Exports and Information Spillovers

Public Disclosure Authorized POLICY RESEARCH Public Disclosure Authorized Exports and Information wp% Spillovers PAPER 2474 A developing country's good (or bad) export performance in one marketcan affectits future export performance Alessandro Nicita not only in the samemarket Marcelo Olarreaga but also in "neighboring" markets.This happens if AlessandroNicita Public Disclosure Authorized WORKING 4"7'4 importers in different countries share information about a particular exporter's performance or if exporters themselvestake advantage of the information acquired while exporting to similar markets.Thus, through Public Disclosure Authorized information spillovers,export success(or failure) becomes cumulative acrossmarkets. The World Bank Development Research Group Trade November 2000 | POLICYRESEARCH WORKINGPAPER2474 Summary findings Exporters' performance in a particular market may affect their future exports to the rest of the world. Importers may base their future transaction decisions on the information revealed by exporters' past performance in other countries. Similarly, exporters acquire valuable information on foreign consumer tastes, product standards, or customs administration that may profitably be used in future transactions with other countries. Nicita and Olarreaga estimate the effects of these information spillovers across markets on the export patterns of four developing countries (Egypt, the Republic of Korea, Malaysia, and Tunisia). A dollar increase in exports to the United States generates on average an extra 2 to 14 cents of exports to the rest of the world in the next period. Social and ethnic networks seem to reinforce these information spillovers, especially in developing countries, where they appear to be geographically more concentrated. The exception is China and to some extent Hong Kong, probably reflecting a geographically more diversified migration pattern. The exchange of information among current and potential export markets can significantly affect a developing country's export performance. Bilateral information spillovers across markets are negligible or nonexistent for exports from the United States, where there is less need to create a reputation in international markets. Similarly, Egypt's good export performance would be more easily noticed in Argentina or India (where the market is small) than would increased exports to France or the United States. This paper-a product of Trade, Development Research Group-is part of a larger effort in the group to understand the role of information in developing countries' integration into world markets. Copies of the paper are available free from the WorldBank, 1818 H StreetNW,Washington, DC 20433. Please contactLili Tabada, room MC3-333, telephone 202-4736896, fax 202-522-1159, email address ltabada@worldbank.org. Policy Research Working Papers are also posted on the Web at www.worldbank.org/research/workingpapers. The authors may be contacted at anicita@worldbank.org or molarreaga@worldbank.org. November 2000. (36 pages) The PolicyResearchWorkingPaperSees disseminatesthe findingsof work in progressto encouragethe exchangeof ideasabout development issues.An objective of the series is to get the findings out quickly, even if the presentations are less than fully polisbed. The papers carry the names of the autbors and should be cited accordingly. The findings, interpretations, and conclusionisexpressedin this paper are entirely those of the authors. They do not necessarilyrepresent the view of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center Exports and Information Spillovers* Alessandro Nicita** Marcelo Olarreaga*** JEL classification numbers: F10, F13, F14 Keywords: Export, Information spillovers, Developing Countries * We wouldlike to thankSimonEvenett,CarstenFink,Jean-MarieGrether,OlivierLumenga-Neso,Jaime de Melo,Garry Pursell,MauriceSchiff,IsidroSoloaga,DavidTarr,AnthonyVenablesandparticipantsin a WorldBank Trade seminarfor very helpfulcommentsand suggestions.We are also gratefulto LiliTabada for excellentassistance. ** Development ResearchGroup(DECRG),WorldBank, 1818H StreetNW, Washington,DC 20433. Email:anicita@worldbank.org. **DECRG,World Bank,Washingtonand CEPR,London.Email:molarreaga@worldbank.org Non-technical summary There is little doubt that in any type of business, individual relationships among trading partners are extremely valuable and can determine the success or decline of a firm. The information acquired through each interaction is seen, by both buyers and sellers, as an investment, which will bring future benefits. The need for information among business partners is probably more pronounced in the case of international transactions. The export success of a firm (and ultimately a country) will depend on the quality of its business relationships. Through repeated interactions, both exporters and importers will acquire valuable information on reliability in terms of the credit and delivery of their trading partners. The information created by the business relationship may spill over to other exporters and importers. These information spillovers may not necessarily be limited to national borders. Importers' direct experience with an exporter may generate valuable information that can be used as an important guide in deciding future transactions with other countries. This implies that a good export performance in one market will not only have positive effects in the same market in the future, but may also positively affect export performance in "'neighboring" markets through information spillovers. Similarly, exporters may acquire valuable information regarding the functioning of customs administrations or foreign consumer tastes, which could profitably be used in future transactions with other countries. This paper focuses on the importance of these international information spillovers on the export performance of four developing countries, which have experienced varied export performance in the last decade: Egypt, Korea, Malaysia and Tunisia. Using the United States as an example of a market where information is generated, we found that an extra dollar exported to the United States by Korea generates on average an increase in Korean exports to I the rest of the world of 14 cents. A similar figure for Malaysia is 10 cents, whereas in Egypt and Tunisia, the figure is close to 2 cents. We also found that some developing countries' export markets, such as India and Argentina, generate larger increases in exports to the rest of the world through information spillovers. Spillovers tend, however, to be geographically more concentrated, in the sense that most of the additional exports occur within a few geographically close countries. The exception in developing countries in terms of geographic concentration as a source of information spillovers to the rest of the world is China and to some extent Hong Kong, where infornation spills over to geographically diversified countries. This is probably due to the less geographically-concentrated migration pattern of China and Hong Kong. These results suggest an important role for public or private export information agencies in developing countries. Diffusion of export information across firms can significantly contribute to the export performance of a country. The analysis also detected the presence of externalities at the two-digit industry level, which suggests that there is room for co-operation through bundling of export offers across firms in the same two-digit industry. The presence of these information spillovers also implies that one bad deal or poor performance in one market not only hurts exporters in that particular market, but may also hurt them in other markets. This suggests an important role for quality controls, such as ISO standards, that can be publicly or privately organized. 2 "Most foreign buyers prefer to give orders to firms that already have considerable export experience and require little instruction and assistance. This is one reason success is cumulative" (Vinod Thomas, John Nash et al., 1993, p. 128). 1. Introduction There is little doubt that in any type of business, individual relationships among trading partners are extremely valuable and can determine the success or decline of a firm. Through repeated business transactions, a certain degree of trust is developed between sellers and buyers, which reinforces the relationship. Future transactions become more profitable through a better understanding and knowledge of each others' needs. The information acquired through each interaction is seen, by both buyers and sellers, as an investment which will bring future benefits.I The need to create bilateral information among business partners is probably more pronounced in the case of international transactions. The export success of a firm (and ultimately a country) will depend on the quality of its business relationships.2 Through repeated interactions, both exporters and importers will acquire valuable information on reliability in terms of the credit and delivery of their trading partners. It will also provide knowledge of the functioning of custom administrations, foreign market tastes, product quality, standards, certification and design (Egan and Mody, 1992, Evenson and Westphal, 1995, Grossman and Helpman 1991, Rhee, Ross-Larson and Pursell, 1984 and World Bank, 1997).3 For a potential exporter to successfully enter an export market, it needs to build a reputation as a reliable supplier and learn about market tastes and structures. The process 'Repeated interactionalso solvesthe problemof asymmetricinformationas shownby Riordan(1986). 