Hofstra University. Abstract. This paper investigates empirically whether industrial organization-oriented FDI theories explain the recent phenomenon of reverse ...
FOREIGNDIRECTINVESTMENTTHEORIES, ENTRYBARRIERS,AND REVERSEINVESTMENTS IN U.S. MANUFACTURING INDUSTRIES Wi SaengKim* Baruch College-CUNY Esmeralda0. Lyn* Hofstra University Abstract.This paper investigatesempiricallywhether industrial FDI theoriesexplainthe recentphenomenon organization-oriented of reverse foreign direct investmentin the U.S. Based on the distributionof FDI in two-digit SIC manufacturingindustries,we find that capitaland advertisingintensitiesact as entry barriersto foreign investmentsin the U.S. We also observe that foreign multinationalsare attractedby the U.S. marketsize, and that they invest heavily in industrieswith intensive R&D combined with marketingefforts.
The extanttheorieson foreigndirectinvestment(FDI) have primarilyfocused on the investment decisionof the U.S.-basedmultinational As the corporation.' in saw the inflow into U.S. form of 1970s the the foreigncapital directinvestment the need to explainthe motivationsbehindnon-U.S.-based intensifying, firms' in the U.S. arose.Althoughthereate some publishedfieldstudies investments which explainthesemotivations,few attemptswere madeto providelinkage betweenextantFDI theoriesandreverseforeigndirectinvestments.2 The objectiveof thispaperis to investigate whetherproductmarket empirically imperfection-oriented theoryof FDI can explainthe inflowof FDI into U.S. We also examinewhetherfactorswhichact as entry industries. manufacturing barriersto domesticcorporations also applyto foreign-based firmsas well. In addition,we examineif reverseforeigndirectinvestments are concentrated in certainindustries, as was the case for FDI by U.S. MNCs[Gruberet al. 1967 and Mansfieldet al. 1979b].The statisticalanalysisto be conductedin this of U.S. manufacturing studyis based on the characteristics industrieswhich receiveforeigndirectinvestmentsratherthanon the firm-specific attributesof multinational foreign-based corporations. of thispaperis organizedas follows.The secondsectionbriefly The remainder reviewsthe FDI theoriesby U.S.-basedfirmsand of entrybarrierswhichare * AssistantProfessors of BaruchCollege-CUNY, andHofstraUniversity, respectively. The authors are gratefulfor comments from anonymous JIBS referees,as well as Dileep Mehta, Vincent Su and StavrosThomadakis. Received:March 1986; Revised:June, Augustand September1986; Accepted:November 1986.
53
Palgrave Macmillan Journals is collaborating with JSTOR to digitize, preserve, and extend access to Journal of International Business Studies ® www.jstor.org
54
BUSINESS JOURNALOF INTERNATIONAL STUDIES,SUMMER1987
the reversedirectinvestment relevantin drawingtestableimplications regarding activitiesof non-U.S.-based firms.The thirdsectiondiscussesthe hypotheses and modelspecification. Data sourcesand researchmethodologyare presented in sectionfour which is followedby empiricaltest results.The final section givesa summaryof thefindings. FDI THEORIESAND REVERSEINVESTMENTS
