).Chnstodoulou, M. ‘‘Intra-Industry Trade in Agrofood Sectors:

The Case of the EEC Meat Market. ” Applied .Ecrmomics

24( 1992):875-84.

Connor, J, M., R.T. Rogers, B.W. Marion, and W.F. Mueller.

The Food Manufacturing Industries: Structures, Strate-

gies, Performance, and Policies. Massachusetts: Lexing-

ton (1985).Cox, D., and R. Harris. “Trade Liberalization and Industrial

Organization: Some Estimates for Canada. ” Journal of

Political Economy 93( 1985): 115+5.

Dayton, JR., and D.R. Henderson. “Patterns of World Trade

in Processed Food. ” OP-32, NC-194, The Ohio State Uni-

versity (1992).Dixit, A., and J.E. Stiglitz. “Monopolistic Competition and

Optimum Product Diversity. ” American Economic Re-

view 67( 1977):297–332.

Drabek Z., and D. Greenaway. “Economic Integration and

Intra-Industry Trade: The EEC and CMEA Compared. ”

Kyklos 37( 1984):444-69.

Falvey, R.E. $‘Commercial Policy and Intra-Industry Trade. ”

Journal of International Economics 11(1981):491-511.

Food Marketing Review, 1988. Economic Research Service,

USDA. Agricultural Economic Report No. 614 (1989),

Fung, K.C. “Collusive lntra-Industry Trade. ” Canadian Jour-

nal of Economics 24(1991):391-404.

Glejser, H., A, Jacquemin, and J. Petit. “Exports in an Im-

perfect Framework: An Analysis of 1,466 Exporters, ”

Quarterly Journal of Economics 84(1980):507-24.

Greenaway, D., and C, Milner. “A Cross-Section Analysis of

Intra-Industry Trade in the U. K.” European Economic

Review 25(1984):319+4.

Greenaway, D., and C. Milner. The Economics of Intra-

Industty Trade. Oxford: Basil Blackwell (1986a).

Greenaway, D., and C, Mihrer. “Growth and Significance of

Intra-hrdustty Trade. ” Paper presented at the International

Economics Study Group ( 1986b).

Grrsbel, H. G., and P.J. Lloyd. Intra-industry Trade. London:

MacMillan (1975).

Hart, T., and B .J. McDonald, [ntra-lndustiy Trade Indexes for

Canada, Mexico, and ~he United States, 1962-S7. Agri-culture and Trade Analysis Division, Economic Research

Service, USDA. Staff Report No. AGES9206 (1992).

Havtylyshyn, O., and E. Civan. ‘‘Intra-industry Trade and the

Stage of Development: A Regression Analysis of Indus-

trial and Developing Countries” in Intra-tndustry Trade

edited by P. K. M. Thasrakan. Amsterdam: North Holland

(1983).

Helpman, E. ‘‘International Trade in the Presence of Product

Differentiation, Economies of Scale, and Monopolistic

Competition. ” Journal of International Economics

11(1981):305-40.

Helpman, E, “Imperfect Competition and International Trade:

Evidence from Fourteen Industrial Countries. ” Journal of

the Japanese and International Economies 1(1987):62–81.

Helpman, E., and P.R. Krtsgman. Market Structure and For-

eign Trade. Cambridge, MA: MIT Press (1985).

Helpman, E,, and P.R. Krugman. Trade Policy and Marker

Structure. Cambridge, MA: MIT Press, (1989).

Hirschberg, J. G., I.M, Sheldon, and JR. Dayton. “An Anal-

ysis of Bilateral Intra-Industry Trade in the Food Process-

ing Sector. ” Applied Economics (forthcoming).

Knrgman, P.R. “Increasing Returns, Monopolistic Competi-

tion and International Trade, ” Journal of International

Economics 9(1979):469-79,

Krugman, P.R. “Scale Economies, Product Differentiation and

the Pattern of Trade. ” American Economic Review

70(1980):950-9.

Krugman, P.R. ‘‘Intra-Industry Specialization and the Gains

from Trade. ” Journal of Political Economy 89(1981):

959-73.

Lancaster, K. ‘‘Intra-hrdustry Trade under Perfect Monopolis-

tic Competition. ” Journal of International Economics

10(1980):151-76.

Learner, E.E, “Testing Trade Theory. ” Working Paper No.

3957, National Bureau of Economic Research (1992).

Lyons, B. “An Empirical Investigation of UK Manufacturing

Trade with the World and the E.EC.: 1968 and 1980” in

The Convergence of International and Domestic Markets.

edited by David Audretsch, Leo SIeuwaegen, and Hideki

Yamawaki. North-Holland: E1sevier (1989).

198 October 1993 Agricultural and Resource Economics Review

McCorriston, S., and I.M. Sheldon. ‘‘Intra-Industry Trade and

Specialization in Processed Agricultural Products: TheCase of the U.S. and the EC.” Review of’ Agricultural

Economics 13( 199 1):173-84.

Pagoulatos, E., and R. Sorenson. “Two Way International

Trade: An Econometric Analysis. ” Weltwirtschaftliches

Archiv 11(1975):454-65.

Posner, M.V. “International Trade and Technical Change. ”

Oxford Economic Papers 13( 1961):323-4 1.

Richardson, J.D. “Empirical Research on Trade Liberalizationwith Imperfect Competition: A Survey. ” OECD Eco-

nomic Studies 12( 1989):7–50.

Schmalensee, R. “Inter-Industry Studies of Structure and Per-

formance. ” in Handbook of Industrial Organization ed-

ited by Richard Schmalensee and Robert D. Willig. Am-

ersterdam: North-Holland (1989).

Sheldon, I.M, ‘‘Intra-Industry Trade and Specialization in Pro-

cessed Agricultural Products: Theory and Issues. ” OP-2,

NC- 194, Ohio State University (1992).

Smith, A., and A. Venables. “Completing the Internal Market

in the European Community-Some Industry Simulations. ”

European Economic Review, 32(1988): 1501-1525.

Spence, A.M. “Trade and Trade Policy with Differentiated

Products. ” Review of Economic Studies 43(1976):217-36.

Sutton, J. Sunk Costs and Market Srructure. Cambridge, MA:

MIT Press (1991).

Tharakan, P.K.M. “Empirical Analyses of the Commodity

Composition of Trade. ‘‘ in Current Issues in International

Trade edited by David Greenaway. London: Macmillan.Toh, K. “A Cross-Section Analysis of Intra-Industry Trade in

U.S. Manufacturing Industries, ” Wekwirtschaftliches Ar-

chiv 118(1982):282–301.

United Nations External Trade Data. Statistical Office of the

United Nations, New York (1989),

Venables, A., and A. Smith. “Trade and Industrial Policy un-

der Imperfect Competition. ” Economic Policy 3(1986):

622-72,

Verdoom, P.J. ‘‘The Intra-Bloc Trade of Benelux” in Eco-

nomic Consequences of Nations, Proceedings of a Confer-

ence Held by the International Economic Association.

London (1960).