2 As reportedin a studydone by Egan andMody (1992)basedon a surveyof US importers,one bad shipmentcan leadto a completebreakof a businessrelationwith low reputationsuppliers. 3Note that the recentempiricalliteraturehas foundvery littleevidenceof "learningby doing" associated with exportactivitieson the productivityof the firm.Seefor exampleClerides,Lachand Tybout(1998)for a studyof Colombianfirms andBemardand Jensen(1999)for a studyof US firms. 3 of building a reputation may be costly. However, reputation building may also show some multiplier effects, as the individual relationship established between an exporter and an importer will typically generate information spillovers beyond the two trading partners. Importers may use other importers, who have had direct experience with potential suppliers, as a source of information on the performance of alternative exporters (World Bank, 1989). This effect may be reinforced by information spillover effects on the exporter side, as export activities generate a better understanding of how foreign markets work. This is valuable information for future transactions. Also, the export success of a firm in some markets may generate demonstration effects for other firms, which become aware of potential opportunities in foreign markets. Export promotion or industry agencies, both public and private, may also help diffuse this information across firms. Information spillovers may not necessarily be limited to national borders. Importers who have had direct experience with an exporter may generate valuable information that can be used as an important guide in deciding future transactions in other countries. This implies that a good export performance in one market will not only have positive reputation effects in the same market in the future, but may also positively affect export performance in "neighboring" markets through information spillovers. Similarly, exporters acquire valuable information regarding the functioning of customs administrations, foreign consumer tastes, shipping procedures, and distribution networks, which could profitably be used in future transactions with other countries.4 Social or ethnic networks may help the international transmission of these information spillovers in various ways: helping to match buyers and sellers across borders; creating market similarities; easing the transmission of these flows across borders; and serving as a deterrent for opportunistic behavior, as in Rauch (1999) or Rauch and Trindade (1999). Even in countries with well-developed judiciary systems, an important share of what makes a successful business deal will typically lie outside the contract established by An alternative explanation for the observation of this export reputation spillovers across borders is the existence of production networks, where plants of the sarne network are located in different countries. Initiating business with one plant in the network allows much easier access to other buyers within the network (see Kaminski and Ng, 1999). 4 4 trading partners (McLaren, 1999). Trust provided by ethnic networks therefore has an important business value. As the empirical section will reveal, the importance of these ethnic networks in explaining the export performance of developing countries is indirectly confirmed in our study. Information flows among countries will be facilitated by the presence of these ethnic networks. The objective of this paper is to try to identify the importance of these international information spillovers to the export performance of four developing countries which have experienced varied export performance in the last decade: Egypt, Korea, Malaysia and Tunisia. The choice of countries is deliberate in the sense that we wanted to have a set of developing countries from different regions and at different stages of development.5 The questions we will be asking are, for example: Does the export performance of Egypt in France affect exports from Egypt to countries which share large information flows with France?; and if so: How important is Egypt's export performance in France in explaining Egypt's export performance in the rest of the world? In the empirical section we found that information spillovers had important effects in the export performance of these four developing countries. Interestingly enough, we also found that information spillovers have little effect on the export pattern of the US. This suggests that the role of information spillovers is more important in developing countries, where the need for building a reputation in international markets is larger. Taking the United States as an example of a market where information is generated, an extra dollar exported to the United States by Korea generates on average an increase in Korean exports to the rest of the world of 14 cents. A similar figure for Malaysia is 10 cents, whereas in Egypt and Tunisia the figure is close to 2 cents. We also found that some developing countries' export markets, such as India and Argentina, generate larger increases in exports to the rest of the world through information spillovers. They tend, mayalsoplayan importantrole.See countrieswhereregionalism we purposelyexcludeLatinAmerican Nicita,OlarreagaandSoloaga(2000)for ananalysisof exportinformation spillovers in a regionalcontext. 5 5 however, to be geographically more concentrated, which is probably due to their less diversified migration pattern. The rest of the paper is structured as follows. In section 2, we discuss the importance of information flows across countries and the different measures used in this paper to capture bilateral information flows. Section 3 develops a simple model with export information spillovers across nations. Section 4 focuses on the econometric model, and section 5 reports the results for the four developing countries in our sample. Section 6 quantifies the importance of export information spillovers for the export performance of these four developing countries. Section 7 outlines the conclusions. 2. Information Flows across countries Related literature has suggested several ways to capture information flows. First, Rauch (1999) suggests that geographic proximity facilitates the exchange of trade related information. The rationale for this is twofold. First, communication costs might affect the flow of information, second, information from other buyers may be more valuable the closer these other buyers are from the domestic market in terms of tastes, product-designs and other cultural factors.6 Thus, information flows would follow a distance decay function and would be larger among relatively close countries. Rauch and Trindade (1999) suggest the use of the share of common ethnic population to capture information flows. The idea is that ethnic networks facilitate the exchange of information. The larger the share is of Chinese population in two countries, for example, the larger the information flow between these two partners. 6 SeeRhee,Ross-LarsonandPursell(1984)or Evensonand Westphal(1995). 6 Two other proxies are suggested by Portes and Rey (1999). These authors measure information flows using bilateral telephone calls and bilateral trade.7 For telephone calls, the intuition is straightforward; whereas for bilateral trade the idea is as discussed above, that business relationships are subject to the exchange of information. Rauch and Trindade (2000) suggest the use of the number of periodicals and newsletters devoted to international trade and commerce as proxy for trade related information. Because we are interested in bilateral information flows, we propose the use of bilateral trade in periodicals and newspapers as proxy for trade-related information flows (SITC rev 1. 8922). The exchange of newspapers will not only include directly trade related information, as in the case of the Journal of Commerce or Export Channel in the US, Made for Export in Europe, Asian Channel in Hong Kong, and Gazeta Mercantil in Latin America. It will also reflect cultural similarities and bilateral immigration patterns, which will also be important determinants of "indirectly" trade related information flows, such as taste similarities across countries. All of these proxies have advantages and disadvantages when it comes to their empirical application. In the empirical section we will test each of them, except the bilateral share of ethnic population between partners. The reason for the exclusion is that we would need a matrix of bilateral migration patterns in the world which is, to our knowledge, unavailable (see Zlotnik (1998) for a discussion of data availability for international migration).8 However, Rauch (1999) has argued that distance may be correlated with the existence of these ethnic networks. We believe that bilateral telephone calls and newspaper trade also capture their presence. However, we will argue that bilateral newspaper trade is the more adequate proxy to capture information flows across countries. Results reported in section 5 are estimated using newspaper trade; but other proxies provided robust results. See footnote 16 in Portes and Rey (1999) for the use of aggregate bilateral trade as a proxy for information flows between countries. 8 Rauch and Trindade (1999) focus on the effects of Chinese networks on bilateral trade relations in the world and therefore had smaller data requirements. 