Liner Shipping Connectivity and Port Infrastructure ... - IngentaConnect

Maritime Economics & Logistics, 2008, 10, (130–151) r 2008 Palgrave Macmillan Ltd All rights reserved. 1479-2931/08 $30.00

www.palgrave-journals.com/mel

Liner Shipping Connectivity and Port Infrastructure as Determinants of Freight Rates in the Caribbean GORDON WILMSMEIER1 & JAN HOFFMANN2 1 Tr a n s p o r t R e s e a r c h I n s t i t u t e , N a p i e r U n i v e r s i t y , C r a i g l o c k h a r t Campus, Edinburgh EH14 1DJ, UK. E - m a i l : g . w i l m s m e i e r @ n a p i e r. a c . u k ; 2 U N C TA D , P a l a i s d e s N a t i o n s , O f f i c e 7 0 4 4 , G e n e v a 1 211 , S w i t z e r l a n d . E-mail: [email protected]

The Caribbean basin lies at the cross roads of major east-west and north-south liner shipping routes. A number of regional ports have been able to take advantage of their geographical position. In other ports, however, the limited scale of hinterlands and the de facto monopolistic situation of ports in Small Island States have had a detrimental effect on port development. Port infrastructure endowment varies between highly equipped global transhipment hubs and rudimentary ports with wooden quays. By the same token, the supply of regular shipping services ranges between highly interconnected routes on the one side, and Small Island States that are heavily dependent on a few limited feeder services on the other. At the same time, freight rates in the region dispose of a high variability. The paper analyses the impacts of port infrastructure and liner shipping connectivity on intra-Caribbean freight rates. The structure of liner shipping services, port infrastructure endowment and liner shipping freight rates are closely related to each other. The paper will analyse these relationships. The empirical methodology includes principal component analysis and ordinary least-squares regressions. Maritime Economics & Logistics (2008) 10, 130–151. doi:10.1057/palgrave.mel.9100195

Keywords: Liner shipping connectivity; transport costs; competition; port infrastructure; principal component analysis.

INTRODUCTION Determinants of international transport costs are the topic of a growing recent literature. Interest in the topic arises from the desire to better explain economic

G Wilmsmeier & J Hoffmann Liner Shipping Connectivity and Port Infrastructure

131

development and international trade patterns, as well as to identify possibilities to reduce transaction costs. Most international trade continues to be transported by sea, and ports are crucial nodes in the global shipping networks. Most Central American and Caribbean countries trade very little with each other. By way of example, in 2005 less than 0.001% of Guatemala’s exports in manufactured goods were destined for Suriname, 0.24% for Jamaica, 1% for the Dominican Republic and around 8% for Costa Rica. What explains these differences? According to the standard gravity model, the participation of country B in global income is the basic determinant of the share of country A’s exports that are destined for country B; that is, if for example country B’s income is 5% of the world’s total income, it can, Ceteris paribus, be expected that 5% of country A’s exports will be destined for country B. As regards the impact of distance, the gravity model would suggest that countries that are further away from each other will trade less (see eg Tinbergen, 1962; Poÿho¨nen, 1963; Linnemann, 1966). However, traditional gravity models ignore the regular shipping liner services configuration. We argue that distance matters only indirectly in the determination of freight rates by shipping companies, because it does neither reflect market structures nor network configurations, which have become increasingly important as hub and spoke network structures have been emerging and potentially impose significant deviations on actual cargo flows. Pure geographic distance only accounts for the fact that the farther two countries are geographically apart the less likely it is that they are well connected by liner shipping services.

