culturaldifferencesthatincreasewithphysicaldistance.Apartfrominformationcosteffects,theremaybesecondaryeffectsthataffecttradepatterns.Onlinetradeopensupapotentiallymuchlargergeographicalcatchmentarea,bothforsuppliersandconsumers, with an increase in variety of available products and in pricecompetition.Bothfactorswouldfavourarelativeshiftawayfromofflineandtowardsonlinetrade.However, new sources of information trade costs may arise online that mayslowdown or even reverse this trend. New information costs may be attributable tolinguistic, cultural and institutional differences and the trade costs related totheoperations of e-commerceinfrastructure.
We apply this framework to a unique dataset of cross-border e-commerceingoodsobtainedfromanonlineconsumersurvey(CivicConsulting,2011)inalinguistically fragmented EU market to explore policy options to boost the EU DigitalSingleMarket.Accordingtothe
EuropeanCommission(2012),tenyearsaftertheadoptionoftheEUE-CommerceDirective,e-commerceisstilllimitedtolessthan4%oftotalEuropeancross-bordere-commercetradeandconsidersthatthisisfarbelowitsfullpotential.TheCommission’sDigitalAgendaforEuropeaimstoget50%ofallEuropeancitizenstobuyonlineand20%toengageinonlinecross-bordertransactionsby 2015. The questioniswhetherthe potential forcross-bordertransactions is higher in e-commerce than in offlinetrade.
We investigate three potential sourcesof changes in onlinetradecosts,comparedto offline trade. First, the shift from ordinary offline trade tointernet-enabled online trade may reduce the importance of geographicaldistance-relatedtradecosts.Whiledistancemaynolongermatterforinformationandpurely digital products and services (Blum and Goldfarb, 2006), goods still need tobephysicallytransported,andsometimescrossbordersbetweendifferentregulatoryregimes,toreachthebuyer.Consequently,onlypartofthetotaltradecostsisaffectedby the shift from analogue to digital information technology. Second, we assess therole of cultural and institutional factors, such as language and the quality oflegalinstitutions,asdeterminantsofonlinetradepatterns.Asdistance-relatedtradecostsdiminish,therelativeimportanceofothersourcesofcostsmayincrease.Third,onlinetradingplatformsforphysicalgoodsrequirespecificinfrastructure,suchasflexibleonline payment systems and cost-efficient parcel delivery systems. We gaugetheircontribution to explaining online trade patterns. Finally, we combine all thesesourcesofonlinetradecostsandlookattheneteffectofpositiveandnegativecontributions,asmeasuredbythedegreeofhomebiasorthe“natural”preferenceforhomemarketproducts. The analytical tool that we use for this purpose is the gravity model ofcross-border international trade, the standard workhorse for explaininginternationaltrade flows in the offline economy (Anderson and Van Wincoop, 2003). This modelisrooted in the Newtonian idea that many of the observed patterns of internationaltradeflows can be explained by the economic size of the trading partners and theirphysicaldistance.“Distance”canbemorebroadlyinterpretedas a catch-allvariableandproxyfor various sources of international trade costs that affect the relative price of
domestic and imported goods. This may include physical transport costs, the costassociatedwithimporttariffsandregulatorybarriers,andrisksrelatedtopoorcontract
enforcement between different jurisdictions. In a traditional bricks andmortareconomy,informationretrievaliscostlyandrequiresphysicaltransport,eithertobringinformationtopotentialcustomersorviceversa.Here,wetrytoseparatetheinformation cost from the physical transport cost dimension.
We find that that the importance of geographical distance-related trade costsisindeedgreatlyreducedinonlinetrade,comparedtoofflinetrade.Ontheotherhand,socio-culturalvariablessuchaslanguageincreaseinimportanceandcounterbalancethe declining cost of distance. Moreover, other sources of trade costs gaininprominenceforonlinetransactions,inparticularpaymentsandparceldeliverysystems.Overall,therearenoindicationsthathomebiasislesssignificantonlinethanoffline, if we compare our online results with others in the offline trade literature.Thismaybeduetothefactthatconsumers(inabusiness-to-consumer(B2C)onlinetradesetting) are more sensitive to these new sources of trade costs than businesses (inabusiness-to-business (B2B) offline trade setting) dealing with each other in moreestablishedofflinerelationships.Wearecautioushoweverininterpretingthesefindingsbecausethesupplychainofintermediariesinvolvedinonlinetradeclearlydiffers from those involved in offlinetrade.