7 7 There are at least two reasonswhy distancemay be an imperfectproxy: it fails to capture size and culturalor historical effects.To illustratethe importanceof size effects,note that using distanceas a proxy would imply that the exchangeof informationbetweenArgentina and Uruguaywould be largerthan betweenArgentinaand Mexico.However,the exchange of information measured by newspaper trade (for example) suggests that the exchange between Argentinaand Mexicowas more than ten times larger in 1995due simply to the relative size of their markets (see Table 1). Regardingthe failure of distance to capture cultural and historicallinks, note that its use would imply that the exchangeof information between Australia and the United Kingdom would be smaller than the exchange of information between China and Australia, since the latter are geographically closer. However,cultural factors such as languageand coloniallinks imply that newspapertrade between Australia and the United Kingdomwas 200 times larger than between Australia and China in 1995. Bilateralaggregatetrade may solve some of the problemsassociatedwith distance,as size and culturallinks are importantdeterminantsof trade. However,bilateralaggregatetrade is also determinedby many other factors,such as comparativeadvantage,and may therefore be poorlycorrelatedwith bilateralinformationflows.9 More importantly,these information flows may be crucial for an exporter in a third country, who may be able to build a valuablereputationin one of the two markets,and may benefit from the informationflows between the two countriesto increase its export performancein the other market. As an example, Ireland's newspapers trade with the United Kingdom represented almost 98 percent of its total trade in newspaperin 1995,whereasthe share of the United Kingdomin Ireland's total trade is close to 30 percent. Similarly,20 percent of Kuwait's newspaper trade is done with Egypt;but Egypt only represents1.5 percentof Kuwait's total trade. Bilateraltelephone calls may also solve the problemsdescribedabove associated with the use of distance or bilateral trade as a proxy for informationflows across countries. But it may raise other problems, when trying to capture informationflows among developing 8 countries. First, the data that is available today at the International Telecommunications Union does not cover a wide range of developing countries (of the 60 countries considered as potential export markets in our sample, 41 are developing countries), and its time dimension is limited (there is no data available before 1985). Second, telephone calls may be a very expensive means of exchanging information for developing countries due to the high cost of international calls. Newspapers are probably the cheapest way to widely disseminate information.1 0 2.1 How largeare bilateralnewspaperflows? The value of world newspaper exports was close to 4 billion dollars in 1995, and represented 0.1 percent of world exports. The growth of bilateral newspaper trade during the period 1969-1997 is close to 10 percent in nominal terms per year. Germany is the world's largest trader of newspapers with a total trade (exports plus imports) of 1.2 billion dollars in 1995. It is closely followed by the United States, which was involved in 25 percent of world newspaper trade in the same year (as either an importer or an exporter). These flows can also be relatively important in developing countries. Brazil's total trade of newspapers was close to 87 million dollars in 1995. The total value of newspaper trade between Brazil and Chile was 30 million dollars, whereas between Singapore and Malaysia it was close to 10 million dollars. Table 1 below provides the share of bilateral newspaper trade and total newspaper trade for a selected number of countries. 3. Export Information Spillovers across countries Export information spillovers are defined as the set of information flows that are generated in a particular export market and that will affect export and import decisions between the original exporter and importers in the rest of the world. They are illustrated in Figure 1. 9 To seethis, acceptforthe momentthat trade flowsare exclusivelydeterminedby factorendowments.Two countrieswith identicalfactorendowmentswillnot trade with eachother,but mayhave a significant exchangeof information. 9 The exporter's performance in country k generates information spillovers (the dashed lines) into two locations. First, it gives feedback to the original exporter on information about customs procedures, product standards, and tastes in foreign markets, which may help in the next period its export performance in a third market: country c. Second, importers in country c learn about the reliability and product quality of the original exporter by observing the exporter's behavior in the rest of the world. This will affect import decisions in the next period. The size of export information spillovers between countries k and c will therefore depend not only on the export performance of the exporter in country k, but also on the extent of information exchange between country c and country k. 3.1 A simple model Firms face constant marginal costs in country 0 (the exporter). Marginal transport costs are also constant. Thus, total cost of exporting from country 0 to country c at period t, denoted CC t, is given by: C", = [a +, Tdc ]xc,(1 where a is the marginal cost of producing the export good; r, is the marginal transport cost; d, is the distance from country 0 to country c; and xc, are exports from country 0 to country c. Export markets for exports originating in country 0 are segmented, which combined with the assumption of constant marginal costs of production allows us to deal independently with each export market. '° Theuseoftheinternetmaychangethisin thefuture,butit wasclearlynotan instrument forexchangeof information duringtheperiodunderexamination here(1969-1997). 10 Demand for each product in each market is derived from a quasi-linear and additive utility function, which freezes substitution and income effects in demand. Each sub-utility function is quadratic so that demand functions are linear. Units are chosen so that the slope of the linear demand function equals 1. There are information spillover effects, in the sense that demand today for exports of country 0 depend on the market share of country 0 in the previous period. The larger the market share of country 0 in period t is, the larger the demand it will face in period t+] .II Information about past export performance also spills over from other countries as suggested in Figure 1. The exporter's past market shares in rest of the world markets also affects the level of demand for its products in country c. Thus, information spillovers are here modeled on the demand side, but could be similarly introduced on the supply side. In our empirical section, we will not be able to disentangle between demand and supply spillovers, and therefore our estimates will be a mixture of both. Inverse demand for exports of country 0 to country c at period t is therefore given by: (2) I c p,,, =a -XCI +A SC,1-I+ 1,1-S, where a is a parameter; pc, is the price of exports in country c at time t; sc/, is the share of country 0 in total import demand of country c in the previous period; X >0 then captures the own-market effect; IC,, is a transposed vector of information flows between country c and all other (potential) export markets of country 0; each element is defined by the share of each rest of the world country in country c's total information flows with the world;'2 S,, is a vector of market shares of country 0 in each market. Thus, 9 > 0 captures the export-market information spillovers across markets. '" See Frootand Klemperer(1989)or Farrell(1986)for a discussionof the relevanceof past marketshares in determiningfuture demand. 12 We use sharesinsteadof actual flowsfor severalreasons.First,it makes interpretationeasier. Second,if the proxy used for informationflowshas a time dimension,then it will avoid our havingto deal with the potentialbiasthat this may introduceinto our econometricestimates.Finally,the power of someof the 11 Free-entry into each export market ensures that: a-x, l +AI "S, r 1 1 =a+ rd (3) Solving (3) for xc,t yields xcI=(a-a)-rdc+Asc,.-l +6Icj ,S 11 (4) Equation (4) implies that an increase in export performance in any location will affect exports to all other countries in the next period through information spillovers. 4. Econometric model and data We will try to capture information spillovers at the product line level and therefore we will use bilateral trade data (60 countries) at the 3-digit of the SITC classification for manufacturing products. For each of the exporting countries in our sample, we will then use the whole sample of potential export products to all countries where a product has been exported at least once during the period 1969-1997. Trade data sources are from national sources compiled by the United Nations in Comtrade's data base. Information flows are captured using the four proxies described in section 2. In the case of distance, we use the matrix of inverse bilateral distance between countries. For total trade, we use the share of bilateral trade with each rest-of-the-world country in the importer's (country c) aggregate trade with the rest of the world. For international phone calls, we use the share of international phone call minutes between country c and each of the rest-of-theworld countries. Finally, newspaper trade is calculated as the share of bilateral newspaper statisticteststhatwe useto testtheerrortermforpotentialcorrelationacrosscountriescruciallydependson the standardizationof this matrix(seeFloraxand Folner, 1992or Anselin,1999). 