This article was downloaded by: [Umeå University Library]On: 06 October 2014, At: 20:29Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Economic JournalPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/riej20

International Service Transactions: IsTime a Trade Barrier in a ConnectedWorld?Bianka Dettmera

a Friedrich-Schiller-University Jena, Jena, GermanyPublished online: 09 Sep 2013.

To cite this article: Bianka Dettmer (2014) International Service Transactions: Is Time aTrade Barrier in a Connected World?, International Economic Journal, 28:2, 225-254, DOI:10.1080/10168737.2013.825305

To link to this article: http://dx.doi.org/10.1080/10168737.2013.825305

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with, in relation to or arisingout of the use of the Content.

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

International Economic Journal, 2014Vol. 28, No. 2, 225–254, http://dx.doi.org/10.1080/10168737.2013.825305

International Service Transactions: IsTime a Trade Barrier in a Connected

World?

BIANKA DETTMER

Friedrich-Schiller-University Jena, Jena, Germany

(Received 6 September 2012; in final form 18 June 2013)

ABSTRACT Firms’ international fragmentation of production has recently widened itsfocus from outsourcing of intermediates to off-shoring of business services such as soft-ware program development and international call center networks. Although many servicesare intangible and non-storable, gravity model estimates show that geographical distancebetween business partners matters less for commercial service transactions. Rather, timezones can be a driving force of international service trade by allowing for continuous oper-ation over a 24 hours business day (continuity effect) when a proper division of labor isfeasible and countries are connected to electronic communications infrastructure (ICT). Buteven when ICT provides alternatives for face-to-face interaction, time zones can act as abarrier when coordination problems with sleeping business partners occur (synchronizationeffect). In this paper, we find empirical evidence for the continuity effect in trade of businessservices, which is robust to measurement and sample size. Even more important is that theeffect of time zones on service trade depends on access to ICT. An improvement of ICTinfrastructure will affect business service trade at long time zone distances significantly morethan trade at short time zone distances.

KEY WORDS: international trade, business services, gravity model, distance, time zones, electroniccommunicationJEL CLASSIFICATIONS: F10, F14, F20

Corresponding Address: Bianka Dettmer Friedrich-Schiller-University Jena, Chair of EconomicPolicy, Carl-Zeiss-Strasse 3, Jena, 07743 Germany. Email: bianka.dettmer@uni-jena.de

© 2013 Korea International Economic Association

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

226 B. Dettmer

1. Introduction

The debate on firms’ international fragmentation of the production chain hasrecently widened its focus from outsourcing of intermediates to off-shoring ofbusiness services. This is evident in international trade flows: producer and busi-ness services have recently grown faster than merchandise trade. Since manyservices are intangible and non-storable, and thus, require that production andconsumption often need to appear simultaneously (e.g. in management consult-ing or tax advisory), services are traditionally regarded as non-tradable. However,innovations in information and communication technology (ICT) seem to renderthe proximity requirement for face-to-face interaction between business part-ners useless. Thus, software program development occurs with colleagues locatedin India, and call center networks around the world provide customer support24 hours, seven days a week. An emerging body of literature focuses on inter-national service transactions (e.g. Grünfeld & Moxnes, 2003; Kimura & Lee,2006) and applies the gravity model, which has been found useful to explainthe location of manufacturing production and FDI. While in the manufactur-ing trade it is argued that the magnitude of the distance effect is still higherthan transportation costs would suggest (in times where transport costs such asair freight rates are rather decreasing),1 geographical distance has an ambigu-ous effect on service trade. Blum and Goldfarb (2006) show that even whentransportation and distribution costs are near zero, distance remains relevantwhen products and services are taste-dependent (e.g. music, gambling, games,and pornography). For non-taste-dependent services (e.g. general information,software, technology and financial information) distance does not representa barrier any more. Although some authors find a significantly negative dis-tance effect in services trade (Ceglowski, 2006; Grünfeld & Moxnes, 2003;Kimura & Lee, 2006; Mirza & Nicoletti, 2004), it is shown that distance doesnot affect transactions in commercial services (Walsh, 2008),2 software services(Tharakan & van Beveren, 2003; and Tharakan, van Beveren, & van Ourti,2005), and communication and financial services (Lejour & de Paiva-Verheijden,2007).

The services sector is very heterogeneous and transaction costs for businessservices need not to be relevant for transportation, travel and tourism services. Incontrast to the broad category of commercial services (which include communi-cation, construction, insurance and financial services, computer and informationservices, royalties and license fees), the specialized business services sector (i.e.leasing, legal, accounting, auditing, book-keeping, tax consulting, business andmanagement consulting, architectural, engineering and other technical services,advertising, and research) belongs to a higher extent to the manufacturing value-added chain (OECD, 2008). Jones and Kierzkowski (1990) argue that these

1See Grossman (1998) and Loungani, Mody, & Razin (2002) for a review of the debate.2The same result holds for travel and government services. In transportation services the distancecoefficient is significantly positive (Walsh, 2008). See also Lennon (2009).

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 227

services are essential to connect fragmented production blocks. In this respect,it has recently been argued that time is a determinant in the location of produc-tion, especially when just-in-time technology is introduced (Evans & Harrigan,2005; Harrigan & Venables, 2006). The role that time plays for internationaltransactions is found to be ambiguous: for the location choice of multinationalenterprises, it is shown that time zones have a negative impact on productivity,because air-traveling in an east–west direction is associated with the jet lag effect.

On the one hand, Stein and Daude (2007) argue that time zones act as a bar-rier even when electronic communication (email and virtual conferences) is anexcellent substitute for face-to-face interaction, because coordination problemswith sleeping business partners occur (synchronization effect). This implies thatlocating business and customer support services in close locations makes sense toensure that time zones allow for simultaneous interactions. On the other hand,Marjit (2007) and Kikuchi and Iwasa (2010) model time zones as a driving forceof service trade and argue that, with few interruptions, business operations cancontinue when a proper division of labor is feasible and countries are connectedvia ICT (continuity effect). They build their time zone argument on operationsin the software industry where programming problems will be emailed from theUSA to India at the end of the day (in the USA) and an Indian software specialiststarts working in her regular office hours when the office in the USA remainsclosed. But the continuity effect can be important for simultaneous interactionsbetween business partners as well, since call center networks provide customersupport outside the customer’s normal working hours.

In this paper, we empirically investigate the argument that time zones are adriving force of business and commercial services trade. In contrast to previousstudies on time zones, we follow the theoretical contribution by Marjit (2007)and Kikuchi and Iwasa (2010) and show that the access to ICT is a preconditionto make use of time zone differences for business services trade. We apply a Tobitmodel to show that the effect of time zones on business services trade dependson the access to information and communication technology. We account forcountries with multiple time zones and apply various measures of distance intime. The research question is relevant for the following reasons: while empiri-cal research focuses increasingly on international service transactions since databecome available, evaluations on trade barriers in business services are scarce.Since specialized services represent an ‘experienced good’ for which the qual-ity cannot be evaluated in advance, the internationalization of business servicesseems to be far more difficult. Many business services firms tend to follow their(manufacturing) customers into global markets (Raff & von der Ruhr, 2007) andare requested to provide support around the clock. In this respect, time zonesmay affect international business service transactions by more than commercialservices trade in general.

The paper is organized as follows. In the next section we derive testable hypoth-esis from a review of the literature on the link between time costs and ICTinterconnectivity in the delivery of commercial- and business services. Section 3presents methodology and data sources. Empirical evidence is discussed insection 4. The last section concludes the paper.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

228 B. Dettmer

2. Economic Theory: Time as a Trade Barrier

The debate on the distance effect in gravity models has prompted a reassessmentof trade costs in international transactions. Instead of estimating transportationcosts in a single distance variable, the literature distinguishes trade costs intofinancial and time dimensions. This is an important step since transportationcosts such as air freight rates decrease (Duranton & Storper, 2005; Hummels,1999), but delivery time has become increasingly important in fragmented pro-duction chains, especially when just-in-time technology is introduced.3 Time todeliver to the market has two effects on trade: first, time can be an entry bar-rier, and second, time can be considered as trade costs.4 A frequently cited paperin this field is Hummels (2001). He finds that time costs associated with the netshipping time between trading countries’ ports reduce the probability that a coun-try will export time-sensitive manufactures. Djankov, Freund, and Pham (2010)argue that a significant part of time costs stems from moving goods from the fac-tory to the ship, which involves time delays due to administrative procedures (e.g.customs and tax procedures, cargo inspection) at the port (see also Hausman,Lee, & Subramanian, 2005), in addition to the quality of physical infrastructurein the country (Limao & Venables, 2001). Thus, when time for export is con-sidered, geographical transaction costs matter less (Nordas, Pinali, & GelosoGrosso, 2006).

Although the physical transportation of goods requires time, some servicesexhibit characteristics that render the time costs associated with delays in trans-portation useless. Services are mostly intangible and non-storable, and thus,require that supplier and consumer are physically located in the same place. How-ever, when telephone, email and virtual conferences become close substitutes forface-to-face interactions, geographical distance should not represent a barrier forinternational service transactions. In this respect, Portes and Rey (2005) arguethat a negative distance effect found in international equity flows can be inter-preted as a kind of information cost.5 When applying the same logic as before,information costs can be split into financial and time elements.

With respect to the time element, the proximity requirement for the deliveryof services rather suggests that a part of the information costs may stem fromtime zone differences between business partners, as one way to provide a serviceand communicate face-to-face is to travel abroad. However, air traveling in aneast–west direction incurs the jet lag effect and requires time to adjust to timedifferences.6 To the extent that electronic communication replaces face-to-face

3See Evans and Harrigan (2005) and Harrigan and Venables (2006) for theoretical models onjust-in-time production and agglomeration.4A high variability in delivery time will prevent firms from entering the market as they will not beshortlisted for contracts that require just-in-time delivery (Nordas, Pinali, & Geloso Grosso, 2006).5See also Buch (2005) for international banking.6According to Paulson (1996), the symptoms of jet lag become important with time zone changesof 5 hours or more. Traveling in both directions causes time to re-establish a circadian equilibrium:traveling from east to west stretches the traveler’s day while traveling from west to east compressesit. However, the time to re-establish is greater with flights eastward than westward. In a similarmanner, Kamstra, Kramer, and Levi (2000) find that sleeping disorders due to daylight saving timechanges affect the response time and problem-solving ability of stock market participants.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 229

interaction, the need to travel should become less relevant. Nevertheless, whenbeing located in distant time zones, coordination problems with sleeping businesspartners occur. In a study on the location of FDI, Stein and Daude (2007) showthat time zones act as a barrier when frequent real-time communications (synchro-nization effect) between headquarters and their foreign affiliates are important(see also Hattari & Rajan, 2008; Zaheer, 2000). This implies that customer sup-port services would be located in the near distance to ensure that time zones allowfor simultaneous real-time interactions.