D E T E R M I N A N T S O F T R A N S P O R T C O S T S A N D R AT E S Transport costs are a major component of overall ‘trade costs’. Anderson and Wincoop (2004) provide an extensive review of trade costs, which are estimated to amount to a 170% ad valorem tax equivalent, including all transport, borderrelated and local distribution costs from the foreign producer to the domestic user. Initial work on the determinants of international transport costs, for example by Radelet and Sachs (1998), uses mainly explanatory variables that are related to distance and connectivity, such as if countries are landlocked, or if trading partners are neighbours, and to country characteristics such as GDP per capita. Martıńez-Zarzoso et al (2003) suggest that greater distance and poor partner infrastructure increases maritime transport costs notably. Inclusion of infrastructure measures improves the fit of the regression, corroborating the importance of infrastructure in determining transport costs. Hummels (1999, 2000, 2001) assesses whether international transport costs have declined over time, and introduces time as a trade barrier. Wilson et al (2003) find that port Maritime Economics & Logistics


132

efficiency has a strong and significant impact on bilateral trade flows in the Asia Pacific region. This paper focuses on the role of port infrastructure and liner shipping services as determinants of international maritime freight rates. The approach and use of explanatory variables, follows up the work of Fuchsluger (1999), Hoffmann (2002), Kumar and Hoffmann (2002), Sańchez et al (2003), Wilmsmeier (2003) and Wilmsmeier et al (2006). Unlike in previous papers, which analysed the international ‘freight’ derived from ‘Cost, Insurance, Freight’ (CIF) and Free on board (FOB) values stated in individual customs declarations, this work bases its analysis on reported freight rates on 189 routes in the Caribbean for 20 foot standard containers. These freight rates are derived from one major liner shipping company, ‘Company A’, providing shipping services throughout the Caribbean region in July 2006. Analysing freight ‘rates’ (published prices per TEU) from one company instead of actual transport costs (prices charged per ton of cargo) of all individual trade transactions of the entire market does not allow incorporating information on the value and volume of the transported goods in the regression analysis. Freight rates also not necessarily express real transport costs, because they are market driven. Further, the results have to be interpreted as applying to one company only. On the other hand, the information from Company A allows incorporating interesting new data on actual routes, journey times and transhipments, which previously could not be used when working with data provided by customs. Finally, the interpretation of results allows understanding the behaviour of Company A in terms of rate setting under specific market conditions. D I S TA N C E , L I N E R S H I P P I N G F R E I G H T R AT E S A N D T H E C O N C E P T O F LINER SHIPPING CONNECTIVITY The ‘geography of trade’, that is, the question of who trades what with whom, depends not only on the demand and supply of goods, but also on the ability to deliver the goods to the market. Relevant aspects include geographical factors such as distance, landlockedness and island character, as well as freight rates. Traditionally distance is assumed to be among the main determinants of freight rates and thus also of the trade competitiveness of countries. The sample of 189 freight rates of one company for the Caribbean, in principle confirms the general positive correlation between distance and freight rates. However, statistically, distance explains only one-fifth of the variance of the freight rate (see Figure 1). Important, yet often neglected, determinants of trade competitiveness are transport connectivity, defined as the access to regular and frequent transport services and the level of competition in the service supply. Maritime Economics & Logistics


133

3,500 y = 0.6206x + 1019.9 R2 = 0.2058

3,000

Freight rate

2,500 2,000 1,500 1,000 500 0 0

500

1,000

1,500

2,000

2,500

Distance Figure 1: Correlation between freight rate (USD) and distance (km). Source: Own elaboration

Most of international containerised cargo is transported by regular liner shipping services. Therefore, access to such services is a determinant of competitiveness and of the geography of trade. In this paper, we use attributes of liner shipping ‘connectivity’ such as the number of available services and the number of shipping lines operating direct services between pairs of countries and evaluate their potential as indicators to describe the market condition and service level between pairs of countries. Recent research has examined various aspects of maritime connectivity. Kumar and Hoffmann (2002), Marquez-Ramos et al (2006) and Wilmsmeier et al (2006) already incorporate measures of ‘connectivity’ into research on maritime transport costs. Angeloudis et al (2006) and Bichou (2004) look at connectivity in the context of maritime security. McCalla et al (2005) measure intermediacy and connectivity for Caribbean shipping networks and Notteboom (2006b) for seaport systems. Notteboom (2006a) also investigates the time factor in liner shipping services. UNCTAD (2006) developed a ‘liner shipping connectivity index’ per country. The remainder of this paper analyses the various relationships between freight rates, distance and various aspects of liner shipping connectivity, including transhipments, competition among shipping companies, port infrastructure endowment and transit times. The structure of the paper is as follows. The next section presents the theoretical the model specification. The further Maritime Economics & Logistics


134

sections describe the data used and identify the connectivity and port infrastructure endowment components for the empirical analysis and then discuss the results of the estimated transport costs equations. The penultimate section summarises the results and the last section draws conclusions and discusses the contribution to the literature.