12 trade with each rest-of-the-world country in country c's total newspaper trade with the rest of the world. When using newspaper trade and total trade, the proxy for information flows contains a time dimension, as data is available throughout the period. However, the data on telephone calls had very little time dimension before 1992 and no data before 1985. Therefore we took the year 1992 as the base year and used the number of bilateral minutes of phone calls for 1992 values, or the closest year (1991 or 1993) for which there is data available, as a proxy for information flows.'3 In the case of distance, there is obviously no time dimension either.'4 More detailed information on variable construction can be found in the Data Appendix. Results reported in section 5 are obtained using newspaper trade as a proxy for information flows. Other proxies yielded robust estimates, though newspaper trade generally yielded more efficient results. We will base the empirical analysis on a stochastic version of equation (4) and test for information spillovers by testing the significance of the paramneter0 . The non-existence of exports in many products across trading partners leads to a large presence of zeroes in our endogenous variables (89 percent of censoring in the case of Egyptian exports, 40 percent for Korea, 60 percent for Malaysia and 91 percent for Tunisia). To correct for the bias introduced by censoring, we estimate equation (4) for each of the four exporting countries (Egypt, Korea, Malaysia and Tunisia) using a tobit technique: 15 X;C, Xp,c = °if 'if x* >0 (5) xp-, <° and x> =(a-a)-rdc +AAs , +0Ic,tlSp, +6p,c 13 Note that due to the lack of data available at the Intemational Telecommunication Union on minutes of bilateral phone calls our estimation for bilateral phone calls only include 41 of the 60 countries in our sample. However, as suggested before, the results were consistent with what we report in section 5 using newspaper trade as a proxy. 14 Note that to avoid identification problems, we need the elements of the information flow vector to be exogenous. Note that this is the case for all proxies. For example, exports of Egypt to France in period t cannot determine the information flows between France and Germany at period t-1. '5 We alternatively used a two-stage tobit technique as in Maddala (1983, p. 221-222) and a two-part model, which yielded similar results to the ones reported in section 5. Generally, results were more efficient when 13 where xpc r are exports of Korea (for example) of product p, to country c in period t; and . p c is the error term. In Anselin (1999) terminology this specification is an implicit pure time-space recursive model. In this paper, the presence of information flows across countries lead to a space recursive model, in the sense that the export performance in country k will affect the export performance in all other countries in our sample through information flows. It is implicit time recursive because the spatial lag of the endogenous variables is expressed in terms of market share and not levels of exports, which allows us to avoid problems related to lagged spatial endogenous variables. The lack of simultaneity in the spatial correlation allows us, in principle, to abstract from problems of correlation between the spatially lagged variable (ICSp,,l) and the error term. However, we will test for error spatial correlation in the next section. One may be tempted to add fixed country or product effects into equation (5), but as suggested by Anselin (1999), this would lead to inconsistent estimates, in which case a random effect specification should then be considered. Assuming country-specific unobserved effects,16 the error term becomes: wC, + where wc, is independently and identically distributed (i.i.d.) across countries and time, and 1p c, is i.i.d across products, countries and time. The estimation of information spillovers (0) through equation (5) may be biased by the absence of some other important variables related to comparative advantage aspects or other types of externalities that are absent in (5). To control for these, we add four variables to the right hand side of (5). using either the simple tobit or the two-stage tobit technique, which may be due to the structure of the data, as discussed in Leung and Yu (1996). 16 In the empirical section we also considered a product-specific component for the error term and obtained similar results to the ones reported in section 5. 14 First, size may matter. The larger the import market, the larger are exports to this market. To control for the size of the import market, we introduce size, which is defined as total imports of productp in country c at period t (purged of exports to country c). Second, comparative advantage aspects may also affect our measure of information spillovers. In some products the exporter may be a "natural exporter" and in others not. To control for this we introduce ca, which is defined as total exports to the rest of the world, denoted ca (again purged). It could also be interpreted as capturing "learning by doing" or economies of scale in export activities, which would not be related to export informnation spillovers across markets.17 Third, bilateral trade preference and cultural links may also affect our estimates. To capture this, we introduce gravity, which is total exports of the source country to country c (again purged). Gravity can also be interpreted as capturing all the explanatory variables of a gravity equation for the export country (including cultural links, language, regional trade agreements, etc.). It may also be seen as across product extemalities within the own market. Finally, we also control for possible within sector externalities by taking the sum of bilateral exports at the 2 digit level of SITC classification (excluding the export product of each observation) and denote it 2digit. Thus:'8 fp',1, Xp, X* ,t > ° if = O , if Xp,C,t < and x*,, =(a-a)- T 0 d, +Asp,c,,_ +OIC,1-Sp,-,+ 01 sizepC, + 0 2 cap, + t03 gravityc, + t, digitp2d,c,t + C",/ + p,C, andJensen(1999). Bernard anddatasources,seeDataandVariableAppendix. ofvariableconstruction Fora moreformaldescription LachandTybout(1998)or 17SeeClerides, 8 2 (6) 15 A time dummy is also introduced in all estimations. All parameters are positive and therefore expected signs are given by the signs in front of the parameters. Results reported in the next section also correct for heteroscedasticity using Huber correction method. The reported R2 is calculated following Veall and Zimnermnann (1994). As shown in their study, the more traditional McFadden (1973) R2 has a downwards bias in large samples with a high degree of censoring.'9 5. Empirical Results We estimated equations (5) and (6) for each of four countries in our sample. Table 2 reports results of these estimations in the first and second column for each country (we discuss results reported in the third column later). All variables have the expected sign and are statistically significant at the 10 percent level or less, except for within two-digit industry externalities in the case of Tunisia.20 Note that the introduction of the four control variables in the estimation of equation (6) does not change the significance of the estimates of equation (5). We further test for the statistical significance of the information spillovers by performing an F-test on the residuals of the estimation of equations (5) and (6) for each of the exporting countries, as suggested by Florax and Folmer (1992) in the presence of spatially lagged variables.21 All F-tests rejected the null hypothesis of absence of information spillovers at the 1 percent level. We also found no evidence of spatial autocorrelation in the residuals of the regressions of equation (6) reported in the second columns of Table 2. As 19 The Veall andZimmermnan (1994) R2 is givenby: (p is the predictedvalueof the endogenousvariable; x7 i, -X, -_;,) +N6 2 ;-where is the meanof this predictedvalue;N is the numberofobservations and a istheTobitmaximumlikelihood estimatedof theerrorvariance. 2 We also try to capture bilateral information spillovers within two-digit industries by weighting the twodigit industry variable by bilateral information flows, but results were insignificant for all countries except Egypt. 21 The F-test is given by: (E'ER - EUEU )/(EUEU /(N - q)), where ER and EU are the error vectorsof the restricted and unrestricted model, N is the number of observations and q the number of explanatory variables. 