Maritime Connectivity and Trade - Policy Issues in International Trade and Commodities, Study Series No. 70

AT MARITIME CONNECTIVITY AND TRADE POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES RESEARCH STUDY SERIES . 70 York Geneva, 2015 AT ii POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES purpose studies Research Study Series analyse policy issues stimulate discussions area international trade development. Series includes studies UNCTAD staff distinguished researchers organizations academia. opinions expressed research study authors official views UNCTAD secretariat member States. studies published Research Study Series read anonymously referee. Comments referees account publication studies. designations employed presentation material imply expression opinion part United Nations legal status country, territory, city area, authorities delimitation frontiers boundaries. Comments paper invited addressed author, / Publications Assistant, Trade Analysis Branch (TAB), Division International Trade Goods Services, Commodities (DITC), United Nations Conference Trade Development (UNCTAD), Palais des Nations, CH-1211 Geneva 10, Switzerland; -mail: tab@unctad.org; fax : +41 22 917 0044. Copies studies Research Study Series obtained address. Studies Research Study Series UNCTAD website http://unctad.org/tab. Series Editor: Victor Ognivtsev Chief Trade Analysis Branch DITC/UNCTAD UNCTAD/ITCD/TAB/72 UNITED NATIONS PUBLICATION ISSN 1607-8291 © Copyright United Nations 2015 rights reserved Maritime Connectivity Trade iii Connectivity crucial determinant bilateral exports. paper presents empirical assessment relationship maritime connectivity exports containerizable goods period 2006-2012. Based unique dataset empirical investigations unequivocally show lacking direct maritime connection trade partner values exports. Estimates point range varying 42 cent 55 cent. assessing effect number transhipments connect country pairs, additional transhipment 20 25 cent exports. Results suggest absence bilateral connectivity indicator impact bilateral distance bilateral exports -estimated. Keywords: Maritime Transport, Sea Distance, Containerizable Trade, Trade Costs JEL Classification: C61, F1, L91 iv POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES paper part multi-year research project maritime connectivity consequences bilateral trade economic development, merging UNCTAD' DITC UNCTAD' DTL analytical expertises. findings reflect views author exclusively, represent views UNCTAD member countries. author gratefully acknowledges comments earlier draft Jan Hoffman Guillermo Valles Victor Ognivtsev support guidance. mistakes errors remain authors' . Maritime Connectivity Trade Executive Summary ................................................................................................................................ vi 1 INTRODUCTION .......................................................................................................................... 1 2 DATA ........................................................................................................................................... 2 3 EMPIRICAL STRATEGY ............................................................................................................. 6 4 RESULTS ..................................................................................................................................... 7 4.1. Core specifications ............................................................................................................ 7 4.2. Robustness checks.......................................................................................................... 11 5 CONCLUDING REMARKS ........................................................................................................ 18 REFERENCES ......................................................................................................................................... 20 List figures Figure 1. Direct sea distance maritime (estimated) distance transhipments ......................... 5 Figure 2. Maritime distance (estimated) number transhipments (country averages) ................ 5 List tables Table 1. Top bottom fifteen countries: average number transhipments .................................. 3 Table 2. Benchmark regressions (Cross section) ................................................................................ 9 Table 3. Benchmark regressions (Panel) ........................................................................................... 10 Table 4. Extended fixed effects ......................................................................................................... 12 Table 5. Poisson regressions ............................................................................................................ 13 Table 6. Negative binomial regressions ............................................................................................ 16 Table 7. -inflated negative binomial regressions ....................................................................... 17 vi POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES Access foreign markets critical determinant export performance. technical literature foreign market access representing foreign market potential country relates inter alia inversely bilateral transport costs. existence direct maritime connection recognized play important role determining trade costs. , theoretical empirical attention devoted impact bilateral exports. lack comprehensive evidence relationship maritime connections bilateral exports due lack data large extent. objective paper fill gap information maritime connections sample 178 countries collected 2006-2012 period. basic statistical analysis reveals period average 14 cent country pairs connected , 11 cent single transhipment, 36 cent transhipments 28 cent transhipments. 