2. The gravity model
Since goods still need to be physically transported to the consumer followinganonlinetransaction,wecanassumethattransportcosts remain important in onlinetrade. Online B 2C trade usually implies transport of individual small parcelswhileofflineB2Bmaybenefitfromeconomiesofscaleinlargecargoconsignments.Consequently, physical transport costs for goods bought online could actuallybehigher than offline. On the other hand, the higher number of intermediaries in offlinetrade (wholesalers, importers, etc.) may add to offline trade costs. We have no datatocompare online and offline trade costs between 27 EU member states and thereforelimittheanalysistoonlinetradecostsonly.Weintroduceanexplicitparceldeliverycostvariableinthegravityequationtotesttheimportanceofphysicaltransportcostsfor onlinetrade.
The gravity equation can also handle observations on domestic trade (i = j). Inthatcase,domesticdistance isameasureofthesizeofacountry.Inline withthemethodologyappliedbyPacchioli,2011,McCallum,1995andWolf,2000,weintroduceadummyvariablefordomestictradeobservations.Thecoefficientofthisdummyisanindicatorofhomebias,ortheextentofconsumerpreference
fordomesticoverforeignproducts.Thehomebiasfactoressentiallymeasuresthecombinedimpactofallthevariablesthatdriveonline(oroffline)sales,includinganyomittedvariablesinthegravityequationsuchas“natural”preferenceforthehomemarket.Wecalculatehomebiasonlyforonlinetradesincewehavenoinformationon
domestic sales for offline products. However, we can compare with homebiasestimates for offline trade produced by otherauthors.
3. Data
Weusedatafromanonlineconsumersurveyinthe27EUMemberStates(CivicConsulting, 2011). The survey contains information on consumer online expenditureongoodsonly,athomeaswellasabroad.Weusethesedatatoconstructa27×27bilateralonlinetradematrixfortheEU27.Wealsoconstructanofflinetradematrixbetween the same trading partners and for the same types of goods, so that wecancompareonlineandofflinetradepatterns.Theofflinetradedataareconstructedonthe basis of Comext data for the corresponding online sales product categoriesreported in the consumer survey. For example, when consumers report buyingbooksor pharmaceuticals online, we use the nearest two- or four-digit CN goodscategoryfromtheComexttradedatabase,inthiscaseCN30 “Pharmaceuticals”andCN4901“printedbooks,brochuresandsimilar printedmaterials”tocalculatethevalueofoffline traditional cross-border trade for these goods. Admittedly, these are notperfectmatches but should represent a goodproxy.
A critical issue in the construction of the online matrix is the extrapolation fromsurvey level to population level. Aggregated at the national level, the surveydataproduceanestimateforaverageexpenditureperconsumerin country i on onlinegoods in country j. We assume that the survey average is representative ofonlineconsumer behaviour in country i. We multiply this with a factor that represents theshare of internet users and the share of users who actually buy online in thetotalpopulationtoextrapolatethesurveyaveragetothenationalaverage.WeuseEurostatdata for the percentage of population that is connected to the internet . However,thereisalargedifferencebetweentheEurostatandthe survey figures for the share ofonline consumers who actually buy online and buy online abroad. Since theEurostatfigures(43and10%ofthepopulationrespectively)arelowerthanthesurvey
figures(63and32%respectively),westicktoEurostattoavoidoverestimation.ThesurveyfigureswouldsuggestthattheEUDigitalAgendapolicytargetsofgetting50%ofallEUconsumerstobuyonlineand20?tuallyshoppingonlineabroadhavealreadybeen reached in2011.
Based on the consumer survey, we estimate the total value of online B2C tradeingoods in the EU at 241 billion � in 2011.2 Out of that total, 197 billion � (80%) istraded domestically. Only about 44 billion � (18%) crosses borders betweenEUMember States, and another 6 billion �(2%) is imported from non-EUcountries.
Comparing the value of estimated online cross border trade (44 billion �) andobserved offline intra-EU trade in the corresponding products categories (491 billion
�) (Comext), we conclude that online trade represents about 8.7% of all cross-border
tradeintheEU.Thisindicatesthatonlineordersfortherelevantcategoriesofgoodsconstitute a significant part of physical cross-border trade in goods.
The question arises to what extent the offline and online trade figures areactuallycomparable. On the one hand, offline and online trade involve the sale ofidenticalconsumerproducts:books,electronics,clothing,etc.Thesearefinalproductsandthetrade volume is determined by consumer demand for these goods. However,theorganisationofbothsupplychainsisverydifferent.Offlinetradeismostlyconductedbusiness-to-business(B2B).Wholesalersexportandimportanduseretailersasintermediariesbeforeagoodreachesthefinalconsumer.Bycontrast,onlinetradeismostly B2C, with online wholesalers selling directly to final consumers. Differencesin supply chains may, in turn, result in differences in the structure of the trade coststhatunderpinthetwosetsoftradeflows.Wholesalersoftenhaveestablishedrelationswiththeir
foreigncustomers,withafixedcostthatcanbeamortizedovermanytransactions.Transaction sizeislikelytobelarger,againinducingeconomiesofscale.Offline B2B cross-border trade figures would have to be augmented with retailgrosspricemarginstoproduceatradevaluefigurethatiscomparabletodirectB2Cestimates.TheaboveestimateofonlineB2Crepresenting8.7%ofB2Bcross-bordertrade should therefore be interpreted with caution.