20 16 suggested by Anselin and Hudak (1992) we performed a Lagrange Multiplier test that corrected for the panel aspect of our data (product-year observations). We could not reject the null hypothesis of no spatial autocorrelation.22 These results suggest that exports from any of the four countries in our sample (Egypt, Korea, Malaysia and Tunisia) to a particular market will be affected by past export performance in the rest of the world, through bilateral information spillovers between the particular export market and rest of the world countries. The rise of globalization in the last decade may also affect these bilateral information flows. As communication costs plummet, information flows across countries may become cheaper.2 3 In order to check for any structural change in the importance of these information flows during the period, we introduce a new variable denoted Globalization. It is constructed as the product of the Information Spillovers vector and a vector that takes the value 0 for any observation before 1985 and 1 otherwise. Results of these estimations are reported in the third column of table 2 for each of the source countries. For Tunisia, the estimated coefficient is positive and significant. This suggests that after 1985, information spillovers increase their importance in the determination of Tunisia's export pattern. Note that none of the other variables changed sign or significance. In the case of Egypt, Korea and Malaysia, the estimated coefficient is negative but insignificant, with the exception of Malaysia, where it is significant at the 10 percent level. This suggests that bilateral information flows tend to lose their relevance for Accordingto Anselin(1999)the Lagrangemultiplieris amongthe most powerfultests for spatial autocorrelationin largesamples.The Lagrangemultipliertest was calculatedas: 22 LM =EYEPJtEP P/ E'E/N/[ n trace(X,j + ,2whereE is thepartitioned vectorofthe error term for productp and time t (i.e., it variesacrosscountries),E is the entireerrorvector,j, is the matrixof standardizedbilateralinformationflowsat time t and npis the numberof products. 23 Note that worldnewspapertradehas been growingat a 10percentrate on averageover the periodand there is littleevidencethat therehas beena structuralchangeforthe worldas a wholeduringthe 1969-1997 period.For a discussionof the evolutionof communicationcostin the last two decades,see WorldBank (1999). 17 Malaysia after 1985.24Again, note that none of the other variables changed sign or significance. Interestingly, when estimating equations (5) and (6) for exports from the United States, the overall fit of the equation and significance were similar to the ones reported in Table 2, except for the absence of information spillovers, which was highly insignificant (t-statistic of 0.3). This suggests that these information spillovers across countries tend to be more important for developing countries where the need for getting noticed and establishing a good reputation as a reliable exporter is higher. In more developed countries, the need for establishing a reputation as a reliable business partner may tend to be less rigid, perhaps due to the existence of more developed legal systems, the country's overall reputation, and a larger market share in world markets. All of these may make less relevant the intercountry information flow between potential export market. In order to explore this hypothesis, we created a new variable to capture the notion of world reputation as an exporter. This is constructed as the market share of a particular product of the source country in world markets, and is denoted WdRep for world reputation. In this variable, the information flow between two potential export markets becomes irrelevant and only its market share in the world market would be of interest for potential importers in other markets.25 Results of the estimation of equation (6) including WdRepare reported in Table 3 for the four source countries (Egypt, Korea, Malaysia and Tunisia) and the United States. In the case of Egypt and Tunisia, WdRep does not enter significantly into the regression (in the case of Egypt it has a negative sign), but all other variables keep their significance including information spillovers. For Korea, Malaysia and the United States, WdRep enters significantly, suggesting that world reputation may be enough to establish a reputation as reliable exporter in these countries. This is particularly true for the United States, where information spillovers do not seem to have any explanatory power. In the 24 Similarresultswereobtainedforthefourcountriesusing1980or 1990insteadof 1985 asthetimebreak. 18 case of Korea and Malaysia, however, information spillovers still play a significant role (in Korea they are significant at the 20 percent level). These results somewhat confirm our hypothesis above: as countries developed, the intercountry information spillovers among potential export markets become less relevant. At low levels of development or country reputation, inter-country information spillovers tend to be relatively more important for exporters. We also tested for the time length of information spillovers by introducing into equation (6) lags of 2 and 3 years for the Information Spillovers variables. The estimated coefficients for these lags were smaller and highly insignificant for Korea, Malaysia and Tunisia, suggesting a short memory in world markets. Again, none of the other variables changed sign or significance. In the case of Egypt, however, the estimates suggested some memory in world markets for Egyptian exports, as the lagged variables were statistically significant. Finally, note that it is difficult to infer from the reported coefficients the importance of information spillovers in determining the export patterns of these four countries. In other words, what is the effect, of a one-dollar increase in export penetration in one particular market today on exports to the rest of the world in the next period? Do some export markets generate more information spillovers than others? The answer to these questions is given in the next section. 6. Where are Export Information spillovers larger? The presence of information spillovers implies that an increase in exports in one particular market will affect the whole system through information spillovers and will therefore be followed by increases in exports to the rest of the world. To see how a one-dollar increase 25 For a more detailed description of the construction of WdRepsee Data Appendix. 19 in exports to country k affect exports to country c in the next period differentiate equation (6) with respect to a dollar increase in exports to market k: dp C, = ) p,C, where Mpk, l c*k,r-- a ( 'p,c,t 0 ic"k,t-l ppc,t-d - SP,k,i-I *p,k,i-I (7) are total imports of product p by country k at time t-l; ic<k,) is the information exchange between countries c and k at period t-1; and 7rP,c,is the probability that x is non-zero conditional on the explanatory variables. Table 4 provides such bilateral estimates for a selected number of countries at the mean (over products and time) using the estimates of the second column of table 2 for each of the source countries (Egypt, Korea, Malaysia and Tunisia). It suggests that, in the case of Korean exports, for example, a one-dollar increase in exports to the United States generates an increase of 0.1 cents in exports to China in the next period and 0.2 cents of extra exports to Germany. Similarly, a dollar increase in Egyptian exports to India generates an increase of 0.1 cents in exports to Hong Kong and 0.04 cents to Great Britain in the next period. To obtain the total effect in the rest of the world of a dollar increase on exports to country k through infornation spillovers, one needs to add the left and right-hand-sides of (6) over all rest-of-the-world countries. That is, Axp= dxp,c,lt= ZrPc,1 ick,1 T (l1- Sp,k,i-l &p,k,i-1 (8) Estimates at the mean (across products and time) of a one-dollar increase in exports to each of the countries in our samnpleby Egypt, Korea, Malaysia and Tunisia are given in Table 5. A one-dollar increase in exports to the United States provides, through information 20 spillovers, an increase in exports to the rest of the world of 14 cents for Korea, 10 for Malaysia, and 2 for each of Egypt and Tunisia. The United States, however, is not the largest market in terms of generating export information spillovers for these countries. The largest spillovers for our four source countries are found in Argentina, Colombia, Hong Kong and India (and France for Tunisian exports). The reason has to do with the size of the United States' market rather than the lack of information flows between the United States and the rest of the world. A large market implies that a dollar increase has very little effect on market shares, which is the force behind the spillovers, and therefore generates smaller spillovers for the same amount of information flow. Countries that generate the lowest spillovers are Panama, Israel, Trinidad and Tobago, and Korea. The reason for the lack of information spillovers from these markets has to do with the small amounts of information flows between these countries and the rest of the world, partly due to their small size. 6.1 Geographic concentration of information spillovers and ethnic networks Information spillovers can be larger in developing countries. However, they tend to be geographically more concentrated. In Table 5, the figures in italics show the share of the top four countries to which information spillovers are generated from each market. The spillovers of Argentina, India, Pakistan, Colombia and Singapore tend to be relatively concentrated (regardless of which is the exporting country) compared to some developed markets such as Spain, United States and Japan. China and Hong Kong, to some extent, are an exception as information spillovers generated from these developing countries appear to be geographically diversified. This probably reflects a less geographically concentrated migration pattern. The large concentration of information spillovers from developing countries also reflects the fact that information spillovers occurred across relatively close markets. In the case of 21 Egyptian exports to Argentina, for example, about 70 percent of the information spillovers generated in Argentina are received by other Mercosur countries (Brazil, Chile, Paraguay and Uruguay). On the other hand, in the case of information spillovers generated in the United States, the top four receivers are Canada, Israel, Trinidad and