61 cent country pairs connected transhipments 90 cent transhipments. paper assessment impact nature maritime connections bilateral exports containerizable goods comprehensive set country pairs observed years. causal relationship remains difficult identify, estimates suggest absence direct connection drop exports varying 42 55 cent depending underlying empirical specification. Results additional transhipment drop exports varying 20 25 cent. find evidence relationship bilateral exports number transhipments transport containerizable goods countries -linear. results suggest quality maritime connectivity preponderant determinant foreign market access. definition landlocked enjoy direct connection trade partner contiguous countries. results provide estimate handicap terms export landlocked countries face top impact quality transit transports. High transport costs continue constitute greatest impediment LDCs’ trade competitiveness, equitable access global markets. improvement quality maritime connectivity core strategy aiming stimulating exports promoting participation domestic economy global chains production. improvement contribute reduction transport costs. Intervening efficiently maritime connectivity easy task. options exist respective desirability reflect country specific characteristics. , investing infrastructures vital options. require financial effort countries bear . International cooperation partnerships crucial. International cooperation partnerships form establishing strategies aiming creating incentives shipping companies serve destinations necessarily profitable place. instance, consist granting companies serving "remote" countries preferential access major maritime hubs world. Maritime Connectivity Trade 1 1. INTRODUCTION Maritime transport core international trade merchandises. 80 cent volume goods exchanged world transported sea (UNCTAD, 2008). predominance maritime transport explained large extent exponential intensification containerized transport services. containerization global liner shipping network, small large exporters importers finished intermediate containerizable goods countries trade -, individual trade transaction economically justify chartering ship transport containers . regular container shipping services transhipment operations -called hub ports, basically countries today connected . empirical study confirmed “… effects Container Revolution World Trade” (Bernhofen al., 2013). North-North trade concerned authors cumulative (concurrent lag effects) average treatment effect containerization 20 year time period amount 790 cent. cumulative effect bilateral GATT membership raise trade average 285 cent, cumulative effect full containerization. growing participation developing countries seaborne trade,1 evidence maritime connections suggests , China, reached full potential. Fugazza al. (2013) dataset find average number direct maritime connections, meaning involving transhipment transported goods country origin destination, developing countries developed . literature emphasized importance transport costs infrastructure explaining trade access international markets. empirical strategies produce estimates level transport costs eventually impact exchange goods. studies ratio imports CIF imports FOB proxy transportation costs, -called cif/fob ratio (.. Baier Bergstrand 2001, Hummels Lugovskyy 2006). Estimates vary essentially level product aggregation. reasonable average estimate ratio computed based total imports CIF FOB country level ranges 6 cent 12 cent. disaggregated product levels dispersion increases. Approximations CIF/FOB ratios higher developing developed regions. UNCTAD estimates decade, freight costs amounted 6.4 cent developed countries’ imports compared 10.6 cent Africa (UNCTAD, 2011). Based estimation gravity model US data, Anderson Van Wincoop (2003) transport costs correspond average ad valorem tax equivalent 21 cent. 21 cent include measured freight costs 9 cent tax equivalent time goods transit. similar empirical approach, Clark al. (2004) estimates reveal Latin American countries, transport costs greater barrier .. markets import tariffs. find ports efficiency important determinant shipping costs. Arvis al. (2013) work extension Jacks al. (2011) contribution. , represents comprehensive country-level analysis trade costs components date. database includes 178 countries covers 1995-2010 period. Estimates trade costs inferred observed pattern production trade countries. Results maritime transport connectivity logistics performance important determinants bilateral trade costs: UNCTAD’ Liner Shipping Connectivity Index (LSCI) World Bank’ Logistics Performance Index (LPI)2 important source variation trade costs geographical distance, effect strong trade relations involving South. 1 1970 2010, developing countries´ share volume seaborne exports rose 18 cent 56 cent world´ total (UNCTAD, 2013). 2 World Bank' Logistics Performance Index (LPI) UNCTAD' Liner Shipping Connectivity Index (LSCI) aim ways provide information countries' trade competitiveness area transport logistics. 2 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES existence direct maritime connection recognized play important role determining trade costs. , theoretical empirical attention devoted impact bilateral exports. nature maritime connections treated component aggregate trade costs function standard theoretical models impact tariff-. empirical evidence limited remains piecemeal. instance, Wilmsmeier Hoffmann (2008) findings based sample 189 freight rates company Caribbean show trade routes indirect services (.. including transhipments) induce higher transport costs. estimates suggest transhipment equivalent impact freight rates increase distance countries 2,612 km. lack comprehensive evidence relationship maritime connections bilateral exports due lack data large extent. objective paper fill gap information maritime connections large sample countries collected 2006-2012 period. attempt assess impact nature maritime connections pair countries bilateral exports containerizable goods. causal relationship remains difficult identify, estimates suggest absence direct connection drop exports 55 cent reference specification. Results additional transhipment drop exports 25 cent. find evidence relationship bilateral exports number transhipments transport containerizable goods countries -linear. rest paper organized . section presents data reviews descriptive statistics. Section 3 discusses empirical strategy implemented retrieve impact maritime connectivity containerizable trade. Results shown section 4. section discusses policy implications concludes. 