The monopolisticadvantagetheoryof FDI advancedby Hymer(1966) and Kindleberger (1969) assertsthatthe multinational corporations possessa rentyieldingasset(for example,productionknow-how)whichgivesthemthe edge in competingwithfirmsin theirhomemarket,as well as withindigenousfirms multinational to thistheory,thenon-U.S.-based abroad.34 According corporations orproductdifferentiation whichenablethemto compete havesuperior technology in marketsaroundthe world.Therefore, it is plausiblethattheywouldoperate in monopolistic industries at homeandabroad.5 whichare createdby both Caves(1971) arguesthatmonopolisticadvantages, characterize andR&D investments, notjust specificfirmsbut rather advertising In fact,Gruberet al. (1967),Caves(1974), industries. firmswithinoligopolistic RomeoandWagner(1979)reported SevernandLaurence (1974),andMansfield, that foreigndirectinvestmenttendsto be associatedwith R&D intensityat the industrylevel.6Knickerbocker (1973) also showedevidencethatthe timing of U.S. MNCs' foreigndirect investmentsis largely determinedby their In a similarcontext,Vernon investments. oligopolisticreactionto competitors' (1974)andGraham(1978)suggestthatreverseforeigndirectinvestments (RFDI) arethereactionsof non-U.S.-based MNCsto theFDIdoneby U.S.-based MNCs. Flowers(1976) suggeststhat the timingof Europeanand CanadianFDI in the U.S. can be explainedby oligopolisticreactions.If this line of reasoning is valid,it is expectedthatRFDIwillbe heavyin R&Dandadvertising-intensive U.S. manufacturing industries,becausethese industriesinvest abroadmore extensivelythanothers. The productcycle theoryof Vernon(1966) suggeststhat new productsare likely to be discoveredand initiallyproducedin the U.S. marketdue to the of the U.S. economy,suchas higherper capitaincome, uniquecharacteristics easeof accessto themarket,andefficientcommunication process.7 Subsequently, otherindustrialized nations'markets(e.g.,Europeor Japan)will be servedby in theseindustrialized export,andthenbe followedby production nations.The production locationwouldultimately move,viaEuropean industrialized countries, to less developedcountries(LDCs)with lowerlaborand/orproductioncosts, andthe U.S. will importfromthesecountries.It is plausiblethat,as the export marketto theU.S.growsandtradeprotectionism increases, foreignmultinationals of newly industrialized countries(e.g., Koreaor Taiwan)may investin the U.S. market,evenif laborcostsaregenerallyhigherin theU.S.thanLDCs.8 Severalstudies[e.g.,Bain(1956), Camanorand Wilson(1967), and Mueller and Rogers(1980)] have concludedthata significantbarrierto entryin U.S. industries is productdifferentiation. manufacturing R&D outlaysand depthof advertising expenditures tend to establishand promoteproductdifferentiation.
FOREIGN DIRECT INVESTMENT
55
If domesticentrybarrieris applicableto foreigndirectinvestment,the R&D U.S. industries andadvertising-intensive areexpectedto receivesmallerportions of RFDI. An additionalentrybarrier,documentedby Scherer(1970), is the amountof capitalnecessaryto establisha plant of minimallyefficientsize. If capital requirements for functionaloperationinhibitforeigndirectinvestment,RFDI inflow will be negativelyassociatedwith the capital intensityof a U.S. manufacturing industry. In view of earlierstudies,this paperinvestigatesthe flow of reverseforeign directinvestmentmeasuredin dollarfigures.As Ajamiand Ricks(1981) and of Commerce(1976) indicate,the sizeof the U.S. market the U.S. Department is a significantfactorfor RFDI. We expectmore RFDI in largerindustries; of incomingRFDI. therefore, industrysizewouldbe a keydeterminant HYPOTHESESAND MODELSPECIFICATIONS
Thesefour variables- monopolisticpower,advertising and R&D intensity, and industrysize - can be consolidatedin a general capitalrequirements, functionalrelationship as: expressed INDFLOWi= ]MPi, IADVi,IRDi, CAPTi,SIZEi,ADVRDi)
where = INDFLOWi MPi= IADVi= IRDi= CAPTi=
(1)
Annualinflowof FDI intoindustryi.9 Measureof monopolistic powerto industryi. Intensityof advertising expensesforindustryi. Intensityof R&Dexpenditures forindustryi. Capitalintensityof industryi.
SIZEi = Size of industryi.