In trade theory, only a few attempts have been made to consider the role oftime zones for trade.7 Marjit (2007) models time zones as a driving force of inter-national services trade. When production of a service is fragmented and takestwo days to be finished, time can be saved if countries located in separate timezones engage in production, each performing one step (sequential interactions).The time zone argument is derived from operations in the software industry.Software problems in the US can be emailed to India at the end of the day (inthe USA). An Indian software specialist is working when the offices are closed inthe USA so that the programming codes can arrive as email attachments whenthe US-offices reopen. A similar argument is proposed by Kikuchi and Iwasa(2010). They assume positive time costs for the delivery of services and considerthat a country’s consumers, who would like to get the service sooner rather thanlater, can import the service cheaper, if there is a significant time zone differ-ence between both countries. Kikuchi (2003, 2009) finds that time zones affectthe structure of comparative advantages when services are used as an intermedi-ate good. Thus, the existence of time zones allow that with little interruptions,business operations can continue if the office locations are in significantly distanttime zones (continuity effect). The first empirical study on the time zone effectin cross border services trade was Head, Mayer, and Ries (2009). They confirmthat the ability to operate around the clock (continuity effect) offsets the needto communicate during business hours (synchronization effect). Christen (2011)finds similar results with respect to service FDI: time zone differences have a sig-nificant positive effect on foreign affiliates’ sales in services. Being further away interms of time zones raises affiliates’ sales compared to the reference group withno time zone difference.

Nevertheless, this type of trade requires two basic preconditions to be satis-fied (Marjit, 2007). First, the time zone difference can be utilized when a properdivision of labor is feasible in the sense that working during regular office hoursis possible. The continuity effect of time zone differences which is employed incases of sequential interaction between business partners can be important forcases of simultaneous interactions as well, since call center networks providecustomer support outside the customer’s normal working hours. Thus, locatingbusiness support services in time-distant countries can be beneficial for servicesfirms, since it allows ensuring that support will be provided without (less produc-tive) shiftwork in the home country. Therefore, we expect that the continuity effect

7Evidence for importance of latitude is already considered in the literature on geography andeconomic development (Acemoglu, Johnson, & Robinson, 2001; Gallup, Sachs, & Mellinger, 1999;Hall & Jones, 1999). The comparative advantage effect of latitude is shown in Melitz (2004).

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

230 B. Dettmer

is more important than the synchronization effect. In line with this argumentationwe test the following hypothesis:

Hypothesis 1. Business service trade increases with time zone differences between businesspartners (continuity effect).

The second precondition to be satisfied is related to the financial element of infor-mation costs: relevant infrastructure has to enable the service to be transportedwith little cost. Keller and Yeaple (2013) model the firms’ choice to transfer intan-gibles and focus on the trade-off between communicating knowledge from oneperson to another versus the costs of moving knowledge embodied in goods. Theyargue that a multinational firm tends to shift its mix from disembodied (directcommunication) towards embodied (trade) knowledge transfer since communi-cation costs (which vary with the degree of knowledge codification) rise fasterwith the knowledge intensity of industries than trade costs (which vary with dis-tance). The empirical evidence confirms that the average knowledge intensity oftrade increases with distance.

Since one way to provide services is to travel abroad, a few authors have shownthat lowering the costs of international business travel (e.g. by a less restrictive for-eign visitor policy) facilitates trade of goods and services by allowing for increasedface-to-face interaction between business partners (Belenkiy & Riker, 2010;Poole, 2010; Yasar, Lisner, & Rejesus, 2012). Another way to provide businessservices is to make use of electronic communication. In this respect, communica-tion costs – measured by, for example, a standard residential rate for internationalcalls – have decreased substantially (Tang, 2006) and allow for the volume of bilat-eral telephone traffic to be extended. Fink, Mattoo, and Neagu (2005) find thattrade in differentiated goods for which significant buyer–seller interaction is nec-essary is more sensitive to telecommunication prices. Similar results are obtainedby Portes, Rey, and Oh (2001) and Portes and Rey (2005) for financial flows: byincluding the volume of bilateral telephone traffic, geographical distance doesnot matter for transactions of homogeneous treasury bonds but remains relevantfor transactions in heterogeneous financial assets (corporate bonds and equities).Loungani, Mody, and Razin (2002) extend the analysis by adding telephone trafficto bilateral FDI flows. However, as bilateral telephone traffic does not reflect thecountries’ capacities to communicate, the focus of analysis shifted to the (qualityof) telecommunications infrastructure as a condition for international transac-tions. Freund and Weinhold (2002) find a significant positive effect of internetadoption abroad on US service trade growth, which is somewhat stronger in asubsample of business, professional and technical services.8 According to Marjit(2007), ICT is a precondition for the ability to make use of the continuity effect oftime zones in international business services. Therefore, we expect Hypothesis 2.

Hypothesis 2. The effect of time zones on bilateral business service trade depends on theavailability of ICT infrastructure.

8Choi (2010) confirms the result. With a higher level of ICT infrastructure, service trade increases inOECD countries as well (Mirza & Nicoletti, 2004). See also Lennon (2009) for commercial services,Tang (2006) for merchandise trade and Loungani et al. (2002) for FDI.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 231

We test the hypothesis with an interaction term of time zone differences and ICTinfrastructure, which is expected to be positive for business and commercial ser-vices trade. When ICT infrastructure is important in making use of the continuityeffect in business services trade, we expect that an increase in the access rates toICT will affect business services trade over longer time-zone distances by morethan over shorter time-zone distances. Moreover, there has been much discussionthat with emerging ICT a technology capacity gap is arising – the digital divide– between information-rich and information-poor countries. Further digitaliza-tion within the country increases the network capacity by lowering the marginalcosts of additional users connected with the network. Since access rates to ICTinfrastructure are highly developed in OECD countries, the marginal utility of anadditional unit of ICT for trade of business services is expected to be decreasing.In contrast, information-poor countries in turn face the threat of being left furtherbehind (Hanafizadeh, Saghaei, & Hanafizadeh, 2009 and references therein). Inthis respect, the degree of interconnectivity between two countries is often deter-mined by the country with the poorer access rates to ICT. Thus, it is reasonablethat the marginal effect of an increasing ICT access on business service trade isstronger in trade with less developed non-OECD countries than in service tradewith OECD partner countries.

3. Empirical Investigation

3.1 Model Specification

The gravity model has been widely used for explaining bilateral trade flowsbetween countries since Tinbergen (1962). The log-linear specification of thegravity equation relates nominal bilateral trade flows from exporting country i toimporting country j to the economic masses (GDP) of the trading partners andto the distance between them, whereas distance proxies for transportation costs.Even though the gravity model first emerged as an empirical relationship, theo-retical underpinning appears (Anderson, 1979; Bergstrand, 1985, 1989). Recently,there have been various attempts to develop structural gravity equations. Thetheoretical foundation can be distinguished in three broad approaches. The firstmodels derive the gravity equation based on product differentiation by the countryof origin (Anderson & van Wincoop, 2003). Thus, countries produce a differen-tiated bundle of products in the country of origin and send them across borders,where it enters the utility of a consumer. The second strand of theoretical foun-dation, which considers product differentiation and monopolistic competition,builds on a similar argument (Krugman, 1980). The third approach builds thegravity equation on models with homogeneous products and heterogeneity inproductivity (Eaton & Kortum, 2002). In contrast to the transaction costs asso-ciated with the shipment of goods, services can also be traded electronically dueto their intangible nature. When including transaction costs that are relevant forservices trade, the structural gravity equation derived from models explainingmerchandise trade, appears to be applicable to trade in services. The follow-ing gravity model specifications are used to evaluate time as a trade barrier for

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

232 B. Dettmer

exporting business and commercial services:

log(EXPijt) = β0 + β1 log(Distij) + β2TIMEij + βk

∑k=i,j

log(GDPkt−1)

+ βm∑

m

Xijm + γi + γj + γt + εijt (1)

where EXPijt is the export of (business or commercial) services from home countryi to host country j at time t. Market size of home and host country is measuredby the log of nominal GDP at time t − 1 to proxy for services demand. We usethe GDP with a one-year lag to mitigate the endogeneity because exports (andimports) are a component in value added and GDP. By controlling for distancebetween countries in the regression, we add TIMEij to the equation, which ismeasured by the differences in hours between the countries’ capitals. We performrobustness checks by a variety of measures in time zone differences, which isexplained in detail in the following section. The bilateral control variables inXijm include a dummy for colonial relationships, English language, and culturalsimilarity respectively. In addition, we add a dummy that covers the presence ofa free trade agreement of a trading pair when services trade according to GATSArticle V is explicitly considered.9

In a second model, we test the impact of time as a trade barrier when thecountries’ degree of connectivity to an information and communication network(ICT) is taken into account. While previous studies use telecommunication pricesand telephone traffic to quantify information costs, we focus on the ICT infras-tructure present in the previous period. Telecommunication costs and bilateraltelephone traffic can be a consequence of increasing international transactionsand, thus, may be endogenous to international trade. In contrast, the presenceof ICT infrastructure in the country better reflects the countries’ capacities tobe involved in international transactions. The theoretical literature suggests thatbeing connected to an ICT infrastructure network is an important preconditionfor the continuity effect of time zones (Kikuchi & Iwasa, 2010; Marjit, 2007).But the degree of interconnectivity between two countries is often determined bythe country with the poorer ICT infrastructure. For this reason we include theminimum ICT infrastructure (min[ICTit−1; ICTjt−1]) and estimate the followingequation:

log(EXPijt) = β0 + β1 log(Distij) + β2TIMEij + β3 log(

mink=i,j

[ICTkt−1])

+ β4TIMEij∗ log

(mink=i,j

[ICTkt−1])

+ βk

∑k=i,j

log(GDPkt−1) + βm∑

m

Xijm + γi + γj + γt + εijt (2)

9Grünfeld and Moxnes (2003) argue that the presence of a free trade agreement (FTA) for goods isinsignificant in services trade as mostly FTAs do not include trade in services explicitly.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 233

The coefficient on the interaction term of time zones and minimum ICT infras-tructure network, β4, indicates whether the effect of time zone differences ontrade depends on the ICT network. A country pair in which the partner withpoorer ICT has a more advanced infrastructure access will trade more over longer(or shorter) time zone distance than a country pair in which the partner withthe poorer infrastructure has a worse ICT network. The interpretation of theinteraction effect depends on the degree of buyer–seller interaction dominant incross-border trade. When the continuity (synchronization) effect is important weexpect that ICT matters more at longer (shorter) time zone distances. In thisrespect, we derive the marginal effects of an increase in time zones and minimumICT on business service trade at various points of the cumulative distributionto test whether ICT equipment is more essential at longer or shorter time zonedistances. The marginal effect indicates the amount of change in business servicetrade with a one unit change in time zones while holding the level of ICT inthe country with the poorer ICT infrastructure constant at various percentiles ofthe cumulative distribution of minimum ICT. Correspondingly, we estimate themarginal effect of an increase in minimum ICT on business service trade at eachhour time zone difference between countries.

When hypothesizing that the degree of interconnectivity between two coun-tries is determined by the country with the poorer ICT infrastructure, we expectthat countries will select into service trade. An empirical problem that ariseswhen using the log of service exports as a dependent variable is how to deal withrecorded zero trade values. Although our data set includes a relatively small num-ber of observations where business service trade would be dropped by taking logs(about 12%), the problem is that those observations can carry important infor-mation with respect to selection into trade. Thus, zero trade can be the result of apoor ICT network so that countries are not able to participate in services trade.Some authors simply exclude the observations in which the dependent variabletakes a value of zero. This would lead to an estimation bias. Instead, we use theTobit model, which assumes that the dependent variable is censored to the left toaccount for zero trade flows and the selection of countries into service trade.