2. DATA present main characteristics data descriptive statistics helped identifying empirical strategy. database Empirical investigation based unique dataset Fugazza al. (2013). shortest maritime liner shipping routes pair countries reported reference sample 178 countries 6 years 2006-2012 period. Information year 2007 missing . Computed maritime liner service distances retrieved original database existing direct liner service connections pairs countries sea distance countries’ respective main container ports.3 connection qualified “direct” implies transhipment country. , ship call ports en route. information existence direct connection retrieved UNCTAD' Liner Shipping Connectivity Matrix (LSCM). information contained database obtained annually, month , Lloyds List Intelligence. data covers reported deployment containerships point time. methodology comparisons time, “sample” complete. , scope activities countries covered, measurement approach, . spite differences, indexes statistically positively correlated, partial correlation coefficient +0.71. Information UNCTAD' LSCI UNCTAD' Review Maritime Transport. detailed description data World Bank, LPI website http://www.worldbank.org/lpi. Sea distance pairs countries represents distance separating coastal country’ main port(). cases large countries coast lines (.. USA, Canada al) main port retained varies trade partner considered. Maritime Connectivity Trade 3 Shortest routes obtained solving shortest path problem frame graph mathematical theory applying modified version Dijkstra (1959) algorithm. identification shortest route, direct connections privileged. case options moving point point , option smallest number transhipments chosen maritime distance travelled longest options. , limit number transhipments ( case origin destination countries landlocked). Larger figures obtained reflect realistic route choices logistical point view. variables retained previous computations: dummy variable assumes 1 direct service countries exists 0 , variable indicating number transhipments connect pair countries , effective maritime distance covered pair countries. Note case direct maritime connection, effective (computed) maritime distance coincides sea distance mentioned . Note information number transhipments connect pair countries symmetric: transhipments move containers country country , number transhipments move containers direction . 178 countries sample 33 landlocked. landlocked countries definition direct access liner shipping services, trade overseas trading partners, making neighbouring countries seaports. Land-locked countries assigned maritime distance corresponds sum distances: distance container port transit country largest share overseas trade passes , computed maritime distance transit country reachable destination. implicitly assume transportation road transportation sea comparable effects transport costs. implicitly assume transhipment road/rail sea impact transport costs transhipment sea sea. assumptions . , road transport costs vary country country cost related transhipment operations. consequence adopt econometric strategy, precisely , accounts explicitly country specific characteristics minimize bias related assumptions. Table 1 Top bottom fifteen countries: average number transhipments Top 15 Bottom 15 GBR 0.73 RWA 3.15 FRA 0.79 MWI 3.15 BEL 0.84 ZMB 3.15 DEU 0.87 BOL 3.16 NLD 0.88 ISL 3.16 ITA 0.92 TKM 3.20 ESP 0.93 NER 3.20 CHN, HKG SAR 0.95 BLZ 3.23 CHN 0.97 SVK 3.31 USA 0.98 HUN 3.31 KOR 1.07 BLR 3.32 MYS 1.11 NRU 3.42 SGP 1.13 MLI 3.53 CHN, TWN Prov. 1.19 MDA 3.62 JPN 1.29 ARM 4.10 4 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES descriptive statistics Straightforward computations provide interesting insights structure global liner shipping network. period average 14 cent country pairs connected , 11 cent single transhipment, 36 cent transhipments 28 cent transhipments. 61 cent country pairs connected transhipments 90 cent transhipments. maximum number transhipments obtained 6. unrealistic circumstances. , 10 cent observations related transhipments involve landlocked landlocked countries. alternative resolution algorithm includes constraint limiting maximum number transhipments 3 (+1 landlocked countries). alternative variable maritime distance robustness check variables. note maritime distances strongly correlated show coefficient pairwise correlation close 0.98. average number connections country level period time reported table 1 (left quadrant) observe Great Britain country smallest average number transhipments, France, Belgium, Germany EU countries. ranking result strong intra-EU trade effect. trade relationships EU members included European countries stay top ten country list. countries top fifteen USA East Asian countries. clear intra-regional effect group countries. quadrant table 1 bottom fifteen countries. geographical composition heterogeneous continents represented. bottom list landlocked countries small island states. direct sea-distance shortest connection distance transhipment expected strongly correlated. maritime distance transhipment(), , increase respect sea distance increases. shown figure 1. Points 45 degree line represent direct connections computed effective distance (including transhipments) sea distance definition coincide. , question maritime distances transhipment number transhipments correlated obvious answer. linear quadratic fit lines reported figure 2 suggest measures weakly correlated. direct connections excluded, sample distance measures remain weakly correlated. findings highlight important features accounted empirical analysis determinants containerized trade. Standard measures distance (sea -crow-fly distances) underestimated case -direct connections, meaning 85 cent observations. direct consequence, impact distance bilateral exports expected -estimated standard measures adopted. existence direct connection number transhipments reach export destination expected affect trade necessarily independently impact effective maritime distance. terms, distance fully reflect incidence transport costs considered number transhipments assessing impact transport costs bilateral exchanges. percentages slightly earlier analysis (UNCTAD, 2013) paper data base includes land-locked countries, connected global shipping network neighbouring transit countries. Maritime Connectivity Trade 5 Figure 1 Direct sea distance maritime (estimated) distance transhipments Note: red line represents linear fit relationship green line quadratic fit. dashed line 45 cent line. Figure 2 Maritime distance (estimated) number transhipments (country averages) 10 15 20 25 10 15 20 25 1 2 3 4 1 2 3 4 2006 2008 2010 2012 Obs. Fitted values Fitted values ar iti ta nc ( st im ed ) Number Transhipments Graphs Year Note: red line represents linear fit relationship green line quadratic fit. 5000 10000 15000 20000 5000 10000 15000 20000 Maritime distance (estimated) ire ct ea di st ce 6 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES 3. EMPIRICAL STRATEGY identification direct connections definition number transhipments connect country pair computation effective maritime distance sample 178 countries 6 years refining assessment impact trade costs bilateral trade determinants. clear causal relationship difficult identify influence endogeneity issues related reverse causality/simultaneity omitted variables fully excluded. Endogeneity