M O D E L S P E C I F I C AT I O N The maritime freight rate charged by Company A (FREIGHTij) per 20 foot standard container of import cargo of country i from country j is assumed to depend on:

distance; various aspects of liner shipping connectivity; trade balance of containerised goods; various aspects of port infrastructure endowment; the countries’ general level of development.

These basic sets of variables are chosen, based on their relevance in previous publications (ie Clark et al, 2004; Sańchez et al, 2003; Wilmsmeier, 2003; Limao and Venables, 2000). Other variables, such as ‘being land-locked’, that have proven to be significant previously, are not relevant for the group of countries under study; and some variables that had been included in previous research were not built into the modelling approach, because this analysis is based on a different type of data set, which does not include information on individual trade transactions, but a published freight ‘rate’. Other variables with an impact on transport costs, such as fuel prices or vessel charter rates, are irrelevant for this analysis, because they generally do not depend on the chosen port or trade route. Unlike most of the previous research, we have chosen not to use an equation in its logarithmic form in the regression. The range of values in the set of observations is relatively narrow (for details see Table 1). Using absolute values, such as USD or km, allows for a more straightforward interpretation of the results. To capture the effect of distance, we include the distance in km between the two main ports of the importing and the exporting country. In order to capture ‘connectivity’ we include six indicator sets. To capture port infrastructure endowment, we include four different indicator sets of information. Connectivity and port infrastructure endowment are developed as ‘components’ using principal component analysis (PCA). In order to capture the trade balance, we include the coefficient of the balance of trade in manufactured goods in volume of tonnes between the two countries. data are described in Maritime Economics & Logistics

Table 1: Description of data Description

Min.

Max.

Average

Median

Standard deviation

n

FREIGHTRATE DISTANCE TRADE BALANCE GDPEXP GDPIMP NUMCAR

USD per TEU Km between main ports Tonnes of containerised imports/tonnes of containerised exports GDP per capita in the exporting country GDP per capita in the importing country Number of liner companies providing direct services between two countries Dummy variable ¼ 1 if there are more than three liner companies providing direct services between two countries Dummy variable ¼ 1 if Company A does not provide a direct service between two countries. It uses at least one transhipment Dummy variable ¼ 1 if there is at least one company in the market that provides a direct service between two countries Container carrying capacity deployed on direct services between two countries Total number of vessels deployed on direct services between two countries Maximum vessel size of ships deployed on direct services between two countries Number of direct services between two countries

650 118 0

3,290 2,158 1,096.75

1,641 1,001 27.40963

1,600 921 0.49972

625 457 98.52503

189 189 189

471 471 0

8,771 10,538 18

3,151 4,213 2.7

3,225 3,594 1

1,840 2,742 4.1

189 189 189

0

1

0.22

0

0.41

189

0

1

0.83

1

0.38

189

0

1

0.51

1

0.50

189

0

265,984

18,836

3,654

38,952

189

0

92

11

4

18.7

189

0

6,742

1,197

707

1,634

189

0

31

6.3

2

7.7

189

FOURCAR TRANSSHIPA DIRECT TEU

SHIPMAX NUMSERV

Source: Own elaboration 135

Maritime Economics & Logistics

NUMVES


Variable


136

more detail in the section ‘Data’. The model is given in equation 1. FREIGHTRATEij ¼ b0 þ b1 DISTANCEij þ b2 CONNECTij þ b3 TRANSHIPAij þ b4 DIRECTAij þ b5 FOURCARij þ b6 INFRAIMPi þ b7 INFRAEXPj þ b8 TRANSITTIMEij þ b9 SPEEDij þ b10 GDPIMPi þ b11 GDPEXPj þ b12 BALANCEROUTEij