In an influential paper, Baldwin and Taglioni (2006) describe in detail what theycall bronze, silver and gold medal mistakes in applying panel data. As it is alreadyknown as ‘the multilateral resistance term’, Anderson and van Wincoop (2003)argue that bilateral trade between any two countries also depends on remoteness,i.e. each country’s resistance to trade with all other trading partners. Neglect-ing the ‘multilateral resistance term’; and estimating trade flows depending onbilateral distance and market size creates an omitted variable bias as bilateraltrade costs with all other countries will be included in the residual. One wayto account for the multilateral resistance term is to include relative prices andto estimate gravity equations with real GDP values and deflated nominal tradeflows. An inappropriate deflation can create spurious correlation owing to globalinflation trends (the bronze medal mistake) in addition to the bias that arises dueto fact that the GDP deflator includes non-traded goods prices as well. Baldwinand Taglioni (2006) argue that including time dummies can correct for mistakendeflation procedure. We try to avoid the bronze and silver medal mistakes (whicharise when using the average of the two-way exports) and estimate the gravity

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

234 B. Dettmer

equation with nominal trade and GDP values as well as uni-directional tradeflows. Country fixed effects will cover what Baldwin and Taglioni (2006) call,‘the gravitational un-constant term’ by removing the cross-section correlationbetween the multilateral resistance term and the independent variables. Althoughtime-invariant pair dummies are superior to nation dummies, estimating the timezone variable in our model does not allow adding pair fixed effects as all time-invariant variables will drop out. As in previous studies on time zone effects, weestimate the model with importer- and exporter country fixed effects and yeardummies.

3.2 Variables and Data Sources

Our main interest is to compare the different time zone effect on business andcommercial services trade. We use bilateral services trade data from the OECDStatistics on International Trade in Services. The statistic covers the period from1999 to 2006 (OECD, 2008). Total services trade data are reported by 27 OECDcountries with their respective 226 partner countries.10 The database allows thedistinction between several service sectors. Regressions for trade in businessservices (which include leasing, legal, accounting, auditing, book-keeping, taxconsulting, business and management consulting, advertising, and research) arecompared with trade in commercial services, which is a broader categorizationof business services (as it additionally considers communication, construction,insurance and financial services, computer and information services, royaltiesand license fees). This is most interesting, as the recent debate on off-shoring ofservice transactions focuses on business and commercial services rather than ontravel, transportation and government services. In sum, these service sectors addup to total service transactions between countries. In addition, we estimate thetime zone effect for merchandise trade based on the OECD International Trade byCommodity Statistics (OECD, 2010) for which we select the same sample period.

Moreover, the services trade data in the OECD statistic are on a balance-of-payments basis and, thus, are an aggregate that covers GATS mode 1 trade(cross-border) and mode 2 transactions (consumption abroad). In contrast, theyinclude only a small part of GATS mode 3 (commercial presence) and mode 4(movement of natural persons) transactions. This classification has been adoptedas a framework for current multilateral negotiations under the GATS. All fourmodes of service transactions considered in the GATS classification reflect tosome extent the suppliers’ choice of services delivery (see, for example, Patterson& Cicic, 1995, for detailed analysis of the entry mode choices of Australian serviceexporters). However, the WTO estimates that up to 50% of all service transactionsis allocated to mode 3 transactions (FDI) whereas mode 1 and mode 2 transac-tions account for 35% and 10 to 15% respectively (see Maurer, Magdeleine, &

10The OECD dataset includes 25 reporter countries for commercial services trade and 23 reportercountries for business services trade for the time period 1999 to 2006 (see Appendix Table A1.1for country and year coverage). Due to missing bilateral data, the number of observations varieswith respect to the service sector: the number of commercial services export observations reducesto 5681; the number of observations for business services exports reduces to 4396 respectively.However, goods trade data are largely reported with 43,680 observations in the sample.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 235

d’Andrea, 2006). While service delivery via FDI is recorded in the foreign affiliate’strade statistics (FATS), the OECD data set that we use here covers mainly cross-border trade where neither the service supplier nor the receiver moves. A highextent of mode 2 service delivery is made up of tourism service, which is notincluded in the service categories considered here. However, the data set we useto evaluate the time zone effect unfortunately limits our analysis since we are notable to split cross-border business services trade according to the degree of buyer–seller interaction (simultaneous or sequential). The analysis allows a conclusionwith respect to transactions that dominate in cross-border business service trade.

For the time zone measure we calculate the shortest time zone differencebetween the countries’ capitals in hours (TIMEcap) based on time zone dataprovided by PTB (2010). The variable varies between 0 and 12 hours when weabstract from daylight saving time. The full sample (henceforth Panel A) alsoincludes (reporter and partner) countries with multiple time zones (e.g. Australiawith UTC + 8 hours in the west and UTC + 10 hours in the east, or the RussianFederation with UTC + 3 hours in the European zone and UTC + 12 hours inVladivostok which is also the largest intra-country time zone distance).11 How-ever, it can be argued that in some countries the capital city is the center ofeconomic activity (e.g. Australia, Indonesia, and the Russian Federation), wheremost of the (internationalized) business service firms locate. In this respect, usingTIMEcap as (our main) measure to capture the time zone effect seems to bevaluable. Nevertheless, in some countries the (internationalized) business servicesector is distributed within the country, with service firms located in regionalmetropolises (e.g. in Canada and at the US East and West coasts). Hence, timezone differences based on the countries’ geographical center (here calculated asTIMEmean) or based on the minimum geographical distance between the countryand the respective trading partner (TIMEmin) is a better approximation to coversuch examples. Thus, for a robustness check we report results on TIMEmean andTIMEmin as well. Moreover, we calculate the number of overlapping office hoursbetween the trading countries’ capitals based on either a normal 8 hour workingday (from 9 am to 5 pm) or an intensive 10 hour working day (from 9 am to 7 pm)which is mostly common in business consulting firms and R&D departments.The variables (Office8, Office10) vary between zero (with no hours overlap: theseinclude trading pairs with a time zone difference from 9 (11, respectively) to 12hours) and 8 hours overlap (10, respectively) when countries are located in thesame time zone.12

In addition, we test the time zone effect on business and commercial serviceswith respect to the precondition that countries need to be connected to ICT to

11See Appendix Table A1.1 for reporter countries included in Panel A and Table A1.2 for countrieswith multiple time zones excluded from the analysis (Panel B). Appendix Table A1.3 contains a listof Non-OECD trading partners.12In addition to variation in the measurement of time zones, we estimate the time zone effectin a restricted sample (henceforth Panel B) which excludes (reporter and partner) countries withmultiple time zones. In this sample, the US, Canada, and Australia which are known as majorservices exporters with respect to the UK, are excluded. The results present further robustnesschecks for the validity of the time zone variables.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

236 B. Dettmer

transfer business services. The time zone variable interacts with the ICT vari-able in the country with the poorer ICT infrastructure to estimate the marginalimpact of minimum ICT and time zone differences on business services trade.The World Bank (2010) reports ICT variables (mobile phones, telephones, per-sonal computers, and internet access), which are expressed in relative terms to thecountry’s population (except for the registered air transport carriers). These vari-ables reflect the distribution of communication infrastructure within the country.Another way to deliver business services requires traveling. Thus, the country’sdomestic takeoffs and takeoffs abroad of air carriers registered in the country isused to proxy for the ability to receive good flight connections. Owing to limitedavailability of data on the infrastructure at the country-level, the minimum ICTvariable mostly restricts our sample to a lower number of observations in theregressions.

The empirical analysis includes control variables which are shown to be relevantfor trade in services: Although previous studies disagree about the impact of geo-graphical distance in certain services, distance needs to be included in the analysisto figure out the additional effect of time zone differences.13 Data on standardgravity variables (distance and colonial ties) are taken from the CEPII-Database(CEPII, 2010). We use distw as the geographical distance between countries,which is calculated following the great circle formula using latitudes and longi-tudes of the most important cities, which are weighted by their population size.We account for language barriers that are relevant in trading business services.Officially spoken languages may not capture the effective language barriers.14

Thus, we construct a bilateral dummy that is equal to one when English is awidely spoken language in both countries. Here, we use data from the CIA WorldFactbook (CIA, 2010). In addition to the countries in which English is an officiallanguage, we treat countries as ‘widely spoken English countries’ in which morethan 50% of the population speaks English (see Appendix Table A2 for a list ofcountries). Since most business services belong to a value added process where thequality cannot be evaluated in advance, the presence of asymmetric informationbetween business partners leads to the fact that trust is a necessary condition tobuild up relationships across borders. Culture and religion are often associatedwith trust and play a prominent role for the ability to build networks.15 Similarto Vietze (2011), we use the CIA World Factbook (CIA, 2010) data to build dum-mies that are equal to one if more than 60% of a country’s population belong toa certain religious denomination (Catholic, Protestant, Orthodox, Muslim, and

13Although both distance and time zone differences reflect geographic division and are thus highlycorrelated (around 0.75), the distance needs to be included in the regression. However, Head et al.(2009) and Stein and Daude (2007) report similar correlations.14Tharakan and van Beveren (2003) and Tharakan et al. (2005) find that India’s endowment with alarge stock of IT professionals proficient in English is an important determinant of India’s successfulsoftware export.15See Glaeser, Laibson, and Sacerdote (2002) on determinants of social capital and Knack andKeefer (1997) on its effects on economic outcomes. Guiso, Sapienza, and Zingales (2009) show thatreligious similarity has a positive impact on trust. Tharakan and van Beveren (2003) stress the roleof co-ethnic networks in bilateral services trade (see also Gould, 1994; Rauch & Casella, 2003;Rauch & Trindade, 2002).