reference sample direct connection countries factor higher potential trade. , existence direct connection countries exist demand goods produced country large . issue reverse causation work underlying intuition obvious. , specialized contributions show establishment/interruption direct connection process circumstances. elements explain existence direct transport services countries independently current relative demand conditions. obvious historically high levels trade, sufficiently good port infrastructures geographical location respect major maritime routes regional maritime hubs. surprisingly, expected important determinants current trade values included empirical model. element reflected current demand conditions time invariant characteristics inherent trade partner countries. elements varying gradually time controlled country time specific effects. inclusion variables reflect influence elements avoid minimize omitted variables biases. , treatment reverse causation/simultaneity remain partial context. empirical strategies implemented order treat specifically simultaneity. Instrumental variable estimation efficient . good instruments existence direct connection easily identifiable. strongly related dependent variable. Due small number trade relationships change connectivity status observed 2006-2012 period strategy based differentiation wake Baier Bergstrand (2007) instance conceivable. Identification based 9 cent worst 3 cent observations constituting sample benchmark specifications. limited variability prevents contemplating difference difference type approaches potentially treat simultaneity explicitly. strategies consist comparative sample trade relationships affected direct connections countries. , group represents 6 cent reference sample annihilates reasonable room precise identification. major element explaining restricted variability connectivity observed reference sample short period time covered. data collection process exclusively oriented remain historical series exist maintained UNCTAD. Keeping constraints mind, core empirical strategy adopted paper set robustness checks implemented explore feasible directions incidence endogeneity bias estimates. 5 Endogeneity plagues estimates obtained version gravity model. stated Bladwin Taglioni (2006) "… gravity model model usual sense – regression endogenous variables endogenous variables…". 6 inter alia Notteboom (2004), Wilmsmeier Notteboom (2009) Ducruet Notteboom (2012). Maritime Connectivity Trade 7 empirical model focus maritime connectivity, dependent variable trade goods highly containerizable natural log depending estimated specification. mentioned previously volume trade transported sea represents 80 cent world trade volume. study reference unit volume . average trade transported sea increasing steadily 50 cent total trade 2006 54 cent 2012. Pairwise correlation series total exports exports highly containerizable 0.93 highly significant. empirical model adopt standard gravity model international trade augmented inclusion maritime connection variables. benchmark specification , ( ) ( ) ijttjtitktjtjkt jktjkjkjkjkt IIIGDPGDPnttranshipmeRTA distmaritimeColonyLanguageBorderX εααααααα ααααα ++++++++ ++++= 111098765 43210 )ln()ln( _lnln dependent variable ln() natural logarithm total exports country country recorded year . Explanatory variables include standard gravity variables. dummies existence common border (Border), common language (Language) dummy indicating trade partner colony source country (Colony). distance variable maritime_dist represents effective maritime distance separating country country computed previously. include RTA dummy indicating trade relationship preferential. dummy expected account impact preferential trade agreements transport costs. natural logarithm GDPs origin destination countries included account aggregate demand supply conditions. versions transhipment variable. version dummy variable assumes 1 direct service exists , 0 . version reports number transhipments connect pair countries computed applying Dijkstra’ algorithm frame graph mathematical theory. case estimate specification includes square number transhipments variable. reflect possibility observing decreasing marginal transhipment costs. Exporter importer specific time invariant effects controlled inclusion exporter importer fixed effects. inclusion sets fixed effects minimize incidence possibly omitted variables expected time invariant country specific. instance importantly, position country respect major maritime routes accounted sets country fixed effects. Global year specific shocks absorbed year fixed effects. 4. RESULTS 4.1. CORE SPECIFICATIONS start sample cross section. average values reference period continuous variables. GDP variables included control exporter importer fixed effects. RTA dummy variable transhipment variable binary version, set reference period. Note choosing rule terms allocation affect significantly results. Results reported table 2. Column (1) shows results obtained standard specification distance variable corresponds direct sea distance major ports. Column (2) include transhipment variable based calculated effective maritime distance direct sea distance. Column (3) refers specification including binary version transhipment variable. Column (4) reports estimates number transhipments indicator maritime connectivity. Column (5) includes square number transhipments. columns coefficients significantly affected 8 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES choice bilateral distance regressions. instance, impact bilateral distance significantly larger column (1) RTA variable coefficient significantly smaller. Coefficients bilateral distance variable, affected inclusion maritime connectivity variable column (3) (5). direct sea distance measure effective measure transhipment variable, impact bilateral distance appears overestimated. cent increase distance 1.3 cent decrease bilateral exports direct sea measure bilateral distance compared 1.1 decrease calculated effective measure maritime distance . figure drops -0.92 -0.88 cent binary transhipment variable number transhipments variable . Estimated coefficients suggest absence direct connection 64 cent exports additional transhipment 31 cent exports. , impact bilateral exports transhipment appears decreasing obtain positive coefficient number transhipments squared variable. Table 3 reports results obtained original structure data. GDP variables reintroduced. qualitative point view patterns similar observed table 2 results. controlling extensive margin maritime connection number transhipments, impact maritime distance appears -estimated. -estimated direct sea distance. case elasticity bilateral exports respect effective maritime distance -1.16 -1.34 . controlling maritime connectivity, elasticity drops -1.01 including binary transhipment variable -0.98 including transhipments count variable. Differences distance (log ) coefficients significant. sensitivity RTA dummy coefficient remarkably affected measure effective maritime distance measure direct sea distance. estimated coefficient increases 0.7 0.88 variation statistically significant. controlling maritime connectivity, estimated coefficients slightly decrease 0.84. difference statistically significant. Results reported column (3) show absence direct connection fall bilateral exports 55 cent. assessing impact total number transhispments reach final destination, results additional transhipment translates fall exports 25 cent. find relationship bilateral exports number transhipments linear represented strictly convex function. table 2 table 3 point importance controlling extensive margin maritime transport. inclusion precise measure effective bilateral maritime distance variable qualifying nature bilateral maritime connection reduces significantly impact bilateral distance bilateral exports. trade costs properly accounted direct geographical distances. Maritime Connectivity Trade 9 Table 2 Benchmark regressions (Cross section) (1) (2) (3) (4) (5) LN(maritime_distance) -1.299a -1.100a -0.929a -0.880a -0.879 (0.0213) (0.0211) (0.0213) (0.0231) (0.0229) Language 0.961a 1.114a 1.071a 1.064a 1.062 (0.0393) (0.0399) (0.0391) (0.0394) (0.0391) Colony 0.768a 0.690a 0.554a 0.602a 0.559 (0.100) (0.106) (0.103) (0.101) (0.102) Border 0.420a 0.821a 0.594a 0.797a 0.656 (0.147) (0.152) (0.144) (0.141) (0.142) RTA 0.885a 1.110a 1.035a 1.064a 1.028 (0.0454) (0.0456) (0.0448) (0.0448) (0.0447) Transhipment -1.020a (0.0384) Transhipments_Number -0.309a -0.797 (0.0155) (0.0322) Transhipments_Number_2 0.134 (0.00832) Exporter FE Importer FE Observations 24760 24760 24760 24760 24760 R2 0.806 0.809 0.813 0.812 0.814 Adjusted R2 0.804 0.806 0.811 0.809 0.811 Note: Robust standard errors parentheses: < 0.10, < 0.05, < 0.01. 10 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES Table 3 Benchmark regressions (Panel) (1) (2) (3) (4) (5) LN(maritime_distance) -1.343a -1.157a -1.014a -0.981a -0.970a (0.0191) (0.0199) (0.0199) (0.0209) (0.0208) Language 0.960a 1.134a 1.088a 1.086a 1.078a (0.0374) (0.0383) (0.0377) (0.0379) (0.0377) Colony 0.781a 0.683a 0.598a 0.625a 0.602a (0.0896) (0.0950) (0.0929) (0.0913) (0.0919) Border 0.392a 0.814a 0.664a 0.795a 0.720a (0.128) (0.134) (0.129) (0.127) (0.127) LNGDPX 0.287a 0.324a 0.318a 0.307a 0.310a (0.0358) (0.0362) (0.0361) (0.0362) (0.0361) LNGDPM 0.795a 0.820a 0.810a 0.800a 0.803a (0.0352) (0.0355) (0.0353) (0.0354) (0.0353) RTA 0.698a 0.879a 0.834a 0.847a 0.830a (0.0363) (0.0374) (0.0368) (0.0368) (0.0367) Transhipment -0.796a (0.0319) Transhipments_Number -0.253a -0.573a (0.0117) (0.0243) 0.0879a Transhipments_Number_2 (0.00569) Exporter FE Importer FE Year FE Observations 125632 125632 125632 125632 125632 R2 0.779 0.771 0.775 0.774 0.775 Adjusted R2 0.779 0.770 0.774 0.773 0.774 Note: Robust standard errors parentheses (clustered country-pair). < 0.10, < 0.05, < 0.01. Maritime Connectivity Trade 11 4.2. ROBUSTNESS CHECKS robustness checks undertaken. mentioned, implement satisfactory statistically validated instrumentalization transhipment variables. opt set robustness checks limit influence endogenity. start inclusion larger set fixed effects. restrict sample trade relationships change transhipment conditions observed period. alternative maritime connectivity variable. Finally treat high incidence zeros matrix bilateral trade. Fixed effects consecutive years enlarge set fixed effects including exporter fixed effects importer fixed effects interacted time fixed effects. Results reported columns table 4. Columns (1) (3) direct counterpart columns (3), (4) (5) table 3. Coefficients variables present tables affected decimal. included country pair fixed effects additional check. , due variability transhipment variables, impact large extent absorbed set country-pair fixed effects. mentioned previously information 2007 missing. excluded year 2006 sample order consecutive years. Results, reported, remain practically unchanged respect benchmark specification. Steady connections order isolate estimates simultaneity bias, drop reference sample trade relationships change connection status (direct - direct) observed period investigation. number total transhipment connectivity related explanatory variable drop trade relationships number transhipments connect underlying pair countries changed. discussed, observations dropped represent 3 cent observations reference sample case. case, percentage 9 cent. Results reported columns table 4. Exporter importer fixed effects interacted time fixed effects considered ease comparison columns table reference light findings commented previous -section. Results line obtained sample. words, generating selection bias, endogeneity, results comparable obtained reference estimations. variance estimated coefficients samples suggest , discussed previously, simultaneity major issue. 12 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES Table 4 Extended fixed effects (1) (2) (3) (4) (5) (6) LN(maritime_distance) -1.016a -0.979a -0.971a -1.019 -0.972a -0.977a (0.0202) (0.0213) (0.0212) (0.0208) (0.0225) (0.0225) Language 1.085a 1.082a 1.075a 1.099 1.076a 1.069a (0.0380) (0.0381) (0.0380) (0.0392) (0.0404) (0.0403) Colony 0.603a 0.628a 0.607a 0.576 0.589a 0.573a (0.0936) (0.0919) (0.0926) (0.0957) (0.0935) (0.0943) Border 0.654a 0.783a 0.710a 0.619 0.693a 0.617a (0.130) (0.127) (0.128) (0.131) (0.127) (0.128) RTA 0.856a 0.868a 0.852a 0.823 0.833a 0.826a (0.0378) (0.0377) (0.0377) (0.0387) (0.0396) (0.0395) Transhipment -0.790a -0.823 (0.0322) (0.0345) Transhipments_Number -0.258a -0.571a -0.302 -0.633a (0.0121) (0.0247) (0.0144) (0.0303) Transhipments_Number2 0.0872a 0.104 (0.00590) (0.00844) Exporter FE*Year FE Importer FE*Year FE Observations 125632 125632 125632 121481 114017 114017 R2 0.779 0.778 0.779 0.779 0.784 0.785 Adjusted R2 0.775 0.774 0.775 0.775 0.780 0.781 Note: Robust standard errors parentheses (clustered country-pair). < 0.10, < 0.05, < 0.01. Maritime Connectivity Trade 13 Table 5 Poisson regressions (1) (2) (3) (4) (5) (6) LN(maritime_distance) -0.460a -0.450a -0.447a -0.499a -0.448a -0.500a (0.0128) (0.0126) (0.0126) (0.0130) (0.0126) (0.0129) Language 0.201a 0.237a 0.236a 0.219a 0.240a 0.222a (0.0367) (0.0364) (0.0364) (0.0358) (0.0364) (0.0358) Colony 0.279a 0.202a 0.201a 0.191a 0.195a 0.188a (0.0424) (0.0427) (0.0425) (0.0433) (0.0427) (0.0434) Border 0.433a 0.436a 0.444a 0.386a 0.440a 0.383a (0.0384) (0.0377) (0.0377) (0.0367) (0.0377) (0.0367) LNGDPX 0.353a 0.354a 0.352a 0.354a 0.354a 0.355a (0.0938) (0.0925) (0.0927) (0.0923) (0.0924) (0.0921) LNGDPM 0.621a 0.621a 0.620a 0.623a 0.621a 0.624a (0.0965) (0.0959) (0.0965) (0.0927) (0.0960) (0.0924) RTA 0.520a 0.443a 0.438a 0.404a 0.434a 0.402a (0.0325) (0.0325) (0.0324) (0.0316) (0.0324) (0.0316) Transhipment -0.610a (0.0272) Transhipments_Number -0.271a -0.264a -0.513a -0.482a (0.0113) (0.0113) (0.0316) (0.0417) Transhipments_Number2 0.0940a 0.0881a (0.00984) (0.0146) Exporter FE Importer FE Year FE Observations 201189 201189 201189 201189 201189 201189 pseudo R2 0.950 0.951 0.952 0.953 0.952 0.953 ll -8.87811e+09 -8.54299e+09 -8.52911e+09 -8.29563e+09 -8.49761e+09 -8.27951e+09 Note: Robust standard errors parentheses (clustered country pair). < 0.10, < 0.05, < 0.01. 