ADVRDi= ADVi*IRDi.Thisvariablewill capturethecombinedeffect of bothadvertising intensityandR&Dintensity. Theliterature on entrybarrierspredictsthattherewill be a negativeassociation betweenINDFLOWandall explanatory variables, forindustrySIZE. controlling Both the monopolisticadvantagetheoryof Hymer(1960) and the industrial organizational theory(or oligopolisticgametheory)of Caves(1971), Vernon (1974), and Knickerbocker (1973) suggestthat the INDFLOWwill be more extensivein R&Dandadvertising-intensive industries. Theresultsof thestudiesby AjamiandRicks(1981),theU.S.Dept.of Commerce (1976),and Franko(1976) revealedthatforeignfirmsareattractedto the U.S. becauseof its marketsize and its marketingand R&D achievements. These observationssuggestthat the combinedeffectsof advertisingand R&D are importantattractions to foreignmultinationals. If theseadvantagesare in fact industry-specific, then research-intensive and advertising-intensive industries shouldattracta substantial portionof RFDI. Thatforeignfirmsareattracted by manyeconomicfeaturesof theU.S.economy, and not in particularby industry-specific resources,is anotherpossibility.If
56
JOURNAL OF INTERNATIONALBUSINESSSTUDIES, SUMMER 1987
so, RFDI inflow would not be systematically associatedwith industryR&D intensity. DATA SOURCEAND METHODOLOGY
The data on foreigndirectinvestmentin the U.S. come from variousissues industries of the Surveyof CurrentBusiness.Data for the U.S. manufacturing are availableonly at a highlevel of aggregation, i.e., for 2-digitSIC industries, startingin 1973. The empiricalanalysiscoveringthe period1974 to 1983 is limitedby the paucityof disaggregated data which,if available,may throw The pricesof light on the finerdifferencesamongreceivingU.S. industries.10 and R&D commonequityand othernecessaryaccountingdatafor advertising AnnualIndustrial are drawnfrom the 1984 versionof the S&P Compustat ratioscome fromthe 1972 Censusof Tape.Four-firmindustryconcentration Manufacturers. ProxyVariables formonopolistic Themostfrequently citedproxyvariables powerarethediscrete concentrationratiosas publishedby the Censusof Manufacturers, and the cumulativeconcentration ratiossuchas the Herfindahl Index.We employthe four-firmconcentration ratioin our analysisfor the followingreasons.First, the literatureindicatesthatthe correlation coefficientsbetweenany variantsof thediscreteandcumulative concentration ratiosrangefrom.91 to .98 [seeBailey and Boyle(1971), and Scherer(1980)].Second,the extracomputational effort in estimatingthe unpublished H-Indexmaynot be worthwhilesincethe results different. maynotbe significantly It hasalsobeenpointedoutby CamanorandWilson(1967)that:"Concentration is simplyone dimensionof marketstructureand is not of itselfa measureof monopolyor markerpower"(p.423).Recentstudies[e.g.,Smirlocket al. (1984), Errunzaand Senbet(1981), and Thomadakis(1977)] have utilizedTobin'sq to measurethe monopolisticpowerof a firm.1'In light of these studies,we use industryaverageTobin'sq-ratioas a proxyvariablefor the marketpower of an industry. Tobin'sq is theratioof themarketvalueof a firmto replacement costsof existingassets.Sincethe marketvalueof a firmvarieswiththe strength of the generaleconomy,the Tobin'sq-ratiofluctuatesconsiderably fromyear to year.12Moreover,becausewe wantto capturethe industryrelativeto U.S. economyaverage,we employindustry relativeq by obtaining theratioof industry averageq to economy-wideaverageq.13
N Thus,q of U.S.economy
N
in whichN standsfortotalnumberof firmsin the Compustat AnnualIndustrial 1984 Tape version,and Rqi
q of industryi q of U.S. economy
(2)
FOREIGN DIRECT INVESTMENT
57