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 237

Others, respectively). Moreover, to control for cultural similarity between tradingcountries, we construct a set of bilateral religion dummies (one dummy for eachreligion), which are equal to one if more than 60% of both the reporter- andpartner-country’s population belong to the same religion, and zero otherwise.We include a dummy for regional trade agreements between countries to accountfor trade barriers when agreements are not signed for GATS Article V (for theservices trade regression), and GATT Article XXIV (for the goods regression)respectively (WTO, 2010).16

4. Results

4.1 Time as a Trade Barrier

Estimations for the time zone effect are presented in Table 1 for trade of special-ized business services and Table 2 for the broader categorization of commercialservices. In Panel A we use all bilateral services trade flows to estimate variousmeasures of the time zone effect. All estimates on control variables are statisticallysignificant and show the expected sign.17

We find a negative distance effect for business services and all sub-sectorsof commercial services trade. In contrast, the time zone effect is significantlypositive for trade in specialized business services. This indicates that the continuityeffect offsets the synchronization effect and, hence, service firms profit from timezone differences by being able to operate over a 24 hour business day. Thus,service providers trade more with business partners located in a distant timezone. The significance of the continuity effect holds for time zones measured inhours between the countries’ capitals (TIMEcap) and for alternative measures oftime zone distance based on countries’ geographical centers (TIMEmean) andthe minimum time distance (TIMEmin). Further robustness checks on time zonemeasures show that business service transactions decrease with the number ofoverlapping office hours in both business partners’ countries.18 However, timezones are not a driving force for trade of commercial services, as shown in the firstcolumn of Table 2. The time zone effect remains insignificant for all alternative

16See descriptive statistics in Appendix Table A3. According to the correlation matrix in Table A4the correlation between home country GDP and the minimum ICT variables are relatively low.Since the dataset includes bilateral trade between OECD and Non-OECD countries, the minimumICT network variables correlate to a somewhat higher extent with the host country GDP, around0.4. Nevertheless, home and host country GDP need to be included otherwise we run into seriousproblems with respect to omitted variable bias.17Colonial ties and language barriers are important for the export of business services. Moreover,we find strong explanatory power of both countries’ lagged GDP, which indicates that firms exportbusiness services to countries with high demand for heterogeneous products. The coefficients forcultural similarity are positive and significant for Orthodox religion and negative and significant forother religion (which includes Hindu, Buddhist and other religions). The impact of Protestant andCatholic religions is insignificant in all cases. As we do not have a reporter country with more than60% Muslim religion, the bilateral dummy for Muslim drops from the estimation. For commercialservices, the dummy for the GATS agreement is significantly positive while EU27 membership isinsignificant. For business services, both the GATS dummy and EU27 membership are insignificant.18The continuity effect remains significant in specialized business services trade when all multipletime zone countries are excluded in sample B (not shown for brevity).

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

238 B. Dettmer

Table 1. Time zones and business services exports

Business services exports

Tobit (A1) (A2) (A3) (A4) (A5) (A6)

Ln Dist −1.813∗∗∗ −1.969∗∗∗ −2.020∗∗∗ −2.001∗∗∗ −1.979∗∗∗ −1.981∗∗∗(0.106) (0.147) (0.139) (0.144) (0.147) (0.147)

TIMEcap 0.073∗∗(0.0277)

TIMEmean 0.107∗∗∗(0.0266)

TIMEmin 0.095∗∗∗(0.0288)

Office 8 −0.088∗∗(0.0324)

Office 10 −0.081∗∗(0.0286)

Ln GDPit−1 2.125∗∗∗ 2.108∗∗∗ 2.100∗∗∗ 2.103∗∗∗ 2.108∗∗∗ 2.108∗∗∗(0.447) (0.446) (0.444) (0.445) (0.446) (0.446)

Ln GDPjt−1 1.240∗∗∗ 1.237∗∗∗ 1.228∗∗∗ 1.233∗∗∗ 1.241∗∗∗ 1.238∗∗∗(0.272) (0.271) (0.271) (0.271) (0.271) (0.271)

English 0.534∗∗∗ 0.509∗∗∗ 0.494∗∗∗ 0.502∗∗∗ 0.506∗∗∗ 0.508∗∗∗(0.146) (0.147) (0.147) (0.147) (0.147) (0.147)

Colony 1.093∗∗∗ 1.056∗∗∗ 1.042∗∗∗ 1.046∗∗∗ 1.064∗∗∗ 1.057∗∗∗(0.176) (0.179) (0.178) (0.179) (0.178) (0.179)

Orthodox 0.966∗∗∗ 0.991∗∗∗ 0.983∗∗∗ 0.993∗∗∗ 1.013∗∗∗ 0.999∗∗∗(0.254) (0.248) (0.249) (0.248) (0.247) (0.248)

Catholic −0.192 −0.183 −0.169 −0.175 −0.186 −0.185(0.153) (0.153) (0.153) (0.153) (0.153) (0.153)

Protestant 0.364 0.306 0.274 0.280 0.294 0.302(0.320) (0.314) (0.314) (0.314) (0.314) (0.314)

Other Religion −1.137∗∗∗ −0.932∗∗∗ −0.984∗∗∗ −0.967∗∗∗ −0.874∗∗ −0.901∗∗(0.286) (0.276) (0.276) (0.276) (0.278) (0.276)

EU27 −0.193 −0.0517 0.181 0.0984 0.0238 −0.0254(0.342) (0.332) (0.331) (0.331) (0.336) (0.333)

FTA GATS V 0.267 0.389 0.427∗ 0.408∗ 0.363 0.392(0.246) (0.248) (0.248) (0.248) (0.247) (0.248)

Constant −56.73∗∗∗ −55.16∗∗∗ −54.33∗∗∗ −54.42∗∗∗ −54.50∗∗∗ −54.31∗∗∗(14.29) (14.32) (14.28) (14.31) (14.33) (14.34)

Obs. 4757 4756 4756 4756 4756 4756Cens.Obs. 550 550 550 550 550 550Log pseudo-

likelihood−9869.69 −9865.31 −9862.23 −9863.89 −9865.00 −9864.86

Pseudo R2 0.2405 0.2407 0.2410 0.2408 0.2407 0.2408

Seemingly unrelated estimationOECD (mean)

TIME−∗ – −0.0357 −0.0173 −0.0181 0.0322 0.0265(0.0345) (0.0345) (0.0370) (0.0409) (0.0345)

Non–OECD (mean)TIME−∗ – 0.180∗∗∗ 0.150∗∗∗ 0.171∗∗∗ −0.181∗∗∗ −0.178∗∗∗

(0.0442) (0.0381) (0.0430) (0.0483) (0.0462)

Chow test: H0: TIME[Non − OECD] − TIME[OECD] = 0Chi(2) – 14.72 10.63 11.17 11.32 12.52p-value – 0.0001 0.0011 0.0008 0.0008 0.0004

Note: See Appendix Table A2 for business services included. All regressions are Tobit estimations and includeexporter and importer fixed effects and year dummies. Results of the seemingly unrelated regressions arebased on Tobit estimations. Standard errors robust to heteroskedasticity reported in parentheses. ∗∗∗denotesignificance at 1% level, ∗∗5% level,∗10% level respectively.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 239

Table 2. Time zones and commercial services trade

Commercial services exports

(A1) (A2) (A3) (A4) (A5) (A6)Communicat. Construct. Computer Financial Insurance

Tobit All services services services services services

Ln Dist −1.465∗∗∗ −2.927∗∗∗ −4.586∗∗∗ −2.463∗∗∗ −3.456∗∗∗ −4.916∗∗∗(0.151) (0.218) (0.304) (0.0139) (0.226) (0.313)

TIMEcap −0.0376 −0.0007 0.364∗∗∗ 0.482∗∗∗ 0.247∗∗∗ 0.283∗∗∗(0.0279) (0.0547) (0.0779) (0.0191) (0.0627) (0.0615)

Ln GDPit−1 1.456∗∗∗ 7.027∗∗∗ 1.372 6.390∗∗∗ 2.458∗∗ −0.0107(0.421) (0.769) (1.048) (0.004) (0.796) (1.050)

Ln GDPjt−1 0.522∗ 0.808 2.734∗∗∗ 1.963∗∗∗ 1.152∗∗ 0.666(0.277) (0.491) (0.665) (0.004) (0.476) (0.667)

English 0.717∗∗∗ 1.579∗∗∗ 2.132∗∗ 3.990∗∗∗ 0.657∗∗ 3.123∗∗∗(0.154) (0.411) (0.859) (0.247) (0.314) (0.382)

Colony 0.965∗∗∗ 1.307∗∗∗ 1.307∗∗ 3.110∗∗∗ 0.927∗∗ 0.493(0.163) (0.339) (0.440) (0.146) (0.306) (0.373)

Orthodox 1.222∗∗∗ 1.037∗ 3.226∗∗∗ – 0.846 2.972∗∗∗(0.305) (0.605) (0.734) (0.539) (0.800)

Catholic −0.242 0.299 −0.413 2.681∗∗∗ 0.0991 0.990∗∗(0.167) (0.280) (0.358) (0.116) (0.283) (0.407)

Protestant 0.465 0.372 0.668 −6.845∗∗∗ 0.843∗ −2.904∗∗∗(0.340) (0.471) (0.636) (0.166) (0.477) (0.585)

Other Religion −1.235∗∗∗ 0.101 −0.773 – 0.632 0.449(0.265) (0.463) (0.758) (0.590) (0.656)

EU27 −0.578 1.193 1.877∗ 9.311∗∗∗ −0.622 1.511∗∗(0.359) (0.757) (1.090) (0.126) (0.810) (0.750)

FTA GATS V 1.022∗∗∗ −1.592∗∗ −0.362 −2.637∗∗∗ −1.429∗∗ −3.039∗∗∗(0.278) (0.523) (0.949) (0.135) (0.515) (0.602)

Constant −23.24∗ −165.2∗∗∗ −56.42∗ −241.1∗∗∗ −50.97∗∗ 41.06(14.04) (24.63) (33.87) (0.114) (24.88) (34.27)

Obs. 5879 4085 3941 25725 4246 4394Cens. Obs. 417 1506 1920 25070 1524 1909Log pseudo-

likelihood−13330.75 −7684.87 −6663.89 −2137.62 −8154.50 −8541.41

Pseudo R2 0.2151 0.2114 0.2118 0.5944 0.2114 0.2120

Seemingly unrelated estimation

OECD (mean)TIMEcap −0.154∗∗∗ −0.162∗∗ 0.0645 −0.060∗∗ 0.168∗∗ 0.0674

(0.0415) (0.0514) (0.0939) (0.031) (0.0661) (0.0788)

Non-OECD (mean)TIMEcap 0.141∗∗∗ 0.135 0.974∗∗∗ 0.020∗∗ 0.539∗∗∗ 0.604∗∗∗

(0.0348) (0.109) (0.131) (0.006) (0.118) (0.115)

Chow test: H0: TIME[Non − OECD] − TIME[OECD] = 0Chi(2) 29.6 6.11 31.89 6.61 7.56 14.80p-value 0.000 0.0134 0.000 0.0101 0.006 0.0001

Note: See Appendix Table A2 for commercial services included. All regressions are Tobit estimations and includeexporter and importer fixed effects and year dummies. Results of the seemingly unrelated regressions are basedon Tobit estimations. Standard errors robust to heteroskedasticity reported in parentheses. ∗∗∗denote significanceat 1% level, ∗∗5% level,∗10% level respectively.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

240 B. Dettmer

measures of time zone differences and does not change when multiple time zonecountries are excluded. Two reasons may explain why time zones do not matterfor commercial services.

First, the heterogeneity of services within the broad category of commercial ser-vices avoids justifying a significant positive or negative time zone effect. Accordingto Table 2, the continuity effect is significant for trade of construction services,computer and information services, financial and insurance services. The coeffi-cients remain robust for the respective services sectors when alternative measuresof time zone differences are included instead. In contrast, for communicationservices, the time zone effect is insignificant and remains insignificant for all alter-native measures of time zone differences.19 The heterogeneity of single servicetransactions suggests the neutralization of the time zone effect in commercial ser-vices with neither the continuity effect nor the synchronization effect dominating.