14 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES Maximum number transhipments mentioned , maximum number transhipments allowed . mentioning landlocked countries Georgia Republic Central Africa, countries Denmark Germany computations 6 transhipments reach destination. logistic point view unrealistic, redefined maritime connectivity variable imposing stringent constraint terms maximum number transhipments solution algorithm. number transhipments larger 3 -landlocked countries larger landlocked . 11 cent total observations affected constraint. dramatic observed terms effective maritime distance. Econometric estimations shown table 5. report specification binary maritime connectivity variable estimated coefficients slightly modified compared shown column (3) table 3. surprising view strong correlation exists calculated maritime distances. Generally speaking coefficients affected, differences respect benchmark estimates dramatic. instance, additional transhipment translates fall exports 22 cent. expected upper bound number transhipment limited maritime distance modified slightly. combination features explains coefficient maritime distance increased absolute . , inclusion interactions exporter time fixed effects importer time fixed effects, reported columns (2), (4) (6), affect results. important remark relates significant change estimated coefficient RTA dummy. systematically smaller counterpart table 3. number transhipment limited, observe amplification maritime distance effect reduction existence RTA. necessarily surprising measures determinants transport costs act independently . Columns (5) (6) show cross-country estimates. Patterns obtained framework. Zeros prevalence zeros asymmetric flows matrix bilateral exports - documented stylized facts. database exception. cent potential export relationships active. 38 cent, 12 cent characterized export directions remaining 26 cent direction. Opting log- linearized empirical model fully compatible existence zeros trade data. imposed truncation sample ( , elimination -trade pairs) solution reference estimations. truncation bias significantly estimated coefficients. implement Poisson pseudo-maximum likelihood (PPML) estimator introduced Gourieroux Monfort Trognon (1984) shown Santos Silva Tenreyro (2006) deal appropriately presence exports observations heteroskedasticity. choice approach Helpman al. (2008) motivated fact consistent estimation structural parameters assumption random components model homoscedastic. clear contrast evidence. Results reported table 5. Columns (3) (5) report estimates obtained restricted number transhipments variable. Coefficients readily interpretable case log-linear model. Coefficients dummy variables, holding variables constant model, represent difference logs exports categories identified dummy. incidence rate ratio exponential coefficients. Estimates dramatically Santos Silva Tenreyro (2014) detailed discussion. Maritime Connectivity Trade 15 obtained log-linear model based intensive margin observations exclusively. instance, transhipment variable, column (2) estimates exports 46 cent direct connection. Results column (4) (5) show additional transhipment fall exports 23 23.5 cent depending maximum number transhipments allowed . Estimates RTA dummy including transhipment variable correspond 55 cent higher exports 41 cent obtained reference estimation. distribution dependent variable displaying signs overdispersion run additional sets regressions. assume bilateral exports follow negative binomial distribution. , negative binomial MLE predict zeros observed number zeros. , set regressions explicitly accounts incidence zeros overdispersion. follow Burger, van Oort Linders (2009) show problem overdispersion excessive zeros addressed appropriately inflated negative binomial MLE. Results reported table 6 table 7 . Results reveal differences estimated impacts pronounced observed Poisson regressions. case classical gravity variables. part trade agreement 240 cent higher bilateral exports. Differences great extent smaller case transhipment variables. Estimated coefficients absence direct connection 42 cent bilateral exports. additional transhipment 20 cent bilateral exports. -inflated negative binomial regression generates separate models combines . logit model generated cases, predicting export relationship active. set explanatory variables considered predicting model. set explanatory variables -- inflated negative binomial model estimates table 7. set exporter importer fixed effects dropped. set comparable inclusion additional country time specific variables. reflect geographical characteristics country. Results marginally affected set explanatory variables included. overdispersion parameter significantly 97 cent confidence level 5 specification. Poisson distribution . Vuong -statistic significantly positive, inflated variant Negative Binomial estimator preferable. 