Tobago and Saudi Arabia, which are geographically dispersed. Note also that they represent only 30 percent of the total information spillovers generated in the American market. To illustrate the relatively high regional concentration of information spillovers generated in developing countries, Figure 2 plots the cumulative distribution for information spillovers generated from Egyptian exports in Argentina, China, France, Germnany,Hong Kong, India, Japan, Tunisia and the United States. The horizontal axis is ordered in terms of geographic distance between each of these countries and the rest of the world. It is clear from figure 2 that spillovers generated from the United States and France tend to be geographically more diversified than spillovers generated from Argentina or India. More than 75 percent of information spillovers from Argentina and India are transmitted within the five closest countries. For China and Hong Kong, however, only around 50 percent of the total information spillovers are generated within the five closest countries, whereas in the case of the United States the figure is around 25 percent.26 7. Concluding Remarks The exchange of information among (potential) export markets can significantly affect the export performance of a developing country. A good (or bad) export performance in one market can affect not only the future export performance in the same market, but also in "neighboring" markets. This will occur if importers in different countries share information on the performance of a particular exporter or if exporters themselves take, advantage of the information acquired while exporting to similar markets. Thus, through 26 Similarresultsare foundfor the otherexportingcountriesin our sample. 22 information spillovers, the overall export success (or failure) becomes cumulative across markets. Exports of the four developing countries in our sample (Egypt, Korea, Malaysia and Tunisia) are significantly affected by these bilateral information exchanges; in particular, for those in earlier stages of development (Egypt and Tunisia). We found, however, that bilateral information spillovers across markets are negligible (or non-existent) for exports from the United States, where the need for creating a reputation in international markets is smaller. In the case of the United States (and to some extent Korea and Malaysia) the overall world reputation of the exporter seems to be a more important determinant than the bilateral exchange of information across export markets. For Egyptian or Tunisian exporters, bilateral information exchanges across export markets is the dominant determinant of their export performance. A dollar increase of Egyptian exports to France generates almost 8 cents of extra exports to the world in the next period through exchange of information between France and the rest of the world. An dollar increase to the United States generates 2 cents of exports to the rest of the world. Similarly, an additional dollar to India can generate as much as 18 cents of exports to the rest of the world. Similar figures for Tunisian exports are 9 cents for information spillovers generated from France, 2 cents from the United States and 7 cents from India. As suggested by the figures above, some developing countries generate larger benefits for exporters through information spillovers. That is the case of India for Egyptian exporters. Also, information spillovers generated in Argentina provide an additional 12 cents in the next period for each dollar exported to Argentina. This is above the benefits from an extra dollar to France or the United States. The reason for this is probably that a relatively good export performance of Egypt in India or Argentina could be more easily noticed than increased exports to the United States or France, given the relatively smaller size of the Indian or Argentinian market. 23 However, export information spillovers generated from developing countries also tend to be geographically more concentrated. The reason has to do with the geographic concentration of the exchange of information flows in developing countries, which partly reflects international migration patterns and the presence of ethnic networks, as suggested by James Rauch's work. The exception in our data is China and Hong Kong. This can be partly explained by the geographically more diversified migration patterns of China and Hong Kong. These results suggest an important role for public or private export information agencies in developing countries. Diffusion of export information across firms can significantly contribute to the export performance of a country. The analysis also detected the presence of externalities at the two-digit industry level, which suggests that there is room for cooperation, for example, through bundling of export offers across firms in the same twodigit industry. More importantly, perhaps, is that the presence of these information spillovers implies that one bad deal or poor performance in one market not only hurts exporters in that particular market, but may also hurt them in other markets. This suggests an important role for quality controls, such as ISO standards, that can be publicly or privately organized. 24 References Anselin, Luc (1999), Spatial Econometrics, mimeo, University of Texas. Anselin, Luc and Sheri Hudak (1992), "Spatial econometrics in practice", Regional Science and Urban Economics 22, 509-536. Bernard, Andrew and Bradford Jensen (1999), "Exceptional exporter performance: cause, effect or both?" Journal of International Economics 47, 1-25. Clerides, Sofronis, Saul Lach and James Tybout (1998), "Is learning by exporting important? Micro-dynamic evidence from Colombia, Mexico and Morocco", Quarterly Journal of Economics, August. Egan, Mary and Ashoka Mody (1992), "Buyer-seller links in export development", World Development 20, 321-334. Evenson, Rovert and Larry Westphal (1995), "Technological change and technology strategy" in T.N. Srinivasan and Jere Behrman, eds, Handbook of Development Economics, vol. 3, North Holland. Farrell, Joseph (1986), "A note on inertia in market share", Economics Letters 21, 73-75. Florax, Raymond and Henk Folmer (1992), "Specification and estimation of spatial linear regression models", Regional Science and Urban Economics 22, 405-432. Frankel, Jeffrey (1997), Regional trading blocs in the World Economic System, Institute for International Economics. Froot, Kenneth and Paul D. Klemperer (1989), "Exchange rate pass-through when market share matters", American Economic Review 79:4, 637-654. 25 Grossman, Gene and Elhanan Helpman (1991), Innovation and Growth in the World Economy, MIT Press. Kaminski, Bartolomiej and Francis Ng (1999), "Central European Economies and Production Fragmentation: participation in EU networks of production and marketing", mimeo, The World Bank, Washington DC. Leung, Siu Fai and Shihti Yu (1996), "On the choice between sample selection and twopart models", Journal of Econometrics 72, 197-229. McLaren, John (1999), "Supplier relations and the market context: a theory of handshakes", Journal of International Economics 48, 121-138. Maddala, G. (1983), Limited dependent variables in economics, Cambridge University Press, Cambridge. Portes, Richard and Helen Rey (1999), "The determinants of cross-border equity flows", CEPR discussion paper # 2225, London, UK. Rauch, James (1999), "Networks versus markets in international trade", Journal of International Economics 48, 139-150. Rauch, James and Alessandra Casella (1998), "Overcoming informational barriers to International Resource Allocation: Prices and Group Ties", NBER Working paper #6628. Rauch, James and Vitor Trindade (1999), "Ethnic Chinese Networks in International Trade", NBER working paper # 7189. 26 Rauch, James and Vitor Trindade (2000), "Information and Globalization: wage comovements, labor demand elasticity, and conventional trade liberalization", NBER working paper # 7671. Rhee, Ung, Bruce Ross-Larson and Garry Pursell (1984), Korea's competitive edge: managing the entry into world markets, John Hopkins University Press. Riordan, Michale (1986), "Monopolistic competition with experienced goods", Quarterly Journal of Economics 48, 139-150. Thomas, Vinod, John Nash et al. (1993), Best Practices in Trade Policy Reform, Oxford University Press. Veall, Michael and Klaus Zimmnermann(1994),"Goodness of Fit Measures in the Tobit Model", Oxford Bulletin of Economics and Statistics 56:4, 485-499. World Bank (1989), Reputation in Manufactured Goods Trade, Industry series paper #21, Washington. World Bank (1997), World Development Report 1997: The state in a changing world, Oxford University Press. World B ank (1999), World Development Report 1998/1999: Knowledgefor development, Oxford University Press. Zlotnik, Hania (1998), "International Migration 1965-96: an overview", Population and Development Review 24:3, 429-467. 27 Appendix VariableConstruction The variablesare constructedas follows: market_share* , == p = ; where xpxc,are exportsof productp from the sourcecountryto countryc at time t; andm[Tp is definedas total importsof countryc of productp at time t; gravity _ effect, Xp,C; xp.C - p comp _ advantagep,t = size, P =mCT Xct -XP _ ; Ct Y,xp,c, 2 digit effect -xpcr -xp,c,/-] F p=2d T z MpC,_,- - MPc,t.