Second, the insignificance of the continuity effect in commercial service trademay depend on the countries considered in the sample. The time zone effect maymatter more in OECD trade with Non-OECD countries, since support servicesmay be outsourced to lower-income countries when ICT infrastructure allows this.We perform a seemingly unrelated estimation, which includes both intra-OECDtrade and OECD–Non-OECD trade in one regression to test the significance ofthe difference in the time zone coefficients. The results are reported in the lowerpart of Tables 1 and 2. The time zone effect is positive and strongly significant inbusiness services trade with low-income Non-OECD countries. For intra-OECDbusiness service trade we find an insignificant time zone effect. According to theChow test, we can strongly reject the null hypothesis of equal coefficients forintra-OECD business service trade and OECD trade with Non-OECD countries.Excluding multiple time zone countries in panel B does not change the results;the Chow test is strongly rejected. Time zones matter more in business servicetrade with Non-OECD countries than in intra-OECD trade.20

The continuity effect present in trading business services is not valid formerchandise trade.21 Rather, we find significantly negative time zone effects.Estimates on alternative definitions of time zones strengthen the results:merchandise trade significantly increases with the number of overlapping officehours between trading partners. Excluding reporter and partner countries withmultiple time zones in sample B confirms the robustness of the negative time zoneeffect. Thus, time zones seem to be associated with additional transaction costs

19According to the IMF’s balance of payments manual, communication services cover two pri-mary categories of international transactions: telecommunication services (which include businessnetwork services, teleconferencing, and support services) and postal and courier services (whichinclude transport and delivery of print media). The insignificant effect of time zones in commu-nication services tends to be reasonable since time zones are associated with additional costs forpostal and courier services.20Moreover, the continuity effect is strong in OECD commercial service trade with Non-OECDcountries as well. However, the time zone effect is significantly negative in intra-OECD commercialservices trade. A breakdown by single services sectors shows that the result stems from commu-nication and computer and information services while the time zone effect remains positive andsignificant in financial service.21Results are not shown for brevity but are available from the author upon request.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 241

in the delivery of goods that are not captured by the distance variable at all. How-ever, our estimates show that the distance effect drops only slightly when addingthe time zone variable to the model and it remains at the level when turning tothe restricted goods sample in which we account for observations on merchan-dise trade when business services export data are non-missing. The time zoneeffect in merchandise trade reverses and becomes significantly positive. In thisrespect, trade in goods follows the same pattern as cross-border business servicetrade, which is perhaps an indication for the fact that business services accompanymanufacturing products. This is especially relevant for high-tech products such asmachinery or ICT equipment, which are bundled with (locally produced) services,i.e. providing financing and installation services as well as customer support.

A large part of OECD merchandise trade takes place in higher-valued productcategories. The seemingly unrelated estimation provides evidence that time zonesrepresent a significant barrier for intra-OECD trade, which is perhaps an indi-cation for additional transaction costs in the form of frequent communicationbetween firms (and the consumer) when in-house customer support is provided.In contrast, this is not the case for merchandise trade with Non-OECD coun-tries: the continuity effect still holds. According to the Chow test, we can stronglyreject the null hypothesis of equal time zone effects in both subgroups.22 In thisrespect, Non-OECD countries play an important part in firms’ fragmentation ofthe production chain and the delivery of intermediates. Thus, transaction coststo connect production steps in distant locations seem to differ from transactioncosts associated with horizontal intra-industry trade in higher valued products.

4.2 Interconnectivity and Time Zones

Note that (accompanying) cross-border business and commercial services trans-actions without movement of service supplier and consumer (as in the case ofcustomer support services) require a good communications infrastructure toenable the interaction between business partners. Therefore, the second modelevaluates the hypothesis that the ability for service providers to make use of thecontinuity effect of time zones depends on the availability of a relevant ICT infras-tructure network in the country with the poorer access rate to ICT. Results arereported in Table 3. The heading of each column shows which ICT variable isincluded in the respective model.

The interaction term on time zone differences and the minimum ICT network ispositive and highly significant for all ICT variables. Thus, the effect of time zoneson business services trade depends on the access rates to an ICT network in thecountry with the poorer infrastructure. And vice versa, the effect of an increase ofminimum access rates to ICT on business services trade depends on the time zonedifference between business partners. The coefficient on the joint impact of timezones and minimum ICT is also highly significant and positive for commercial

22The time zone effect remains strong in OECD–Non-OECD merchandise trade when excludingmultiple time zone countries from the sample, but time zone differences do not affect intra-OECDmerchandise trade. Nevertheless, according to the Chow test, the null hypothesis of equal time zonecoefficients in both subsamples is rejected at the 2.6% level.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

242B.D

ettmer

Table 3. Time zones and ICT interconnectivity in commercial and business services trade

Tobit Business services exports Commercial services exportsMinICTincluded: MOBILE PHONE PC NET AIR MOBILE PHONE PC NET AIR

Ln Dist −1.974∗∗∗ −1.999∗∗∗ −2.082∗∗∗ −1.995∗∗∗ −1.937∗∗∗ −1.487∗∗∗ −1.532∗∗∗ −1.631∗∗∗ −1.516∗∗∗ −1.443∗∗∗(0.147) (0.147) (0.146) (0.146) (0.145) (0.152) (0.152) (0.151) (0.151) (0.149)

TIMEcap −0.0876 −0.328∗∗∗ −0.167∗∗∗ −0.0481 −0.763∗∗∗ −0.123∗∗ −0.272∗∗∗ −0.209∗∗∗ −0.120∗∗∗ −0.923∗∗∗(0.0556) (0.0873) (0.0456) (0.0405) (0.125) (0.0403) (0.0559) (0.0364) (0.0341) (0.115)

minICT −0.0966 −0.581∗∗ −1.253∗∗∗ −0.302∗∗ −0.439∗∗∗ −0.0766 −0.560∗∗ −1.144∗∗∗ −0.242∗∗ −0.550∗∗∗(0.0943) (0.193) (0.119) (0.0956) (0.0896) (0.101) (0.204) (0.119) (0.107) (0.0779)

TIMEcap∗ minICT 0.047∗∗∗ 0.097∗∗∗ 0.094∗∗∗ 0.046∗∗∗ 0.070∗∗∗ 0.0293∗∗ 0.0635∗∗∗ 0.0838∗∗∗ 0.0399∗∗∗ 0.0749∗∗∗(0.013) (0.019) (0.012) (0.010) (0.010) (0.00982) (0.0132) (0.0103) (0.00877) (0.00902)

Const −54.05∗∗∗ −59.49∗∗∗ −62.55∗∗∗ −60.62∗∗∗ −61.91∗∗∗ −22.80 −27.16∗ −27.94∗∗ −24.98∗ −29.09∗∗(14.42) (14.35) (14.39) (14.37) (14.35) (13.94) (13.97) (13.97) (14.03) (14.02)

Obs. 4755 4756 4751 4756 4756 5875 5879 5873 5879 5871Cens. Obs. 550 550 550 550 550 417 417 417 417 417Log pseudo-

likelihood−9857.46 −9853.79 −9804.02 −9855.22 −9843.50 −13318.86 −13320.22 −13268.66 −13320.28 −13278.38

Pseudo R2 0.2412 0.2416 0.2446 0.2415 0.2424 0.2153 0.2157 0.2179 0.2157 0.2168

Seemingly unrelated estimation

Non-OECD (mean)TIMEcap∗ minICT 0.00298 0.00789 0.00645 0.00463 0.0775∗∗ −0.0247∗∗ −0.0379∗∗ −0.395∗∗ −0.0269∗∗ 0.0140

(0.0180) (0.0294) (0.0225) (0.0173) (0.0254) (0.0118) (0.0169) (0.180) (0.0120) (0.0245)

OECD (mean)TIMEcap∗ minICT 0.0257 0.081∗∗ 0.099∗∗∗ 0.028∗∗ 0.061∗∗∗ 0.0547∗∗ 0.206∗∗∗ 0.149∗∗∗ 0.0753∗∗∗ 0.0858∗∗∗

(0.0218) (0.0399) (0.0188) (0.0141) (0.0102) (0.0261) (0.0455) (0.0180) (0.0147) (0.0102)

Chow test: H0: TIMEcap∗minICT[Non-OECD] − TIMEcap∗minICT[OECD] = 0Chi(2) 0.64 2.18 9.92∗∗ 1.11 0.37 7.71 25.26 55.02 28.93 7.32p-value 0.4225 0.1394 0.0016 0.2924 0.5451 0.0055 0.000 0.000 0.000 0.007

Note: All regressions are Tobit estimations and include the log of importer and exporter GDP, dummies for common language, common religion, colonial ties,dummy for service trade agreements and EU27 membership. Importer and exporter fixed effects as well as time dummies are included. Standard errors robustto heteroskedasticity are reported in parentheses. ∗∗∗denote significance at 1% level, ∗∗5% level,∗10% level respectively.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 243

services trade, but we find mixed results when considering different types ofcommercial services: the interaction term is significantly positive in the trade offinancial services, insurance services, and computer and information services,but less robust in construction services and not significant in communicationservices trade.

However, in business service trade we obtain, for example, the lowest coeffi-cient on the interaction term of time zones and access to the internet, which is0.046. The magnitude of the interaction effect is, on average, 0.4 log points(=0.046∗3.959∗2.213 = coeff∗mean(TIMEcap)∗mean(log[minNET])) and doubleswhen considering a maximum internet access rate in the country group with thepoorer network (0.8 log points = 0.046∗3.959∗4.472). However, the magnitude ofthe interaction term triples when the maximum time zone distance of 12 hoursis accounted for (1.2 log points = 0.046∗12∗2.213) and reaches 2.5 log points forboth the maximum time zone difference and maximum internet access rate.23

Since the access rates to ICT infrastructure are on average less developed inNon-OECD countries, the magnitude of the interaction would presume that thosecountries can increase business trade with partner countries in distant time zonesby increasing the access rates to electronic communications equipment. Althoughit seems reasonable that the joint impact of time zones and ICT on business servicetrade differs for both groups, the lower part of the table provides the results of theChow test. We fail to reject the null hypothesis: The magnitude of the interactionterm is the same for both groups of trading partners.24

Moreover, by including the interaction term in the model, the coefficient onthe time zone variable turns negative and is interpreted as a single effect whenthe access rate to ICT infrastructure in the country with the poorer network iszero. The same holds for the ICT variables for which now negative coefficientsare observable: the interpretation can only be made if no time zone differencesexist. In this respect, we estimate the average marginal effect of an increase in timezone differences on business and commercial services trade while holding the levelof minimum ICT in the country with the poorer ICT infrastructure constant atvarious percentiles of the cumulative distribution. The results are reported inTable 4. In general, there is a positive average effect of time zones on businessservices trade in countries that are at the mean of the minimum ICT distribution.The time zone effect is significant at the 10% level. In the group of countriesthat are in the bottom 50th percentile of the distribution (median) the averageeffect of time zones on business services trade is positive as well. The coefficient is

23According to Appendix Table A3, the internet infrastructure variable has a mean of 2.213 anda maximum of 4.408, which indicates that at worst 9% of the population in the country withthe poorer infrastructure has access to the internet, while at best 82% of the population uses theinternet.24For commercial services, however, the Chow test validates significant differences in the coefficientfor intra-OECD trade and OECD trade with Non-OECD partner countries. The strong resultis misleading since commercial services are very heterogeneous: according to the estimates onsingle services sectors, we cannot reject the hypothesis of the equality of coefficients across countrygroups for all types of commercial services except for computer and information services. The sameargument applies to merchandise trade: the hypothesis of equality in the joint impact of time zonesand minimum ICT is rejected.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

244 B. Dettmer

quantitatively higher and statistically significant at the 5% level. Moving to the top75th percentile, the effect of time zones on business services trade is statisticallysignificant at the 1% level and quantitatively about 1.5 to 2 times the size of theestimated average marginal effect of the bottom 50th percentile. In contrast, theaverage time zone effect on business services trade with partner countries that arein the bottom 25th percentile of the ICT distribution is insignificant. The resultimplies that ICT infrastructure tends to be a bottleneck to trade business servicesat increasing time zone distance. A minimum level of ICT needs to be establishedin the country with the poorer ICT infrastructure.