3.1. Cultural and institutionalvariables
Contemporary applications of the gravity trade model routinely include sharedlanguage between trading partners as an explanatory variable, and in most
casesthisturnsouttobesignificant.Thiscouldbeconsideredasaproxyfor“culturaldistance”(Blum and Goldfarb, 2006). In a B2C trading environment a shared languageisessential, though the relative importance of language may vary by type of good. It is likely to be more important for cross-border trade in books for instance, thanforelectronic goods that are more or less standardized across the world. Our dataset
doesnotallowustoseparatetradebytypeofgoodhowever.Wealsointroduceadummyfor the largest and most widely shared language groups in the EU, English, Frenchand German, as another measure of language influence on cross-bordere-commerce.
To measure the role of institutional quality in online trade, we constructanindicator of the quality of the legal system, based on the World Bank dataset ofglobalgovernance indicators. This is meant to capture the differences in expected tradecostsrelated to dispute settlement between importers and exporters in online trade.Onepeculiar aspect of online B2C in the EU is that consumers buying abroad arestillprotected by consumer laws at home, not the law in the exporting country. Thismeansthat consumers do not really have a choice of legal regime in which they carry outtheironlinetransactions.Still,consumersmaynotbeawareofthis;hence,whentheychoose between foreign regimes, they choose the one that looks moretrustworthyafter comparing the quality of both legal systems. A coefficient close to zero wouldindicate that consumers are aware of the legalissues.
3.2. Quality of the online enabling environment
It is important to identify possible trade costs linked to the specificorganisationalneeds of online transactions in goods. Though they may be subsumed in thecatch-all“distance”variableweintroducethreeexplanatoryvariablesexplicitlyrelatedtotheoverall enabling environment for online trade in goods. The first two are relatedtoonline payment systems, the third to transport costs. Since consumers need tohaveeasy access to online means of cross-border payments to settle a transaction atthelowest possible transaction cost. We capture the maturity of online payment systemsintwoways.First,themarketshareofcashpaymentsondeliveryisconsideredtobean indicator of the relative underdevelopment of payments systems, combined withanabsence of trust in online payments and high transaction costs (the transport ofmoney). Compared to credit or debit card payment systems, it is a costly andriskysystem as it involves the transport of large amounts of cash, and transporterandconsumerneedtobeavailableatthesamelocationandatthesamepointintime.Second, the market share of Pay Pal is taken as a proxy of the maturity ofonlinepayment systems whereby consumers trust a non-bank financial intermediary. It mayhoweveralsopointtodeficienciesinthelocal bankingsystemsothatPayPalhelpsconsumers to circumvent these deficiencies. Credit and debit cards are widelyavailableinalmosteverycountryandsupportedbythebankingsystem.Wedonottake the share of credit and debit cards as an indicator. These cards are very commonin all EU countries and their share of transactions is highly negatively correlated withthe previous two variables. In fact, cash-on-delivery and Pay Pal are alsonegativelycorrelated.Toavoidmulticollinearityproblemsweusethesevariablesinseparateregressions. Both cash-on-delivery and Pay Pal indicators are obtained from theWorldPayments Report by CapGemini et al.(2011).
Anefficientparceldeliverysystemneedstobeinplacetophysicallyshipthegoodsfrom their warehouses to the consumer and to minimize physical transport costsanddelivery time. As argued above, the shift from offline to online trade only reduces theinformation cost component of trade costs, not the physical transport cost; onthecontrary,becauseofdiseconomiesofscaleinparceldeliverycomparedtobulkcargo,physicaltransportcostsmayactuallyincrease.Theroleoftransportcostsontradeisnotunambiguousintheliterature.Lendleetal.(2012)showthatthedistancecoefficientisalmostnotaffectedby theinclusionofshippingcosts.InthislineMartinez-ZarzosoandNowak-Lehman
(2007)analyzethedeterminantsofmaritimetransportandroadtransportcostsforexportsand find that distance is not a goodproxyfortransportation costs.Kuwamori(2006)showthatdistanceisimportanttodetermine transport costs, but not decisive. We capture this by introducing a parceldeliverycostindicator:theratio offoreigntodomestic parcel delivery costs, takenfromMeschietal.(2011).Wetakeforeignparceldeliverycostsbycountrypairanddirection of trade. Parcel transport costs are asymmetric for a given country pair.Thedata cover parcel delivery costs by postal services, not commercial courier services.They are officially reported prices, not negotiated price rates for large online retailers.