Selectinga proxyvariableforR&Dintensityentailsweighingalternative measures of the R&D effortsof an industry.Commonlyusedmeasuresare 1) totalR&D of totalsalesand 2) the numberof scientistsand as a percentage expenditures engineersas a percentageof total industryemployment[Gruberet al. (1967), RomeoandWagner(1979)].Since SevernandLaurence(1974) andMansfield, R&D outlaysarecommittedoverseveralyears,thereis a learningcurveeffect on R&D productivity such expenditures [Scherer(1967)].Furthermore, yield benefitsin the future[Weiss(1969) and Telser(1961)] and thus,we employ valueof R&Doutlaysandthatof advertising thecapitalized expenditures.'4 To be consistent withearlierstudies[Scherer(1970)andOrr(1974)],we employ FixedAssets Sales as a proxyvariablefor capitalintensity.The proxyvariableshowsthe amount of fixedassetemployedper dollarsales.The naturallogarithmof the industry saleslevel is usedas a proxyvariablefor industrysize in orderto reducethe skewnessof theuntransformed distribution. Regression Models A multipleregression modelis employedto determinethe associationbetween RFDIto U.S.industries andindustry-specific variables. INDFLOWi bo+ biRqi+ b2IADVi+ b3IRDi+ b4CONi + b5CAPT, + b6LogSIZEi + b7ADVRD, + e, (3) Theregression modelof equation(3) maysufferfrommulticollinearity problem, becausestudies[e.g.,BuckleyandCasson1976,Smirlocketal. 1984,andSalinger 1984]indicatethatINDFLOWandRq may be simultaneously determined by R&Dandadvertising intensity.As shownin theAppendix,thesimplecorrelation betweenRq andIADVis .624 and thatof Rq andIRD is .71. However,since our objectiveis to determineif an associationbetweenreverseforeigndirect investment andR&D intensityandadvertising intensityexists,we employtwostageregression to overcomethe multicollinearity procedure problem. = + bo + b3IRDi+ b4CONi INDFLOWi b,Rqi + b2IADVi + b5CAPTi + b6LogSIZE, + b7ADVRD, + e, (4) WhereRqi- Rqi- RqiandRqi = ao + a IADV1+ a2IRDi+ a3CONX. Therefore, Rqirepresents thevalueof Rqiafterremovingfromit theeffectsof concentration ratio, R&D intensity,and advertisingintensity.This approachallows us to determine theassociation betweenreverseforeigndirectinvestment andindustry monopolistic powerfromwhichbothR&Dandadvertising effectsareremoved. We also utilizean alternativespecification,where it is assumedthat the relationshipbetweenreverseforeigndirect investmentand the explanatory variablesis nonlinear. LogINDFLOWi ao + alRq, + a2IADVi+ a3IRDi+ a4CONi + a5CAPT + a6Log SIZE + a7ADVRD+ ui (5)
58
JOURNALOF INTERNATIONAL BUSINESS STUDIES,SUMMER1987 TEST RESULTS
the RFDI statisticsfor the sampleperiod1974-1983.For Table1 summarizes the 10-yearperiod,the total RFDI flow to U.S. manufacturing industrieswas Selectedindustries, over$54billion,withanaverageannualgrowthrateof 17.6%. which make up 84%of the total RFDI in manufacturing industries,had an annualgrowthrateof 17.3%.The chemicalindustryaccountedfor 24.4%of industries. Table1 showsthatthe RFDI flow the RFDIflow to manufacturing is widely distributedamong manufacturingindustries.The non-electrical machineryindustryachievedthe highestannualgrowthrate, 20.10%,in the sampleperiod. of industries Table2 presentsthe characteristics whichreceivedRFDI.Analysis of the meanvaluesof the key variablesshowthatadvertising intensityis higher in boththe chemicalandfoodproductsindustries thanin otherindustries. Both TABLE1
FDIGrowth Rate by Industry,1974-1983 Selected Industries
FDIFlow ($ Millions)
Percentageof Manufacturing Industries
GrowthRate of FDI(%)
Food Products Chemicals Petroleum PrimaryMetals Non-electricalMachinery
6982 13209 13666 4464 7122
12.9 24.4 25.2 8.2 13.3
19.1 18.7 14.4 18.9 20.1
TotalSelected Industries
45543
84.0
17.3
TotalAllManufacturing
54239
100.0
17.6
TABLE2 Industry Means (Standard Deviations) of Key Variables
Industry Food and Kindred Products Chemicals Petroleum Primary Metals Non-electrical Machinery
INDFLOW q mean mean (S.D) (S.D)
IADV mean (S.D)
IRD mean (S.D)
608.2 (657.01)
1.133 (.0913)
.0672 (.0134)
.0145 (.0050)
1320.9 (1076.5) 1366.6 (1037.58) 446.4 (331.047) 712.20 (993.954)
1.596 (.337) 1.139 (.1088) .94527 (.06972) 1.1523 .11226
.0731 (.0038) .00146 .00058 .00127 (.00075) .0125 .00113
.0833 (.0095) .01595 .00441 .02010 (.005523k .056619 .01484
Note: INDFLOW =AverageannualFDIinflowsto industry IADV =(Capitalizedadvertisingexpenses)/Sales IRD =(CapitalizedR&Dspending)/Sales CON =Four-firmconcentrationratio CAPT =Gross FixedAssets/Sales SIZE