The empirical results for commercial services are more mixed, mainly due tothe varying impact of time zones and minimum ICT by type of services sectorincluded. Although the coefficient on the interaction term is highly significant andpositive, time zones are not a driving force of trade in the broad categorizationof commercial services. Table 4 shows that the effect of time zone distance oncommercial services trade is insignificant at nearly all percentiles of the cumulativedistribution of minimum ICT.25 The significant interaction term tends to be drivenby the effect of minimum ICT on trade.

The lower part of Table 4 contains the marginal effects of an increase in min-imum ICT on business and commercial services trade at selected time zonedifferences between countries. In fact, we find that the marginal effect of anincrease of the minimum ICT (i.e. in the country with the poorer network) byone unit from the mean on business services trade depends on the time zonedistance between countries. Business services trade at longer time zone distancesincreases with rising access to minimum ICT, while at shorter time zone distancesthe marginal effect of minimum ICT on trade is insignificant or even negative.26

The same holds for commercial services trade: an increase of the minimum accessrates to ICT by one unit from the mean affects commercial service trade positivelyat time zone distances of more than 9 hours. At short time zone distance of lessthan 9 hours the effect of minimum ICT is insignificant or even negative.27

At longer time zone distances, country pairs increase trade of business and com-mercial services with a rising access to ICT infrastructure network. In this respect,service firms export business services to countries in significant distant time zonesin order to save time when the established ICT infrastructure network allows this.

25A detailed analysis shows that time zones do not affect trade of communication services irrespec-tive of the ICT variable included. But this is not the case for construction services, computer andinformation services, financial and insurance services. The marginal effect of an increase in timezone distance on trade is positive and strongly significant at all percentiles of the minimum ICTdistribution.26The negative marginal effect of access to personal computers on business and commercial servicestrade might be explained by the circumstance that personal computers do not connect countriesto an international network per se, like the worldwide web, mobile phones, and international airtransport availability.27The access to air transport becomes important at longer time zone distances for trade of alltypes of commercial services. With respect to electronic communications equipment we find mixedresults: ICT used for simultaneous interaction (mobile phone and fixed-line telephony) is importantfor communication services trade at all time zone distances, but it does not matter for constructionand financial services trade. A minimum access to the internet is only important for computerservices at long time zone distances.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

InternationalServiceT

ransactions245

Table 4. Marginal effects of time zones and minimum ICT on commercial-and business service trade

Business services exports Commercial services exports

MOBILE PHONE PC NET AIR MOBILE PHONE PC NET AIR

Marginal effects: (dydx) TIMEcapAt minICT (P25) 0.0258 0.002 −0.025 0.014 −0.047 −0.053∗ −0.056∗ −0.078∗∗ −0.066∗∗ −0.145∗∗∗

(0.032) (0.032) (0.033) (0.032) (0.034) (0.028) (0.028) (0.029) (0.029) (0.032)(mean) 0.053∗ 0.050∗ 0.042 0.049∗ 0.003 −0.037 −0.026 −0.020 −0.036 −0.085∗∗

(0.029) (0.028) (0.029) (0.029) (0.030) (0.028) (0.028) (0.028) (0.028) (0.029)(P50) 0.069∗∗ 0.063∗∗ 0.042 0.057∗∗ 0.028 −0.025 −0.010 −0.016 −0.026 −0.062∗∗

(0.028) (0.028) (0.029) (0.028) (0.029) (0.028) (0.029) (0.028) (0.028) (0.028)(P75) 0.101∗∗∗ 0.123∗∗∗ 0.135∗∗∗ 0.097∗∗∗ 0.061∗∗ −0.007 0.024 0.069∗∗ 0.009 −0.015

(0.028) (0.028) (0.028) (0.027) (0.028) (0.030) (0.031) (0.030) (0.029) (0.027)(P99) 0.126∗∗∗ 0.155∗∗∗ 0.219∗∗∗ 0.142∗∗∗ 0.180∗∗∗ 0.009 0.044 0.142∗∗∗ 0.046 0.111∗∗∗

(0.030) (0.031) (0.031) (0.030) (0.030) (0.032) (0.033) (0.035) (0.033) (0.030)Marginal effects: (dydx) minICT

At TIMEcap = 0 −0.097 −0.581∗∗ −1.25∗∗∗ −0.302∗∗ −0.439∗∗ −0.077 −0.560** −1.144∗∗∗ −0.242∗∗ −0.550∗∗∗(0.094) (0.193) (0.119) (0.096) (0.090) (0.101) (0.204) (0.119) (0.107) (0.078)

mean (3.9) 0.074 −0.229 −0.913∗∗∗ −0.137 −0.186∗∗ 0.039 −0.309 −0.813∗∗∗ −0.085 −0.255∗∗∗(0.095) (0.183) (0.103) (0.087) (0.073) (0.096) (0.189) (0.104) (0.093) (0.063)

6 0.185∗ 0.001 −0.692∗∗∗ −0.029 −0.021 0.099 −0.179 −0.641∗∗∗ −0.003 −0.101(0.109) (0.191) (0.102) (0.089) (0.070) (0.099) (0.187) (0.102) (0.090) (0.062)

9 0.326∗∗ 0.291 −0.411∗∗∗ 0.108 0.188∗∗ 0.187∗ 0.012 −0.390∗∗∗ 0.117 0.124∗(0.134) (0.216) (0.112) (0.099) (0.077) (0.110) (0.190) (0.107) (0.093) (0.070)

10 0.373∗∗ 0.388∗ −0.318∗∗ 0.154 0.258∗∗ 0.217∗ 0.076 −0.306∗∗ 0.157 0.199∗∗(0.144) (0.226) (0.117) (0.105) (0.082) (0.115) (0.193) (0.110) (0.095) (0.075)

11 0.420∗∗ 0.485∗∗ −0.224∗∗ 0.199∗ 0.328∗∗∗ 0.246∗∗ 0.139 −0.222∗ 0.197∗∗ 0.274∗∗∗(0.155) (0.238) (0.124) (0.111) (0.088) (0.121) (0.197) (0.115) (0.099) (0.081)

12 0.467∗∗ 0.582∗∗ −0.131 0.245∗∗ 0.397∗∗∗ 0.275∗∗ 0.203 −0.138 0.236∗∗ 0.349∗∗∗(0.166) (0.251) (0.132) (0.118) (0.094) (0.127) (0.202) (0.120) (0.103) (0.087)

Note: Marginal effects based on Tobit estimations which include the log of importer and exporter GDP, dummies for common language, common religion, colonialties, dummy for service trade agreements and EU27 membership. Importer and exporter fixed effects as well as time dummies are included. Standard errors robust toheteroskedasticity are reported in parentheses. ∗∗∗denote significance at 1% level, ∗∗5% level, ∗10% level respectively.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

246 B. Dettmer

Thus, an increase of ICT access rates in countries with a poorer ICT infrastructurenetwork would favor business trade with time distant countries. The continuityeffect is relevant when providing 24-hour customer support. The positive aver-age effect of an increase in ICT infrastructure used for simultaneous interactions(mobile and fixed-line telephony) might indicate that this is in fact the case.

Trade of specialized business services with time-distant countries is supported,especially by ICT infrastructure. In this respect, the continuity effect of the timezone distance in business service trade depends on the ICT infrastructure network.At long time-zone distances, the ICT infrastructure is more important to connectcountries for services delivery. When business services are transferable via ICTnetworks, the digitalization becomes a serious task for countries to profit fromthe outsourcing decisions of international services firms in general.

5. Conclusion

Due to innovations in information and telecommunication technology (ICT)the proximity requirement for face-to-face interaction between business part-ners seems to have become less important. While geographical distance in gravitymodels on merchandise trade is found to overestimate the true cost of transporta-tion, in the firms’ fragmentation of the value-added chain time has become animportant determinant in the location of intermediate production. In this paper,we analyze whether time is a trade barrier for business and commercial serviceswhen countries become connected to ICT networks. We estimate gravity modelsthat include various measures of time zones and ICT infrastructure variables.According to the theory provided by Marjit (2007) and Kikuchi and Iwasa (2010),we argue that a proper division of labor in the sense of working during regularoffice hours is a necessary condition to make use of time zone differences. We findevidence for a positive time zone effect indicating that the continuity effect dom-inates the synchronization effect at a given geographical distance. Thus, servicessuppliers rather profit from time zone differences. When time zones allow time tobe saved by operating over a 24-hour business day, time seems to be relevant forcross-border business services trade, as previously found for time-sensitive inter-mediates. In this regard, time represents a barrier in trading with business partnerslocated in the same time zone as they can continue to work half a day later when weabstract from shift work. Has the firms’ ability to profit from time zones emergeddue to improved tradability of business services? In line with previous theoreti-cal contributions, we empirically validate that the access to an ICT network is aprecondition for the continuity effect in business services trade. In addition, wepresume that the ICT network is often determined by the country with the pooreraccess rates to ICT infrastructure. By including an interaction term of time zonedistance and minimum ICT access into the model, we find, indeed, that time costsfor cross-border business services trade are significantly dependent on the ICTnetwork. The marginal effect of an increase in time zone distance on business ser-vices trade is insignificant when the minimum ICT is below a certain threshold.This applies for countries that are in the bottom 25th percentile of the cumulativedistribution of minimum ICT. An increase of ICT access rates in countries witha poorer ICT network will significantly favor business and commercial services

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 247

trade at time zone distances of 9 hours and more while the effect is insignificantfor business services trade at shorter time zone distances. Thus, in order to savetime, service firms export business services to countries in significantly distanttime zones when ICT infrastructure networks allow this. When timely deliveryplays a role in business services, time zones may affect the market entry strat-egy of service providers. For transferable business services via ICT, the trade-offbetween time zone costs and degree of buyer–seller interaction may become morerelevant – especially when there is a high degree of uncertainty to go abroad. Thisis an important question for further research because countries in distant timezones become relevant for business services outsourcing. To benefit from businessservices outsourcing, further digitalization is a serious task.

Acknowledgements

The author would like to thank two anonymous referees and the associate editor of the journal for veryuseful comments. The article has benefited from further suggestions by Andreas Freytag, Matthias Geissler,Viktor Slavtchev, and Christoph Vietze. I also thank Sarah Al Doyaili and Nils Laub for their helpful researchassistance. All remaining errors are my own. An earlier version of this paper was presented at the thirteenthAnnual Conference of the European Trade Study Group at Copenhagen Business School and University ofCopenhagen in September 2011.