16 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES Table 6 Negative binomial regressions (1) (2) (3) (4) (5) (6) LN(maritime_distance) -1.374a -1.259a -1.196a -1.337a -1.179a -1.338a (0.0175) (0.0185) (0.0201) (0.0207) (0.0196) (0.0207) Language 1.320a 1.299a 1.301a 1.226a 1.291a 1.223a (0.0346) (0.0349) (0.0348) (0.0353) (0.0349) (0.0353) Colony 0.922a 0.878a 0.873a 0.916a 0.869a 0.915a (0.0704) (0.0709) (0.0698) (0.0706) (0.0703) (0.0707) Border 1.719a 1.603a 1.655a 1.470a 1.611a 1.448a (0.137) (0.142) (0.143) (0.147) (0.144) (0.147) LNGDPX 0.0417 0.0496 0.0492 0.0507 0.0441 0.0505 (0.0968) (0.0968) (0.0967) (0.0978) (0.0966) (0.0978) LNGDPM 0.679a 0.667a 0.637a 0.646a 0.639a 0.650a (0.0992) (0.0978) (0.0975) (0.0964) (0.0970) (0.0964) RTA 0.926a 0.891a 0.891a 0.757a 0.872a 0.754a (0.0321) (0.0317) (0.0317) (0.0314) (0.0317) (0.0314) Transhipment -0.545a (0.0287) Transhipments_Number -0.195a -0.212a -0.409a -0.337a (0.0138) (0.0130) (0.0256) (0.0321) Transhipments_Number2 0.0497a 0.0390a (0.00678) (0.0105) Exporter FE Importer FE Year FE 201189 201189 201189 201189 201189 201189 pseudo R2 0.086 0.086 0.086 0.087 0.086 0.087 ll -1275803.1 -1275443.7 -1275407.3 -1274375.8 -1275268.4 -1274353.1 Note: Robust standard errors parentheses (clustered country pair). < 0.10, < 0.05, < 0.01. Maritime Connectivity Trade 17 Table 7 -inflated negative binomial regressions (1) (2) (3) (4) (5) (6) LN(maritime_distance) -1.004a -0.879a -0.847a -0.975a -0.837a -0.970a (0.0131) (0.0135) (0.0141) (0.0150) (0.0140) (0.0150) Language 1.148a 1.116a 1.121a 1.044a 1.104a 1.036a (0.0276) (0.0278) (0.0277) (0.0279) (0.0278) (0.0280) Colony 0.805a 0.761a 0.760a 0.797a 0.761a 0.801a (0.0522) (0.0526) (0.0520) (0.0523) (0.0523) (0.0525) Border 1.577a 1.442a 1.490a 1.355a 1.446a 1.322a (0.0785) (0.0805) (0.0802) (0.0817) (0.0805) (0.0818) LNGDPX 0.125 0.138c 0.129 0.138c 0.131 0.135c (0.0820) (0.0816) (0.0818) (0.0827) (0.0815) (0.0823) LNGDPM 0.663a 0.662a 0.634a 0.647a 0.646a 0.653a (0.0720) (0.0706) (0.0709) (0.0704) (0.0703) (0.0706) RTA 0.922a 0.877a 0.885a 0.770a 0.865a 0.794a (0.0242) (0.0238) (0.0240) (0.0239) (0.0239) (0.0238) Transhipment -0.606a (0.0206) Transhipments_Number -0.197a -0.209a -0.469a -0.434a (0.00965) (0.00930) (0.0196) (0.0244) Transhipments_Number2 0.0734a 0.0747a (0.00587) (0.00832) Exporter FE Importer FE Year FE 201189 201189 201189 201189 201189 201189 ll -1248838.2 -1247483.0 -1248168.6 -1247012.7 -1247678.2 -1247305.9 Note: Robust standard errors parentheses (clustered country pair). < 0.10, < 0.05, < 0.01. 18 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES 5. CONCLUDING REMARKS importance trade costs drivers exchanges merchandise goods countries geographical pattern economic activity contributions understanding remain piecemeal. paper assessment impact nature maritime connections bilateral exports containerizable goods comprehensive set country pairs observed years. variables, gravity literature, qualify nature maritime connections: variable indicating existence direct maritime connection (.. existence operational shipping service) countries, variable resulting optimization algorithm indicating number transhipments connect countries, including landlocked . estimates suggest absence direct connection drop exports varying 42 55 cent depending underlying empirical specification. Results additional transhipment drop exports varying 20 25 cent. find evidence relationship bilateral exports number transhipments transport containerizable goods countries -linear. causal relationship remains difficult identify. recognize existence direct connection countries expected rely heavily intensity exchanges. recognize sensitivity established direct transport service demand side expected depend location country respect major maritime routes. underline existence hysteresis direct connections necessarily sections major maritime routes. inclusion country fixed effects demand related variables GDP contribute limit incidence omitted variable bias. Reverse causality bilateral exports existence direct transport services excluded scope minor empirical set . Traditionally sea distance assumed main determinants freight rates trade competitiveness countries. findings presence indicator bilateral maritime connectivity, simplest form, impact maritime distance diminishes. impact standard gravity variables significantly affected inclusion indicator. results suggest quality maritime connectivity preponderant determinant foreign market access eventually export performance. provide instance estimate handicap terms export landlocked countries face top impact quality transit transports. LDCs development partners set ambitious goal doubling share LDCs’ exports global exports 2020. development partners agreed realize timely implementation duty free quota free market access, lasting basis, LDCs, simple, transparent predictable rules origin; reduction elimination arbitrary unjustified tariff barriers trade distorting measures. , measures sufficient. results improvement quality maritime connectivity core strategy aiming stimulating exports. improvement contribute reduction transport costs continue constitute greatest impediment LDCs’ trade competitiveness. Maritime Connectivity Trade 19 Intervening efficiently maritime connectivity easy task. options exist respective desirability reflect country specific characteristics. , investing infrastructures vital options. require financial effort countries bear . International cooperation partnerships crucial. International cooperation partnerships form establishing strategies aiming creating incentives shipping companies serve destinations necessarily profitable place. form granting companies serving "remote" countries preferential access major maritime hubs world. 20 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES REFERENCES Anderson van Wincoop (2003). Gravity Gravitas: Solution Border Puzzle. American Economic Review, 93, 170–192. Arvis -, Shepherd , Duval Utoktham (2013). Trade costs developing world : 1995 - 2010. Policy Research Working Paper Series . 6309, World Bank. Baier SL Bergstrand JH (2007). free trade agreements increase members' international trade Journal International Economics, 71(1), 72-95. Baier SL Bergstrand JH (2001). growth world trade: tariffs, transport costs, income similarity. Journal International Economics, 53(1), 1-27. Baldwin Taglioni (2006). Gravity Dummies Dummies Gravity Equations. CEPR Discussion Papers .5850. Bernhofen DM, El-Sahli Kneller (2013). Estimating effects container revolution world trade. CESifo Working Papers . 