-l pc2d where2digitincludesall the tariff lineswithinthe same2-digitcategoryofthe SITCclassification; . MC.PJ-1 The elementj of the vectorof informationflows, IC,,, are definedas: Xc÷*kj-l where Ycek ,_, is the bilateralflowof informationbetweencountryc andcountryk; exportsand importsof newspaperswhenproxiedwith newspapertradeor minutesof phoneincomingand outgoingphonecalls betweencountriesc andk, whenproxiedwith telephonephonecalls. IC,-] Sp,,1 1, - k,p,p-j iC-k,p,1-; k¢c where Sk,p,t, is the market shareof productp in countryk and S..,-, is its vector form; globalization = I, . S,,,, LO , if year>1985 ,if year< 1985 28 Data Sources: Trade data sourcesare from the nationalsourcescompiledby the UnitedNationsin Comtrade'sdatabase. The analysisis carriedusingmanufacturingproductstradedata givenof SITCrev.I classificationat the 3 digit level. Dataon newspapertradeare alsoprovidedby the Comtrade'sdatabase(SITCrev. 1, code8922) Distancedata are calculateusinggeographicaldistancebetweencountries'capitals. Data on telecommunicationhas been providedby STARSdatabaseby the InternationalTelecommunication Union. 29 Figure 1: The Role of Information Spillovers for Export Performance *.Informnation .Spillovers at t-1 V , m-m-""-X""aountry C......... o :, ANe~wspapet Lt_> >,/ Flowst-I Country ~~k \ ffi :' +at Export Flows time t ~~~Market Sare ~~~~at t-1\I .Information .Spillovers att-1I * /g" t 30 ~Exporter ) Figure 2: Geographic concentration of information spillovers (selected countries) 0 Argentina China Frace Germayna Hol<Kong g India 0 o , i_J__ S1 X -1 _ ,~~~~~5 _ Turisia USA 1 59 Ranked Distance 31 Table 1: Bilateral Newspaper flows in 1995 for selected countries (percentage of importing country total flow).a Reporfern Partner AM.O. AOgantao 0.0% B-rant Cil. 22.0% 41.0% ChiN Spain Fnnc 0.0°A 4.0% 0.0% HongKong IhdoI_W Ihd UK G.n,rAY 0.0% 0.0% 0.0% 0.0% 0.0% Konz Meniro k Wepe. PhkWstn Shaoow. 0.0% 0.0% 0.0% 3.0% 0.0% 0.0% 0.0% rub *- na ThlOnn USA 0.0% 0.0% B-11 24.0% 0.0% 47.0% 0.0% 2.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 1.0% 3.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 1.0% Chile 33 0% 34.0% 0.0% 0.0% 1.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 1.0% 0.0% 0.0% 0.0% 0.0% 0.0%h Chine 0.0% 0.0% 00% 0.0% 0.0% 0.0% 0.0% 0.0% 10.0% 0.0% 0.0% 0.0% 2.0% 0.0% 0.0% 0.0% 0.0% 1.0% 2.0% 0.0% SFai 16 0% 6.0% 3.0% 0.0% 0.0% 5.0% 10.0% 5.0% 0.0% 0.0% 0.0% 3.0% 0.0% 1.0% 17.0% 0.0% 0.0% 0.0% 0.0% 0.0% France 1.0% 2.0% 0.0% 2.0% 15.0% 0.0% 9.0% 9.0% 1.0% 7.0% 1.0% 27.0% 4.0% 1.0% 2.0% 0.0% 0.0% 0.0% 0.0% 3.0% UK 1.0% 2.0% 1.0% 6.0% 28.0% 8.0% 00% 7.0% 3.0% 30.0% 4.0% 15.0% 15.0% 0.0% 0.0% 12.0% 5.0% 14.0% 0.0% 7.0% Gennany 3.0% 4.0% 0.0% 8.0% 20.0% 13.0% 11.0% 0.0% 2.0% 6.0% 4.0% 23.0% 10.0% 3.0% 2.0% 0.0% 5.0% 1.0% 2.0% 3.0% HongKg 0.0% 0.0% 0.0% 42.0% 00% 0.0% 0.0% 0.0% 0.0% 9.0% 2.0% 0.0% 10.0% 26.0% 0.0% 6.0% 4.0% 8.0% 33.0% 1.0% ndnnenos 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 1.0% 0. 0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0,0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 1.0% 0.0% 3.0% 0.0% 11.1y 1.0% 50% 0.0% 0.0% 4.0% 11.0% 6.0% 6.0% 1.0% 0.0% 1.0% 0.0% 1.0% 3.0% 0.0% 0.0% 0.0% 0.0% 2.0% 1.0% Japn 0 0% 5.0% 0.0% 16.0% 0.0% 0.0% 2.0% 1.0% 18.0% 1.0% 00% 0.0% 0.0% 39.0% 0.0% 3.0% 0.0% 27.0% 42.0% 2.0% Ko... 0.0% 0.0% 0.0% 0.0% 00% 0.0% 0.0% 0.0% 10.0% 13.0% 0.0% 0.0% 8.0% 0.0% 0.0% 0,0% 0.0% 5.0% 0.0% 0.0% Menno 3.0% 0.0% 1.0% 0.0% 4.0% 0.0% 0 0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 4.0% UClanoeS 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 2.0% 0.0% 2.0% 0.0% 1.0% 0.0% 0.0% 0.0% 0.0% 17.0% 1.0% 0.0% PIahs- 0.0% 0.0% 0.0% 0.0% 0 0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 1.0% 0.0% 0.0% gara 0.0% 0.0% 0.0% 4.0% 0.0% 0.0% 1.0% 0.0% 8.0% 0.0% 18.0% 0.0% 15.0% 13.0% 0.0% 50.0% 6.0% 0.0% 2.0% 0.0% aTordn 0.0% 0.0% 0.0% 5.0% 00% 0.0% 0.0% 0.0% 16.0% 0.0% 0.0% 0.0% 11.0% 0.0% 0.0% 2.0% 0.0% 1.0% 0.0% 0.0% 15.0% 2.0% 3.0% 10.0% 3.0% 10.0°h 8.0% 25.0% 2.0% 17.0% 9.0% 65.0% 12.0% 6.0% 5.0% 12.0% 0.0% UK G.,r-ny HongKong Indsb J".n Kor- Mco Moy Pakistn SIg.pona Tawan USA 813.5 1255.0 115.1 23.1 62.1 22.9 7.5 64.7 29.6 1098.0 USA 2.0% Tolal AUI-aina MilL S 78.8 12.0% Brand 86.7 1.0% Chin nn Spaid 63.1 14.5 285.8 Fr 852.0 61.4 2.5 Wa Itly 5.5 348.4 aThe cells in each column give the trade share of each of the partner countries in the total exports and imports of newspaper of the reporter. For example, 24 percent of Argentina's total trade of newspaper is undertaken with Brazil and 22 percent of Brazil's total trade of newspapers is undertaken with Argentina. The total value of newspaper trade for each country (export plus imports) is given in the last row of the table in million dollars. 32 Table 2: Estimating Export Information Spilloversa Tunisia Egypt Constant Market share (t-1) Distance Information Spillovers Timetrend -1567.87** (406.89) 6568.25*** (2342.55) -0.1209*(0.3427) 84614.15** (21631.4) 31.28** (7.52) -1489.38*** (386.75) 5729.15** (2044.65) -0.1000*** (0.028) 39446.7*** (8916.44) 12.94** (4.17) 0.0002*** (0.00004) 0.0370** (0.0059) 9131.5** (3985.88) 0.0368*** (0.0093) -1489.12*** (389.46) 5728.90*** (2045.96) -0.1000** (0.028) 39292.6*" (10394.4) 12.93*** (4.24) 0.0002** (0.00004) 0.0370*** (0.0059) 9129.83** (3999.24) 0.0367*** (0.0095) 356.19 (8023.57) -1852.8** (374.65) G8154.9*** (25202.3) -0.2997*** (0.1070) 95156.5*** (28702.0) 9.897*** (2.722) -2239.4*** (447.67) 59717.6-* (21205.28) -0.2430*** (0.0810) 51373.7*** (17212.7) 8.332*** (1.967) 0.0005** (0.0002) 0.3844*** (0.0044) 19449.36 (20422.5) 0.0321*** (0.0099) -2222.33*** (445.81) 59420.6(21130.7) -0.2413*(0.0807) 46501.7*** (16311.3) 6.314*** (1.722) 0.0005** (0.0002) 0.0385*** (0.0044) 19403.5 (20409.3) 0.0318*** (0.0099) 167519.4*** (39092.8) 0.187 20.21*** 131424 0.2376 157.28-* 131424 0.2375 226.99*131424 0.300 16.80*** 131419 0.314 138.20** 131419 0.314 140.26*** 131419 Size Gravityeffects 2digit effects Compadvantage Perioddummy (>1985) Rsquared Wald chi squared #observations Malaysia Korea -11833.2*** -9539.5*** -12599.7*** -9305.1*** (4713.7) (2990.6) (3050.03) 107235.9'** (34942.6) -1.1245*** 82452.8'** (23982.7) -1.0679*** 82413.3*** (24027.6) -1.068*** -11802.5* -11263.6** (4808.1) (3341.28) Market share (t-1) Distance 81482.4*** (25871.7) -0.7221- 64010.9*** (16437.5) -0.4214*** (0.2001) (0.1552) (0.1540) (0.2683) Information Spillovers Timetrend 46707.5** (16900.0) 555.05*** (155.23) 30371.3** (12164.8) 154.92*** (50.377) 0.0259(0.0039) 0.0058*-* (0.0005) 19379.6* (9936.4) 0.0101(0.0031) 47538.3** (20903.95) 187.72-* (61.574) 0.0259*** (0.0039) 0.0058** (0.0005) 19467.1* (9978.6) 0.0102*** (0.0032) -30628.2* (16831.3) 154895** (56221.9) 589.11*** (163.49) 62507.1** (24535.8) 296.02*** (83.43) 0.0131 *** (0.0018) 0.0084*** (0.002) 15538.4*** (5210.8) 0.0148*** (0.0053) 101672.3*** (32941.3) 310.12*** (86.503) 0.0131*** (0.0018) 0.0084-* (0.0020) 15389.8*** (5209.92) 0.01512** (0.0053) -56689.4' (29245.3) 0.110 20.01*** 131418 0.328 2151.87*** 131418 0.328 2484.01*** 131418 0.195 18.54"* 131417 0.303 336.01*** 131417 0.305 344.55*** 131417 Constant - Size Gravityeffects 2digiteffects Compadvantage Perioddummy (>1985) Rsquared Wald chi squared #observations (3534.8) 64726.2*** (16641.6) -0.4173* (0.2751) Estimation technique is Tobit to control for censored data. Figures in parenthesis are White-Robust standard errors. *** stands for significance at the 1 percent level; * at the 5 percent level and * at the 10 percent level. 33 (0.2753) Table 3: World Reputation and Information Spilloversb Egypt Tunisia Korea Malaysia USA Information 41929- 331230*' 24805 29225' -1765 Spillovers (12666) (16674) (18645) (17235)) (1701) -1931 (1253) 24558 (38487) 39454** (15979) 225845(84786) 43714* (14683) 0.24 0.32 0.33 0.31 0.19 304-** 131424 166- 2853*** 424*** 19191*** 131419 131418 131417 131830 World Reputation R squared Wald chi squared # observations Estimation technique is Tobit to control for censored data. Figures in parenthesis are White-Robust standard errors. * * * stands for significance at the I percent level; * * at the 5 percent level and * at the 10 percent level. For the four countries, all other coefficients are within two standard deviations of the results reported in the third column of Table 2. For the United States all other coefficients are qualitatively similar to the ones reported for the other four countries in Table 2 (with the exception of distance, which is insignificant). b 34 Table 4: Bilateral export information spillovers for selected countries (dollars)a EGYPT ARGENTINA CHINA 0.00000 ARGENTINA FRANCE UK GERMANY HONG KONG JAPAN SINGAPORE USA 0.00005 0.00001 0.00021 0.00000 0.00000 0.00001 0.00000 0.00007 0.00014 0.00117 0.00225 0.00030 0.00683 0.00008 0.00031 0.00011 INDIA CHINA 0.00001 FRANCE 0.00029 0.00002 UK 0.00006 0.00007 0.00129 GERMANY 0.00289 0.00041 0.00388 0.00238 HONG KONG 0.00000 0.00095 0.00002 0.00024 0.00004 INDIA 0.00003 0.00033 0.00011 0.00163 0.00105 0.00711 JAPAN 0.00004 0.00110 0.00014 0.00021 0.00054 0.00078 0.00007 SINGAPORE 0.00000 0.00003 0.00002 0.00065 0.00002 0.00643 0.00129 0.00129 USA 0.00059 0.00014 0.00035 0.00103 0.00042 0.00018 0.00020 0.00051 TUNISIA ARGENTINA FRANCE UK CHINA 0.00000 ARGEN77NA 0.00098 0.00175 0.00002 0.00002 0.00013 0.00002 0.00022 0.00142 0.00039 0.00021 0.00065 0.00087 0.00006 0.00045 0.00149 0.00003 0.00083 0.00128 0.00105 0.00586 0.00018 0.00071 0.00364 0.00044 0.00114 0.00043 HONG KONG INDIA 0.00012 0.00013 JAPAN SINGAPORE USA 0.00000 0.00011 0.00000 0.00010 0.00000 0.00004 0.00004 0.00005 0.00018 0.00149 0.00000 0.00000 0.00000 0.00000 0.00001 0.00015 0.00000 0.00000 0.00011 0.00014 0.00280 0.00004 0.00005 0.02137 0.00065 0.00068 0.00021 0.00002 0.00407 0.00049 0.00114 0.00000 0.00000 0.00000 0.00037 0.00002 0.00011 0.00004 0.00043 0.00000 0.00003 0.00000 GERMANY 0.00027 CHINA 