References

Acemoglu, D., Johnson, S., & Robinson, J. A. (2001). The colonial origins of comparative development. TheAmerican Economic Review, 91(5), 1369–1401.

Anderson, J. (1979). A theoretical foundation for the gravity model. The American Economic Review, 69(1),106–116.

Anderson, J., & van Wincoop, E. (2003). Gravity with gravitas: A solution to the border puzzle. The AmericanEconomic Review, 93(1), 170–192.

Baldwin, R., & Taglioni, D. (2006). Gravity for dummies and dummies for gravity equations. NBER WorkingPaper No. 12516, Cambridge, MA.

Bergstrand, J. (1985). The gravity equation in international trade: Some microeconomic evidence and empiricalevidence. The Review of Economics and Statistics, 67(3), 474–481.

Bergstrand, J. (1989). The generalized gravity equation, monopolistic competition, and factor-proportionstheory in international trade. The Review of Economics and Statistics, 71(1), 143–153.

Belenkiy, M., & Riker, D. (2010). Face-to-face exports: The role of business travel in trade promotion. Journalof Travel Research, 51(5), 632–639.

Blum, B. S., & Goldfarb, A. (2006). Does the internet defy the law of gravity? Journal of International Economics,70(2), 384–405.

Buch, C. M. (2005). Distance and international banking. Review of International Economics, 13(4), 787–804.Ceglowski, J. (2006). Does gravity matter in a service economy? Review of World Economics, 142(2), 307–329.CEPII. (2010). dist-cepii.xls. Centre d’Études Prospectives et d’Informations Internationales. Download at

http://www.cepii.fr/anglaisgraph/bdd/distances.htm, access: January, 2010.Choi, C. (2010). The effect of the Internet on service trade. Economics Letters, 109(2), 102–104.Christen, E. (2011). Time zones matter: The impact of distance and timezones on service trade. Paper presented at

the thirteenth annual Conference of the European Trade Study Group, Copenhagen, Denmark, September2011.

CIA. (2010). The World Factbook. Retrieved May 2010 from https://www.cia.gov/library/publications/the-world-factbook/.

Djankov, S., Freund, C., & Pham, C. S. (2010). Trading on time. The Review of Economics and Statistics, 92(1),166–173.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

248 B. Dettmer

Duranton, G., & Storper, M. (2005). Rising trade costs? Agglomeration and trade with endogenous transactioncosts. CEP Discussion Paper No 683, Centre for Economic Performance.

Eaton, J., & Kortum, S. (2002). Technology, geography and trade. Econometrica, 70(5), 1741–1779.Evans, C. L., & Harrigan, J. (2005). Distance, time, and specialization: Lean retailing in general equilibrium.

The American Economic Review, 95(1), 292–313.Fink, C., Mattoo, A., & Neagu, I. C. (2005). Assessing the impact of communication costs on international

trade. Journal of International Economics, 67(2), 428–445.Freund, C., & Weinhold, D. (2002). The internet and international trade in services. The American Economic

Review, 92(2), 236–240.Gallup, J. L., Sachs, J., & Mellinger, A. (1999). Geography and economic development. International Regional

Science Review, 22(2), 179–232.Glaeser, E. L., Laibson, D., & Sacerdote, B. (2002). An economic approach to social capital. The Economic

Journal, 112(483), 437–458.Gould, D. M. (1994). Immigrant links to the home country: Empirical implications for US. Bilateral trade flows.

The Review of Economics and Statistics, 76(2), 302–316.Grossman, G. (1998). Comment. In J. A. Frankel (Ed.), The regionalization of the world economy. Chicago and

London: The University Press of Chicago.Grünfeld, L. A. & Moxnes, A. (2003). The intangible globalisation: Explaining the patterns of international

trade in services. Working Paper 657, Norwegian Institute of International Affairs.Guiso, L., Sapienza, P., & Zingales, L. (2009). Cultural biases in economic exchange? The Quarterly Journal of

Economics, 124(3), 1095–1131.Hall, R. E., & Jones, C. I. (1999). Why do some countries produce so much more output per worker than others?

The Quarterly Journal of Economics, 114(1), 83–116.Hanafizadeh, M. R., Saghaei, A., & Hanafizadeh, P. (2009). An index for cross-country analysis of ICT

infrastructure and access. Telecommunications Policy, 33(7), 385–405.Harrigan, J., & Venables, A. J. (2006). Timeliness and agglomeration. Journal of Urban Economics, 59(2),

300–316.Hattari, R., & Rajan, R. S. (2008). Sources of FDI flows to developing Asia: The role of distance and time zones.

ADB Institute Working Paper No. 117, Asian Development Bank Institute.Hausman, W. H., Lee, L., & Subramanian, U. (2005). Global logistics indicators, supply chain metrics and

bilateral trade patterns. World Bank Policy Research Working Paper No. 3773, The World Bank.Head, K., Mayer, T., & Ries, J. (2009). How remote is the offshoring threat? European Economic Review, 53(4),

429–444.Hummels, D. (1999). Have international transportation cost declined? University of Chicago. mimeo. available

at: https://www.gtap.agecon.purdue.edu/resources/download/1238.pdfHummels, D. (2001). Time as a trade barrier. GTAP Working Paper No. 18, Purdue University.Jones, R. W., & Kierzkowski, H. (1990). The role of services in production and international trade: A theoretical

framework. In R.W. Jones & A. O. Krueger (Eds.), The political economy of international trade: Essays inhonor of Robert E. Baldwin (pp. 31–48). Cambridge, MA: Blackwell.

Kamstra, M. J., Kramer, L. A., & Levi, M. D. (2000). Losing sleep at the market: The daylight saving anomaly.The American Economic Review, 90(4), 1005–1011.

Keller, W., & Yeaple, S. R. (2013). The gravity of knowledge. The American Economic Review, 103(4),1414–1444.

Kikuchi, T. (2003). Interconnectivity of communications networks and international trade. Canadian Journalof Economics, 36(1), 155–167.

Kikuchi, T. (2009). Time zones as a source of comparative advantage. Review of International Economics, 17(5),961–968.

Kikuchi, T., & Iwasa, K. (2010). A simple model of service trade with time zone differences. InternationalReview of Economics and Finance, 19(1), 75–80.

Kimura, F., & Lee, H.-H. (2006). The gravity equation in international trade in services. Review of WorldEconomics, 142(1), 92–121.

Knack, S., & Keefer, P. (1997). Does social capital have an economic payoff? A cross-country investigation. TheQuarterly Journal of Economics, 112(4), 1251–1288.

Krugman, P. (1980). Scale economies, product differentiations, and the pattern of trade. The American EconomicReview, 70(5), 950–959.

Lejour, A., & De Paiva-Verheijden, J.-W. (2007). The tradability of services within Canada and the EuropeanUnion. The Service Industries Journal, 27(4), 389–409.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 249

Lennon, C. (2009). Trade in services and trade in goods: Differences and complementarities. WIIW WorkingPapers 53, Vienna Institute for International Economic Studies.

Limao, N., & Venables, A. J. (2001). Infrastructure, geographical disadvantage, transport costs, and trade. TheWorld Bank Economic Review, 15(3), 451–479.

Loungani, P., Mody, A., & Razin, A. (2002). The global disconnect: The role of transactional distance and scaleeconomies in gravity equations. Scottish Journal of Political Economy, 49(5), 526–543.

Marjit, S. (2007). Trade theory and the role of time zones. International Review of Economics and Finance,16(2), 153–160.

Maurer, A., Magdeleine, J., & d’Andrea, B. (2006). International trade in services – GATS, statistical conceptsand future challenges. World Trade Organization, mimeo.

Melitz, J. (2004). Geography, trade and currency union. In A. Volbert, G. Furstenberg & J. Melitz (Eds.),Monetary union and hard pegs: Why, Effects on trade, financial development and stability. Oxford: OxfordUniversity Press.

Mirza, D., & Nicoletti, G. (2004). What is so special about trade in services? GEP Research Paper 2004/02,Leverhulme Centre for Research on Globalization and Economic Policy.

Nordas, H., Pinali, E., & Geloso Grosso, M. (2006). Logistics and time as a trade barrier. OECD Trade PolicyWorking Papers No. 35, OECD Publishing.

OECD. (2008). Statistics on international trade in services, vol. II. Detailed tables by partner countries. Paris:OECD publishing.

OECD. (2010). ITS international trade by commodity statistics, SITC revision 3. Paris: OECD publishing.Patterson, P. G., & Cicic, M. (1995). A typology of service firms in international markets: An empirical

investigation. Journal of International Marketing, 3(4), 57–83.Paulson, E. (1996). Travel statement on jet lag. Canadian Medical Association Journal, 155(1), 61–66.PTB. (2010). Zeitzonen. retrieved January 2010 from http://www.ptb.de/de/org/4/44/441/zeit.htm.Poole, J. P. (2010). Business travel as an input to international trade. Department of Economics, University of

California, Santa Cruz.Portes, R., & Rey, H. (2005). The determinants of cross-border equity flows. Journal of International Economics,

65(2), 269–296.Portes, R., Rey, H., & Oh, Y. (2001). Information and capital flows: The determinants of transactions in financial

assets. European Economic Review, 45(4–6), 783–796.Raff, H., & von der Ruhr, M. (2007). Foreign direct investment in producer services: Theory and empirical

evidence. Applied Economics Quarterly, 53(3), 299–321.Rauch, J. E., & Casella, A. (2003). Overcoming informational barriers to international resource allocation:

Prices and group ties. The Economic Journal, 113, 21–42.Rauch, J. E., & Trindade, V. (2002). Ethnic Chinese networks in international trade. The Review of Economics

and Statistics, 84(1), 116–130.Stein, E., & Daude, C. (2007). Longitude matters: Time zones and the location foreign direct investment. Journal

of International Economics, 71(1), 96–112.Tang, L. (2006). Communication costs and trade of differentiated goods. Review of International Economics,

14(1), 54–68.Tharakan, P. K. M., Van Beveren, I., & Van Ourti, T. (2005). Determinants of India’s software exports and

goods exports. The Review of Economics and Statistics, 87(4), 776–780.Tharakan, P. K. M., & Van Beveren, I. (2003). Exports and distance in a digitized world. GEP Research Paper

2003/10, Leverhulme Centre for Research on Globalisation and Economic Policy.Tinbergen, J. (1962). Shaping the world economy: Suggestions for an international economic policy. New York:

Twentieth Century Fund.Vietze, C. (2011). What’s pushing international tourism expenditures? Tourism Economics, 17(2), 237–260.Walsh, K. (2008). Trade in services: Does gravity hold? Journal of World Trade, 42(2), 315–334.World Bank. (2010). World development indicator database. Washington, DC: The World Bank.WTO. (2010). Agreement list. Retrieved March 2010 from http://rtais.wto.org/ui/PublicAllRTAList.aspx,

Geneva: World Trade Organization.Yasar, M., Lisner, D., & Rejesus R. M. (2012). Bilateral trade impacts of temporary foreign visitor policy. Review

of World Economics, 148(3), 501–521.Zaheer, S. (2000). Time zone economics and managerial work in a global economy. In P. C. Early & H.