ð1Þ

where b0 is a constant term; FREIGHTRATEij is the freight rate per 20 foot standard container (TEU), published to its clients by Company A; DISTANCEij is the distance in km between the main port of country i and the main port of country j; CONNECTij is an indicator describing liner shipping connectivity between the importing country i and the exporting country j. In the empirical analysis, different aspects and components of CONNECT are evaluated and interpreted in regard to their impact on freight rates. CONNECT itself is derived from PCA and includes the following attributes of connectivity: number of carriers, TEU deployed, number of services, number of vessels, shipping options and maximum size (TEU) of a ship on a specific route; TRANSHIPAij is a dummy variable, which is true (1), because service from company A involves at least one transhipment between countries i and j; DIRECTAij is a dummy variable, which is true (1), if there is at least one competing company in the market, which provides a direct service between countries i and j; FOURCARij is a dummy variable, which is true (1), if the number of shipping companies operating between countries i and j is greater than four; INFRAIMPi is an indicator for port infrastructure endowment in the importing country i. It is derived from PCA; INFRAEXPi is an indicator for port infrastructure in the exporting country j. It is derived from PCA; TRANSITTIMEij is the average transit time in days calculated from the liner shipping schedules between country i and j, in the case of transhipment routes transit times to and from transhipment ports were calculated and added, including also an estimated dwell time in the transhipment port; SPEEDij is the average speed in knots of the ships calculated from the distance, including transhipment deviation, and transit time between countries i and j; GDPIMPi is the gross domestic product (GDP) per capita in the importing country i; GDPEXPj is the GDP per capita in the exporting country j; and BALANCEROUTEij is the coefficient of the imports of containerisable cargo of country i received from country j divided by the exports of containerisable cargo from country i to country j. Maritime Economics & Logistics


137

CONNECTIVITY AND PORT INFRASTRUCTURE ENDOWMENT Data Approximately half of the 189 routes1 covered in our database are served by direct liner shipping services, whereas the other half includes transhipments in ports of third countries. By way of example, between Costa Rica and Colombia, there are 14 companies that provide direct services, deploying a total of 50 container ships, with a combined container carrying capacity of around 61,000 TEU; the largest vessel being of 2,500 TEU. Between Costa Rica and Jamaica, there are five companies/16 ships/17,400 TEU/2,105 TEU maximum size. Between Costa Rica and Guyana, there are no direct services. The freight rates of Company A on 189 routes of our sample range between 650 and 3,290 USD per TEU. Data are obtained from a major carrier in July 2006 (Hernańdez, 2006). Table 1 shows a summary of the explanatory variables, the expected sign, the data sources and the summary statistics.

Connectivity and port infrastructure – PCA The maritime network structure and its service supply are described by various indicators. We try to develop a connectivity measure by employing this complexity. Many of the possible determinants of liner shipping freight rates describing connectivity and port infrastructure endowment are closely correlated to each other. In order to control multicollinearity in the regression model and to better identify and separate the effects of shipping connectivity and port infrastructure, we use PCA to calculate non-collinear variables from the set of variables described above. In the first PCA model, we introduced 14 variables (see also Table 1 for description). The correlation matrix and the Kaiser-Meyer-Olkin (KMO) test (0.720) proved to be significant, which indicates the sampling adequacy of the chosen variables. The PCA extracted three factors, explaining 75.26% of the intrinsic variance of the data fulfilling the Kaiser criterion with eigenvalues over 1. Based on the given set of variables, the first component is referred to as CONNECT. It comprises the following variables:

NUMCAR: Number of carriers that provide direct services between two countries. The impact of this variable has already been discussed in more detail above. TEU: Total capacity of vessels deployed on direct services between two countries. NUMVES: Number of vessels deployed on direct services between two countries. Maritime Economics & Logistics


138

SHIPMAX: Size of the largest vessel that is deployed on direct services between two countries. SHIPPOS: Total number of shipping possibilities between the ports in country i and the ports in country j. NUMSERV: Number of direct services between two countries.