4136. CESifo, Center Economic Studies Ifo Institute. Burger MJ, van Oort FG Linders GJM (2009). specification gravity model trade: Zeros, excess zeros -inflated estimation. Spatial Economic Analysis, 4, 167-190. CEPII (2013). Centre de recherche français dans le domaine de ’économie internationale. GeoDist. http://www.cepii.fr/anglaisgraph/bdd/distances.htm Clark Dollar Micco (2004). Port efficiency, maritime transport costs, bilateral trade. Journal Development Economics, 75(2), 417-450. Duval (2011). UN ESCAP: Trade Facilitation Asia Pacific: Policies Measures affect Trade Costs Bangkok, 2011. http://www.unescap.org/ tid/publication/swp111.pdf. Ducruet Notteboom (2012). Worldwide Maritime Network Container Shipping: Spatial Structure Regional Dynamics. Global Networks, 12(3), 395-423. Dijkstra EW (1959). " note problems connexion graphs". Numerische Mathematik 1: 269-271. Gourieroux , Monfort Trognon (1984). "Pseudo maximum likelihood methods: Applications Poisson models", Econometrica, 52, 701-720. Hummels Lugovskyy (2006). " Matched Partner Trade Statistics Usable Measure Transportation Costs," Review International Economics, 14(1), 69-86. Hummels, David Georg Schaur (2012). Time Trade Barrier, NBER Working Paper 17758, January 2012, http://www.nber.org/papers/w17758.pdf. Notteboom TE (2004)."Container shipping ports: overview". Review Network Economics, 3(2), 86-106. Santos Silva JMC Tenreyro (2006), log gravity, Review Economics Statistics, 88, 641-658. Maritime Connectivity Trade 21 Santos Silva JMC Tenreyro (2014). Trading Partners Trading Flows: Implementing Helpman-Melitz-Rubinstein Model Empirically, Forthcoming Oxford Bulletin Economics Statistics. UNCTAD (2008). Transport Newsletter #38, March 2008. http://www.unctad.org/en/docs/sdtetlbmisc20081_en.pdf. UNCTAD (2011). Review Maritime Transport 2011. Geneva: United Nations. http://www.unctad.org/rmt. UNCTAD (2013). Review Maritime Transport 2013. Geneva: United Nations. http://www.unctad.org/rmt. Wilmsmeier Hoffmann (2008). Liner Shipping Connectivity Port Infrastructure Determinants Freight Rates Caribbean. Maritime Economics & Logitics, 130-151. Wilmsmeier Notteboom (2009). Determinants liner shipping network configuration: region comparison, Geojournal, November. 22 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES UNCTAD Study Series POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES . 30 Sam Laird, David Vanzetti Santiago Fernández de órdoba, Smoke mirrors: Making sense WTO industrial tariff negotiations, 2006, Sales . .05.II..16. . 31 David Vanzetti, Santiago Fernandez de órdoba Veronica Chau, Banana split: EU policies divide global producers, 2005, 27 . Sales . .05.II..17. . 32 Ralf Peters, Roadblock reform: persistence agricultural export subsidies, 2006, 43 . Sales . .05.II..18. . 33 Marco Fugazza David Vanzetti, South–South survival strategy: potential trade developing countries, 2006, 25 . . 34 Andrew Cornford, global implementation Basel II: Prospects outstanding problems, 2006, 30 . . 35 Lakshmi Puri, IBSA: emerging trinity geography international trade, 2007, 50 . . 36 Craig VanGrasstek, challenges trade policymaking: Analysis, communication representation, 2008, 45 . . 37 Sudip Ranjan Basu, link development institutions, policies geography, 2008, 50 . . 38 Marco Fugazza Jean-Christophe Maur, -tariff barriers computable general equilibrium modelling, 2008, 25 . . 39 Alberto Portugal-Perez, costs rules origin apparel: African preferential exports United States European Union, 2008, 35 . . 40 Bailey Klinger, South–South trade testing ground structural transformation, 2009, 30 . . 41 Sudip Ranjan Basu, Victor Ognivtsev Miho Shirotori, Building trade-relating institutions WTO accession, 2009, 50 . . 42 Sudip Ranjan Basu Monica Das, Institution development revisited: nonparametric approach, 2010, 26 . Maritime Connectivity Trade 23 . 43 Marco Fugazza Norbert Fiess, Trade liberalization informality: stylized facts, 2010, 45 . . 44 Miho Shirotori, Bolormaa Tumurchudur Olivier Cadot, Revealed factor intensity indices product level, 2010, 55 . . 45 Marco Fugazza Patrick Conway, impact removal ATC Quotas international trade textiles apparel, 2010, 50 . . 46 Marco Fugazza Ana Cristina Molina, determinants exports survival, 2011, 40 . . 47 Alessandro Nicita, Measuring relative strength preferential market access, 2011, 30 . . 48 Sudip Ranjan Basu Monica Das, Export structure economic performance developing countries: Evidence nonparametric methodology, 2011, 58 . . 49 Alessandro Nicita Bolormaa Tumurchudur-Klok, traditional trade flows economic crisis, 2011, 22 . . 50 Marco Fugazza Alessandro Nicita, importance market access trade, 2011, 35 . . 51 Marco Fugazza Frééric Robert-Nicoud, ‘Emulator Effect’ Uruguay United States regionalism, 2011, 45 . . 52 Sudip Ranjan Basu, Hiroaki Kuwahara Fabien Dumesnil, Evolution -tariff measures: Emerging cases selected developing countries, 2012, 38p. . 53 Alessandro Nicita Julien Gourdon, preliminary analysis newly collected data -tariff measures, 2013, 31 . . 54 Alessandro Nicita, Miho Shirotori Bolormaa Tumurchudur Klok, Survival analysis exports developed countries: role comparative advantage, 2013, 25 . . 55 Alessandro Nicita, Victor Ognivtsev Miho Shirotori, Global supply chains: Trade Economic policies developing countries, 2013, 33 . . 56 Alessandro Nicita, Exchange rates, international trade trade policies, 2013, 29 . . 57 Marco Fugazza, economics -tariff measures: Theoretical insights empirical evidence, 2013, 33 . . 58 Marco Fugazza Alain McLaren, Market access, export performance survival: Evidence Peruvian firms, 2013, 39 . . 59 Patrick Conway, Marco Fugazza . Kerem Yuksel, Turkish enterprise-level response foreign trade liberalization: removal agreements textiles clothing quotas, 2013, 54 . . 60 Alessandro Nicita Valentina Rollo, Tariff preferences determinant exports -Saharan Africa, 2013, 30 . 24 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES . 61 Marco Fugazza, Jan Hoffmann Rado Razafinombana, Building dataset bilateral maritime connectivity, 2013, 31 . . 62 Alessandro Nicita, Marcelo Olarreaga Peri Silva, Cooperation tariff waters World Trade Organization, 2014, 39 . . 63 Marco Fugazza Claudia Trentini, Empirical insights market foreign direct investment, 2014, 33 . . 64 Marco Fugazza, éline Carrè, Marcelo Olarreaga Fréderic Robert-Nicoud, Trade unemployment, 2014, 36 . . 65 éline Carrè Christopher Grigoriou, mirror data capture informal international trade, 2014, 42 . . 66 Denise Penello Rial, Study average effects -tariff measures trade imports, 2014, 26 . . 67 Cristian Ugarte, Weak Links diversification, 2014, 28 . . 68 Marina Murina Alessandro Nicita, Trading conditions: effect sanitary phytosanitary measures income countries' agricultural exports, 2014, 20 . . 69 Olivier Cadot, Alan Asprilla, Julien Gourdon, Christian Knebel Ralf Peters, Deep regional integration -tariff measures: methodology data analysis, 2015, 36 . . 70 Marco Fugazza, Maritime connectivity trade, 2015, 30 . Copies UNCTAD study series Policy Issues International Trade Commodities obtained Publications Assistant, Trade Analysis Branch, Division International Trade Goods Services, Commodities, United Nations Conference Trade Development, Palais des Nations, CH-1211 Geneva 10, Switzerland (Tel: +41 22 917 4644). studies http://unctad.org/tab. Printed United Nations, Geneva 1501294 () – February 2015 – 250 UNCTAD/ITCD/TAB/72 United Nations publication ISSN 1607-8291