Singh (Eds.), Innovations in international and cross-cultural management (pp. 339–353). Thousand Oaks:Sage.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

250 B. Dettmer

Appendix

Table A1.1. Reporter countries (Panel A)

Country Business services Commercial services Merchandise

Australia 1999–2006 1999–2006 1999–2005Austria 1999–2005 1999–2005 1999–2006Belgium 2002–2005 2002–2005 1999–2006Canada – 1999–2005 1999–2006Czech Republic 2000–2005 2000–2005 1999–2006Denmark 1999–2005 1999–2003,2005 1999–2006Spain 1999–2005 1999–2005 1999–2006Finland 1999–2005 1999–2005 1999–2006France 1999–2005 1999–2005 1999–2006Germany – 1999–2005 1999–2006Greece 1999–2005 1999–2005 1999–2006Hong Kong – – 1999–2004Hungary 1999–2005 1999–2005 1999–2006Ireland 1999–2005 1999–2005 1999–2006Italy 1999–2005 1999–2005 1999–2006Japan 1999–2005 1999–2005 1999–2006Korea 1999–2005 1999–2005 1999–2006Luxembourg 2002–2005 – 1999–2006Netherlands 1999–2005 1999–2005 1999–2006Norway 1999–2005 1999–2001,2004–2005 1999–2006New Zealand 2004–2005 1999–2005 1999–2006Poland 2004–2005 2004–2005 1999–2006Portugal 1999–2005 1999–2005 1999–2006Slovak Republic 1999–2005 1999–2005 1999–2006Sweden 1999–2005 1999–2005 1999–2006United Kingdom 1999–2004 1999–2005 1999–2006United States – 1999–2006 1999–2006

Table A1.2. Reporter and partner countries excluded in Panel B

Australia UTC + 8 hours to UTC + 10 hoursBrazil UTC − 3 hours to UTC − 5 hoursCanada UTC − 3.5 hours to UTC − 8 hoursIndonesia UTC + 7 hours to UTC + 9 hoursKazakhstan UTC + 5 hours to UTC + 6 hoursDR Congo UTC + 1 hours to UTC + 2 hoursMexico UTC − 6 hours to UTC − 8 hoursMongolia UTC + 7 hours to UTC + 8 hoursRussian Federation UTC + 3 hours to UTC + 12 hoursUnited States UTC − 4 hours to UTC − 10 hours

Table A1.3. Non-OECD partner countries

Albania Costa Rica Liechtenstein SamoaAlgeria Cote d’Ivoire Lithuania Saudi ArabiaAntigua and Barbuda Croatia Macao SenegalArgentina Cyprus Madagascar Serbia and Montenegro

(Continued)

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 251

Table A1.3. Continued.

Azerbaijan Dominica Malawi SingaporeBahamas Dominican Republic Malaysia South AfricaBahrain Ecuador Malta Sri LankaBangladesh Egypt Mauritania SudanBarbados El Salvador Mauritius SwazilandBelarus Estonia Mongolia Syrian Arab RepublicBermuda Fiji Morocco ThailandBhutan Guatemala Myanmar TokelauBolivia Honduras Nepal Trinidad and TobagoBotswana Hong Kong New Caledonia TunisiaBrazil India Nicaragua UkraineBrunei Indonesia Nigeria Unit Arab EmiratesBulgaria Iran Norfolk Island UruguayCape Verde Iraq Pakistan VanuatuCayman Island Jamaica Papua New Guinea VenezuelaChina Jordan Peru VietnamChina Taipei Kazakhstan Philippines Virgin IslandColombia Latvia Romania YemenCook Island Lebanon Russia

Source: own compilation.

Table A2. Documentation of variables and data sources

Variable Description Source

Business ServicesExports

Include leasing, legal, accounting, auditing, book-keeping, tax consulting, business and managementconsulting, advertising, and research

OECD (2008)

CommercialServices Exports

Include communication services, construction services,insurance services, financial services, computer andinformation services, royalties and license fees, otherbusiness services, personal, cultural and recreationalservices

OECD (2008)

MerchandiseExports

Includes total trade based on SITC Rev. 3. OECD (2010)

Distance The weighted distance (lndistw) expresses thepopulation-weighted average of the great-circledistance between the 20 highest populated cities(regional metropolises) within the (reporter- andpartner-) countries.

CEPII (2010)

TIME capital Shortest time zone difference between trading partners’capital city, daylight saving time is not included.

PTB (2010)

TIME mean Time zone difference between trading partners’ basedon the mean time zone in the country for countrieswith multiple time zones. Daylight saving time is notincluded.

PTB (2010)

TIME min Time zone difference between trading partners’ basedon the minimum time zone difference between thecountries with multiple time zones and its tradingpartners. Daylight saving time is not included.

PTB (2010)

(Continued)

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

252 B. Dettmer

Table A2. Continued.

Variable Description Source

Office hoursoverlap

Number of overlapping office hours between tradingpartners based on the capital time zone difference.(ranges between zero and 8 hours or zero and 10hours overlap)

PTB (2010)

GDP Nominal GDP in current USD World Bank (2010)Mobile telephone

subscriptions(per 100 people)

Mobile cellular telephone subscriptions are subscrip-tions to a public mobile telephone service usingcellular technology, which provide access to the publicswitched telephone network. Include post-paid andprepaid subscriptions.

World Bank (2010)

Telephonesubscribers (per100 people)

Total telephone subscribers are mobile and fixed-linesubscribers

World Bank (2010)

Personal computers(per 100 people)

Personal computers are self-contained computersdesigned to be used by a single individual.

World Bank (2010)

Internet users (per100 people)

Internet users are people with access to the worldwidenetwork.

World Bank (2010)

Air transport,registered carrierdeparturesworldwide

Registered carrier departures worldwide are domestictakeoffs and takeoffs abroad of air carriers registeredin the country

World Bank (2010)

English language The dummy for English language is equal to one if theEnglish language is spoken by at least 50 per centof the population in both countries. In addition tothose countries where English is an official language,the following countries are considered as widelyspoken English countries: Korea, Denmark, Iceland,Norway, Egypt, Ethiopia, Aruba, NetherlandsAntilles, Panama, Suriname, Israel, Bahrain, Kuwait,Qatar, Jordan, Indonesia, Maldives, Nepal, Pakistan,Sri Lanka, Thailand, Vietnam, Samoa, Tonga, andAmerican Samoa.

CIA (2010)

Colonial ties Dummy if countries have colonial relationship CEPII (2010)Religion Dummy if both countries have the same religion.

A country is considered as a Muslim, Catholic,Protestant, Orthodox, or other religion (Buddhism,Hinduism) if at least 60 per cent of the populationbelongs to the denomination.

CIA (2010)

EU27 Dummy for both countries membership in the EuropeanUnion

WTO (2010)

GATS Dummy if both countries have a trade agreement forservices according to GATS Article V, into force until2002

WTO (2010)

GATT Dummy if both countries have a trade agreement forgoods according to GATT Article XXIV, into forceuntil 2002

WTO (2010)

Source: own compilation.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

International Service Transactions 253

Table A3. Summary statistics

Obs. Mean Std.Dev Min Max

Panel AMerchandise export (million USD) 43680 932.278 6435.391 2.0e–06 316700.000Commercial services exports (million

USD)6108 514.399 1954.467 0 32820.000

Business services (million USD) 4972 147.326 611.246 0 11730.000Communication services (million USD) 4257 9.771 36.563 0 512.254Construction services (million USD) 4118 16.189 67.881 0 1285.665Computer and information services

(million USD)31819 0.372 7.540 0 537.200

Financial services (million USD) 4433 42.384 265.321 0 6514.336Insurance services (million USD) 4584 24.524 171.703 0 5474.386Log Merchandise export 43680 16.267 3.536 0.693 26.481Log Commercial services exports 5681 17.063 3.067 3.912 24.214Log Business services 4396 16.121 2.776 6.846 23.185Log Communication services 2649 14.288 2.535 6.745 20.054Log Construction services 2095 14.936 2.526 7.030 20.975Log Computer and information services 672 14.503 2.445 2.996 20.102Log Financial services 2849 14.378 3.128 6.709 22.597Log Insurance services 2557 14.963 2.720 3.401 22.423Log Distance 6023 8.362 1.048 5.042 9.880Time capital 6108 3.959 3.507 0 12Time mean 6108 4.085 3.508 0 12Time min 6108 3.548 3.184 0 12Office 8 6108 4.298 3.077 0 8Office 10 6108 6.102 3.388 0 10Log GDPit−1 6108 26.497 1.512 23.729 30.147Log GDPjt−1 5879 25.734 1.711 19.252 30.147Log minMobilet−1 6103 2.983 1.382 −3.912 4.681Log minPhonet−1 6108 3.907 0.965 −0.953 5.584Log minPCt−1 6096 2.296 1.228 −2.438 4.472Log minNett−1 6105 2.213 1.427 −4.605 4.472Log minAirt−1 6086 11.217 1.438 6.114 16.115

Panel A (OECD)Time capital 2908 4.006 3.841 0 12Log minMobilet−1 2907 3.671 0.645 1.258 4.664Log minPhonet−1 2908 4.460 0.422 2.635 5.081Log minPCt−1 2903 2.965 0.772 0.971 4.472Log minNett−1 2908 2.898 0.892 −0.357 4.406Log minAirt−1 2908 11.578 1.370 7.505 16.115

Panel A (Non-OECD)Time capital 3200 3.916 3.173 0 12Log minMobilet−1 3196 2.358 1.564 −3.912 4.681Log minPhonet−1 3200 3.405 1.043 −0.953 5.584Log minPCt−1 3193 1.688 1.249 −2.438 4.356Log minNett−1 3197 1.493 1.542 −4.605 4.408Log minAirt−1 3178 10.886 1.420 6.114 16.115

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014

254B.D

ettmer

Table A4. Correlation matrix

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)

(1) Dist 1(2) Time cap 0.760 1(3) Time mean 0.741 0.964 1(4) Time min 0.737 0.970 0.956 1(5) Overlap8 −0.771 −0.976 −0.943 −0.944 1(6) Overlap10 −0.766 −0.996 −0.961 −0.965 0.988 1(7) GDPit−1 0.079 0.085 0.111 0.042 −0.096 −0.089 1(8) GDPjt−1 −0.273 −0.069 −0.056 −0.093 0.047 0.062 0.011 1(9) minMobilet−1 −0.148 0.052 0.042 0.056 −0.048 −0.051 0.057 0.414 1(10) minPhonet−1 −0.177 0.089 0.085 0.087 −0.082 −0.087 0.045 0.439 0.857 1(11) minPCt−1 −0.125 0.120 0.111 0.115 −0.113 −0.120 0.043 0.310 0.674 0.757 1(12) minNett−1 −0.123 0.127 0.115 0.125 −0.121 −0.126 0.066 0.381 0.866 0.847 0.729 1(13) minAirt−1 0.044 0.141 0.148 0.112 −0.140 −0.142 0.300 0.283 0.384 0.391 0.534 0.407 1

Source: own calculations.

Dow

nloa

ded

by [

Um

eå U

nive

rsity

Lib

rary

] at

20:

29 0

6 O

ctob

er 2

014