0.00000 FRANCE 0.00000 0.00000 UK 0.00092 0.00003 0.00037 0.00164 GERMANY 0.00006 0.00006 0.00242 HONGKONG 0.00043 0.00000 0.00000 0.00001 0.00003 INDIA 0.00003 0.00000 0.00000 0.00006 0.00005 0.00000 JAPAN 0.00232 0.00001 0.00022 0.01575 0.00519 0.00001 0.00001 SINGAPORE 0.00001 0.00000 0.00000 0.00177 0.00215 0.00000 0.00013 0.00090 USA 0.00027 0.00006 0.00010 0.00015 0.00045 0.00067 0.00001 0.00011 KOREA ARGENTINA FRANCE UK GERMANY HONG KONG INDIA JAPAN 0.00000 0.00050 0.00007 0.00214 0.00002 0.00187 0.00000 0.00067 0.00005 0.00082 0.00098 0.00002 0.00000 0.00022 0.00062 0.00179 0.00047 0.00378 0.00000 0.00000 0.00038 0.00000 0.00034 0.00214 0.00003 0.00051 0.00004 0.00004 0.01415 0.00852 0.00004 0.00188 0.00048 0.00239 0.00000 0.00003 0.00000 0.00601 0.00032 0.00065 0.00093 0.00002 0.00041 CHINA 0.00000 ARGENTINA 0.00022 CHINA 0.00004 FRANCE 0.00000 0.00000 UK GERMANY 0.00068 0.00014 0.00004 0.00015 0.00034 0.00502 0.00354 HONG KONG 0.00820 0.00001 0.00000 0.00009 0.00058 INDIA 0.00079 0.00000 0.00000 0.00191 0.00176 0.00000 JAPAN 0.00242 0.00001 0.00025 0.01490 0.00499 0.00001 0.00001 SINGAPORE 0.00001 0.00021 0.00002 0.00019 0.00547 0.00000 0.00009 0.00012 USA 0.00292 0.00055 0.00089 0.00152 0.00454 0.00736 0.00007 0.00123 MALAYSIA ARGENTINA FRANCE UK CHINA 0.00000 0.00000 0.00032 ARGENTINA 0.00006 CHINA 0.00004 FRANCE 0.00000 0.00000 UK 0.00063 0.00004 GERMANY HONG KONG 0.00053 USA 0.00000 0.00063 0.00044 0.00221 0.00000 0.00000 0.00037 0.00000 0.00035 0.00002 0.00004 0.01272 0.00003 0.00046 0.00052 0.00005 0.00754 0.00158 0.00225 0.00000 0.00001 0.00000 0.00272 0.00051 0.00097 0.00147 0.00002 0.00041 0.00047 0.00379 0.00202 GERMANY 0.00015 0.00016 0.00474 0.00329 HONG KONG 0.00370 0.00000 0.00000 0.00004 0.00030 INDIA 0.00136 0.00000 0.00000 0.00290 0.00261 0.00000 JAPAN 0.00230 0.00001 0.00025 0.01421 0.00478 0.00001 0.00001 SINGAPORE 0.00004 0.00197 0.00067 0.00044 0.00007 0.00072 0.00060 0.00109 0.01733 0.00318 0.00000 0.00443 0.00029 0.00006 USA 0.00095 SINGAPORE 0.00094 0.00035 USA 0.00116 0.00025 0.00135 0.00003 JAPAN SINGAPORE 0.00001 0.00000 0.00005 0.00114 INDIA 0.00038 0.00002 0.00037 0.00080 0.00300 0.00039 The value in each cell indicates the additional export value in dollars to the row-country (in italics) due to an extra one-dol ar increase in exports to the column-country. For example, an additional dollar of Egyptian exports to India, will generate an increase of 0.00 129 dollars to Hong Kong and 0.00039 dollars to the UK. I 35 Table 5: Total Exports informationspilloversby market' Exporterl Egypt Tunisia Korea ARGENTINA Total return of 1 extra dollar to: 0.1215 Share of top 4 receivers from: 69.74% AUSTRALIA 0 0247 66.38% 0.0199 75.9% AUSTRIA 0.0461 95.16% 0.0558 94.3% BANGLADESH 0.0431 90.51% 0.0222 BELUX 0.0284 91.62% 0.0643 BOLIVIA 0.0242 85.37% 0D0156 BRAZIL 0.0453 76.38% 0.0612 CANADA 0.0202 93.78% CHILE 0.0885 CHINA COLOMBIA Market Total dollar to: 0.0768 1 Share receivers from 64.6% return of dollar to: 0.9898 Malaysia of top 4 receivers from 77.1% Total return of 1 dollar to: 0.6161 0.1950 76.1% 0.2292 75.2% 0.0610 90.2% 0.0606 90.0% 93.8% 0.1521 87.2% 0.2430 87.9% 96.0% 0 0661 90.0% 0.0623 88.8% 77.1% 0.2541 8.4% 0.1396 89.6% 8.6% 01976 70.3% 0.1470 77.5% 0.0128 90.4% 00834 93.8% 0.0646 922% 78.96% 0.0485 72.4% 0 7363 610% 0.4467 82.4% 0 0045 71.29% 0.0021 6S.7% 0.0352 87.4% 0.0378 84.3% 0.1323 61.14% 0.0757 629% 1.2123 60.0% 0.6070 58.2% COSTA RICA 0.0343 78.70% 0.0172 72.3% 0.2888 80.2% 01570 79.6% DENMARK 0.0348 88.57% 0.0409 87.0% 0.0939 88.1% 0.0937 87 4% ECUADOR 0.0179 78.42% 0.0137 79.5% 0.1544 83 0% 0 0733 78.0% 0.1882 95.5% 0.6677 96.7% 0.7849 96.9% 59.5% 0.1 27 174 27.7% 94.4% 0.1919 96.4% 0.1914 96.5% 59.4% 0.1848 47 7% 0.1786 46.0% 53.0% 0.1650 45.5% 01817 47.7% 44.1% 0.1739 333% 0.1734 34.1% 57.7% EGYPT 1 | Share of top 4 receivers from 79.0% SPAIN C00360 FINLAND 0.0681 94.83% - 0.0789 FRANCE 0.0789 51.53% 0.0939 UK 0.0479 46.32% 0.0513 GERMANY 0.0629 4210% 0.0772 GREECE 0.0113 68.70% 0.0163 629% 0.0263 568% 0.0273 GUATEMALA 0.0228 70.70% 0.0128 67.2% 0.2028 74.4% 0.1138 70.9% HONG KONG 0.1159 55.50% 0.0548 58.6% 0.7307 55 8% 1.1317 58 6% HONDURAS 0.0128 66.64% 0.0077 65.4% 0.1222 73.0% 0.0663 71.4% INDONESIA 0.0026 83.86% 0.0012 72.4% 0.0181 89.4% 0.0242 91.9% INDIA 0.1836 88.39% 0.0680 88.5% 0.4561 84.2% 7 0.6513 84.5% IRELAND 0 0237 98.20% 0.0310 98.7% 00667 98.1% 0.0610 98.0% ISRAEL 0.0036 72.28% 0.0068 78.9% 0.0129 65.3% 0.0122 65.7% ITALY 0.0354 48.58% 0.0553 66.3% 0.0928 39 .5% 0 0870 38.5% JAPAN 0.0355 62.95% 0.0180 59.9% 0.1653 63.6% 0.3434 66.4% KOREA 0.0041 88.63% 0.0016 82.2% 0.0463 96.1% KUWAIT 0.0522 82.77% 0.0S27 88.0% 0.2582 79.7% 0.3391 80.5% SRI LANKA 0.0442 88.30% 0 0209 90.3% 0.1535 84.2% 0.2382 79 5% MOROCCO 0.0191 8765% 0.0676 96.9% 0.0656 90.0% 0.0668 89.5% MEXICO 0.0331 48.25% 0.0237 46.4% 0.3150 540% 0.1579 54.6% MALAYSIA 0.0253 93.86% 0.0112 93.0% 0.1850 40. 60% ! 1 0 0444 . 1 4 1 , T 95.5% NICARAGUA 0.0174 81.95% 0.0090 78.7% 0.1619 0517% 0.1073 86.7% NETHERLANDS 0.0305 50.92% 0.0347 55 8% 0.0650 39.9% 0.0S84 39.3% NORWAY 0.0295 92.3% 0.0338 92.0% 0.0794 929% 00792 92.5% NEWZEALAND 0.0204 92.77% 0.0182 96.5% 0.1716 94.1% 0.2007 94.2% OMAN 0.0501 89.83% 0.0695 94.4% 0.2678 91.5% 0.3428 90.0% PAKISTAN 0.1038 93 70% 0 0327 90.8% 0.1787 79.0°/ 0.2598 PANAMA 0.0057 66.82% 0.0038 69.2% 0.0476 667% 0.0262 73.0% 0.2528 80.8% 0.1310 76.5% 887.8% 0.1069 93.3% 0.1521 94.0% 86.1% 01090 93.2% 0.0748 90.2% 864% 01768 951% 0.1066 93.6% 88.5% 071489 76.3% 01958 77.9% 75.2% | 67 6% |PERU 0.0286 72.13% 0.0209 PHILIPPINES 00172 91.13% 0.0061 F PORTUGAL 0.0238 86.74% 0.0259 - PARAGUAY 0.0186 86.98% 0.0135 SAUDI ARABIA 0.0303 82.52% 0.0384 SINGAPORE 0.0476 71.45% 0.0183 687% 0.3150 77.0% 0.2251 69.6% |SLOVENIA 0.0140 71.80% 0.0084 64.5% 0.1470 79.3% 0.0793 80.0% |SWEDEN 0.0516 91.71% 90.4% 0.138 93.3| 0.0540 70.99% 0.0691 78.5% 0.0975 THAILAND 0.0163 63.01% 0.0068 TRIN&TOB. 61.9% 0.0699 0.0027 93.90% 0.0018 TUNISIA 00119 TURKEY 0.0060 85.93% 0.0066 TAIWAN 0.0091 89.64% 0.0037 URUGUAY 0.0532 91.65% 00209 0.0207 SWITZERLAND USA VENEZUELA - | | 0.0591 T | | 91.% 91.03% , , 93.5% 0.1028 | 54.9% 533% 0.1003 . 59.8% .e135 905% 0.0100 V 88.7% 0.0509 93.8% 0.0496 93.4% | 80.8% 00099 78.1% 00101 ! 77.7% | 66.1% 0.0879 955% 0.1110 | 960% 0.0371 88.4% 93.6% 0.3005 3046% | 00175 044747 33.9% 0139 321% 0.0997 7554% 0.0177 76.0% 01533 75.1% 0.0798 94.6% 1 25.2% 73.6% 'For each exporting countrv, the first column gives the value of additional exports to the rest of theworlddue to a one-dollar increase in exports to each market. The second column gives the share of the4 largest receivers of information spillovers from each of thesemarkets. In the case of Egypt for example, an extra dollar exported to Argentina provides 0.12 15 dollars of additional exports to the rest-of-the-world and 68.74 percent of these additional exports are concentrated in the four largest receivens of Argentina's information spillovers. - 36 i Policy Research Working Paper Series Title Author WPS2455The Effectson Growth of Commodity Jan Dehn Price Uncertaintyand Shocks Date Contact for paper September2000 P. Varangis 33852 WPS2456Geographyand Development J. VernonHenderson ZmarakShalizi AnthonyJ. Venables September2000 R. Yazigi 37176 WPS2457 Urbanand RegionalDynamicsin Poland UweDeichmann VernonHenderson September2000 R. Yazigi 37176 WPS2458 ChoosingRuralRoadInvestments to Help ReducePoverty Dominiquevan de Walle October2000 H. Sladovich 37698 October2000 P. Sader 33902 WPS2459 Short-LivedShockswith Long-Lived MichaelLokshin MartinRavallion Impacts?HouseholdIncome Economy in a Transition Dynamics WPS2460LaborRedundancy,Retraining,and Outplacementduring Privatization: The Experienceof Brazil'sFederal Railway AntonioEstache Jose AntonioSchmitt de Azevedo EvelynSydenstricker October2000 G. Chenet-Smith 36370 WPS2461VerticalPriceControland Parallel Imports:Theoryand Evidence KeithE. Maskus YongminChen October2000 L. Tabada 36896 WPS2462 ForeignEntry in Turkey'sBanking Sector, 1980-97 CevdetDenizer October2000 I. Partola 35759 WPS2463 PersonalPensionPlansand Stock MarketVolatility MaxAlier DimitriVittas October2000 A. Yaptenco 31823 WPS2464The Decumulation(Payout)Phaseof EstelleJames DimitriVittas DefinedContributionPillars:Policy Issuesin the Provisionof Annuities and OtherBenefits October2000 A. Yaptenco 31823 WPS2465 ReformingTax ExpenditurePrograms CarlosB. Cavalcanti ZhichengLi in Poland October2000 A. Correa 38949 WPS2466El Nino or El Peso?Crisis,Poverty, And IncomeDistributionin the Philippines GauravDatt Hans Hoogeveen October2000 T. Mailei 87347 WPS2467Does FinancialLiberalizationRelax FinancingConstraintson Firms? Luc Laeven October2000 R. Vo 33722 Policy Research Working Paper Series Title Author Date Contact for paper IanWalker WPS2468Pricing,Subsidies,and the Poor: Demandfor ImprovedWater Services FidelOrdohez PedroSerrano in CentralAmerica JonathanHalpern November2000 S. Delgado 37840 WPS2469RiskShiftingand Long-Term Contracts:Evidencefrom the RasGas Project MansoorDailami RobertHauswald November2000 W. Nedrow 31585 WPS2470Are LargerCountriesReallyMore Corrupt? StephenKnack OmarAzfar November2000 P. Sintim-Aboagye 38526 WPS2471ValidatingOperationalFood InsecurityIndicatorsagainsta DynamicBenchmark:Evidence fromMali Luc J. Christiaensen RichardN. Boisvert John Hoddinott November2000 L. Christiaensen 81463 WPS2472Uzbekistanand Kazakkhstan: A Tale of TwoTransitionPaths AsadAlam Arup Banerji November2000 L. Henson 84026 WPS2473BankingRisksaroundthe World: The ImplicitSafetyNet Subsidy Approach Luc Laeven November2000 R. Vo 33722