=-Industrytotal Sales in $ billion
CAPT mean (S.D)
SIZE mean
.432 (.0597)
18314.4
.40 .4942 (.0298) 36.4 .775 .13679 .422 .664 .1117 .39 .400 .0595
14502.3 106412.2 11520
CON -
.476
-
-
5675.4 -
FOREIGN DIRECT INVESTMENT
59
the chemicalandnon-electrical industries exhibithigherIRD relative machinery The Tobin'sq-ratiois highestin the chemicalindustryand to otherindustries. lowest in the primarymetalindustry.Since Tobin'sq representsthe market costs of its assets,an industry value of a firm in excessof the replacement willexhibita highTobin's withmonopolistic poweror highgrowthopportunities q-ratio.TheTobin'sq-ratiosrangefrom.945 to 1.596. resultsforequation(3) andforthe variousforms Table3 presentsthe regression (models3.1-4.1)we usedto addresspossiblemulticollinearity problems.Table of Rq,theproxyvariableformonopolistic 3 revealsthatthesignof thecoefficients power,is negative,suggestingthat monopolisticpowerof U.S. manufacturing industriesis an entrybarrierto non-U.S.-based MNCs.At the same time, it also shows that concentrationratio is not a significantdeterrentto foreign investmentsin the U.S. The coefficientof capitalintensityis negativeand statisticallysignificantin all regressionmodels, suggestingthat foreign tend to investin industriesnot requiringlargecapitaloutlays. multinationals This findingis consistentwith the view that the capitalintensityneededfor efficientoperationacts as an entrybarrierto foreignmultinationals [Scherer 1970andOrr1974].As suggested by earlierstudies,the coefficientof the SIZE variable is positive and statisticallysignificant,indicating that foreign areattracted multinationals by thesizeof theU.S. market. The coefficientof R&D intensityvariableis positiveandstatistically signiflcant (models3.1-3.3),suggestingthatR&D-intensive U.S. manufacturing industries tendto attractnon-U.S.-based multinationals' in the U.S. However, investments whentheinteraction variable of advertising andR&Disinserted intotheregression estimation(models3.4 and 4.1), the coefficientsof IRD becomestatistically insignificant. But,thecoefficientof theadvertising andR&Dinteraction variable is positiveand statisticallysignificant.This suggeststhat the flow of reverse foreigndirectinvestmentis heavy in industriescharacterized by innovations throughintensivemarketingand R&D efforts.The evidenceis also consistent with Vernon(1974), Flowers(1976), Graham(1978) and the earlierstudies of Commerce1976,AjamiandRicks1981 andFranko [e.g.,U.S. Department 1976]. The regression resultsof model4.1 are obtainedby the two-stageleastsquares method.As depictedby equation(4), we obtainRq fromthedifference between Rq and Rq, whereRq is estimatedby usinga regressionmodel.15It reveals thatthecoefficientof IADVis negativeandsignificant at the5%level,suggesting thatadvertising-intensive industries the flow of reverseforeigndirect discourage in U.S. manufacturing investments industries. Thisindicatesthatadvertising acts as an entrybarrierto foreigndirectinvestment,consistentwith Bain (1956) and Camanorand Wilson (1967).16
Table4 presentsthe regressionresultsin equation(5) and its variousforms (models5.1-5.6)in whichthe associationbetweenRFDI and industry-specific variables is assumedto be nonlinear. Theestimated coefficients of theexplanatory variablesare,in general,consistentwith the resultsin Table3. The regression resultsin Table4 indicatethatabout64%of reverseforeigndirectinvestments to U.S.manufacturing industries areexplainedby theseexplanatory variables.17
60
BUSINESS JOURNALOF INTERNATIONAL STUDIES,SUMMER1987
'
(0 0
0)
a:
Lo
Lo
00
0
cm
a)
cm
(0
00
co
co
cc0
q
I
LO~ LO~
N
00 o -COJ a)o 0) t~~~~~~~
0
C:
cm
.
)
co
w LO~~~~~~~LO C/
_
\
m"iP
00 ~~~~~~q0)-0
ao-
_
C
-
aC)
_
L-O
q '
P
Q.
C6-
-
C CM c\jco00c00 v
E~~~~~o
C6
60
O cm a) Q W ao .2~~~~~~~~~~~~~~.
.0
L
-
a) -
_
_
U
-et
,
?0
0 PI
CY)~ n
,
(0)~~~~~~~~~~~~~~~~0 _ Cy j Cm X c' g~~~~~~o Cm
t Cy It ,
U
0
U
ct
--
CO0
LO
LO8
0
;1
QP 0)
~~~
t~~~
C)~~~~~~~
W
CY)
E.
o,
0 CO
t Cj
co E4 Z CO 0)C
C' Cy)
C OR
:u:u