The second and third components are denominated INFRAEXP and INFRAIMP. They comprise information on the port infrastructure endowment of the exporting and importing country, respectively:

AREAEXP and AREAIMP: Port area in the country’s main port. STOREEXP and STOREIMP: Storage area in the country’s main port. LENGTHEXP and LENGTHIMP: Length of the quays in the country’s main port. MAXDRAFTEXP and MAXDRAFTIMP: Maximum draft in the country’s main port (Table 2).

E M P I R I C A L A N A LY S I S This section presents the estimates for different determinants influencing freight rates from Company A within the Caribbean basin. In total, 19 models based on the specified model in the section Model Specification were tested. Competition among carriers The estimation results presented in Table 5 aim at understanding the impact of distance, competition and transhipments on freight rates. The constant term ranges between around 1,500 and 1,700 USD. Independent of the maritime distance, the average fixed component in the freight rate most likely reflects port handling, repositioning and fixed administrative costs. The parameter for distance ranges between 0.28 and 0.36 USD per km. Roughly speaking, every 3 km of distance adds around 1 USD to the freight rate. In order to estimate the impact of competition on Company A’s freight rate setting on a specific route, three different variables are tested: NUMCAR (Models 1, 2, 3 and 4), DIRECT and FOURCAR. The results for NUMCAR in Model 1 (Table 5) can be interpreted that each additional carrier leads to a reduction of the freight rate by 79 USD. The underlying reasons may be increased competition and economies of scale. Figure 2, however, suggests that the relationship between the number of carriers and the freight rate is not quite that straightforward. Three groupings of freight rates can be identified in the sample. The first grouping contains the freight rate for those services that involve at least one transhipment, that is, no company provides a direct service on these particular routes. The lowest freight Maritime Economics & Logistics

Table 2: Partial correlation coefficients between variables included in principle components NUMCARIJ TEUij NUMVESij SHIPMAXij SHIPPOS NUMSERV EXPAREA EXPSTOR EXPLENGTH EXPMAXDRAF IMPAREA IMPSTOR IMPLENGTH IMPMAXDRAF 1.000

Export country Port area (m2) Storage area (m2) Quay length (m) Max draft (m)

0.912

0.550

0.880

0.939

0.224

0.258

0.157

0.074

0.293

0.373

0.431

0.293

1.000

0.923

0.679

0.641

0.690

0.204

0.213

0.207

0.168

0.266

0.313

0.498

0.359

1.000

0.630

0.816

0.881

0.220

0.243

0.192

0.123

0.291

0.360

0.515

0.357

0.468

0.602

0.108

0.104

0.091

0.111

0.234

0.284

0.435

0.329

1.000

0.908

0.209

0.256

0.232

0.115

0.271

0.363

0.498

0.328

1.000

0.201

0.239

0.152

0.104

0.299

0.397

0.482

0.357

1.000

0.929 1.000

0.629 0.602

0.333 0.434

0.027 0.031

0.013 0.034

0.000 0.006

0.000 0.009

1.000

0.582

0.038

0.044

0.038

0.030

1.000

0.002

0.025

0.003

0.033

1.000

0.937 1.000

0.662 0.678

0.380 0.466

1.000

0.601 1.000

Source: Own elaboration 139


Import country Port area (m2) Storage area (m2) Quay length (m) Max draft (m)

0.735


Number of carriers (ij) Deployed TEU (ij) Number of deployed vessels (ij) Maximum ship size (ij) Shipping possibilities port level(ij) Number of services (ij)


140

3,500

y =1814.9e-0.0671X R R2 = 0.4348

3,000

Freight rate

2,500 2,000 1,500 1,000 500 0 0

2

4 6 8 10 12 14 16 Number of Carriers providing direct services

18

20

Figure 2: Correlation between freight rate and the number of carriers providing direct services. Source: Own elaboration

rate, involving a minimum of one transhipment, between Costa Rica and Barbados, is 1,580 USD. The group also includes the highest freight rate of the sample, which is the rate between Belize and St Kitts and Nevis, at 3,290 USD (Table 3). A second grouping includes the freight rates for routes on which one to four companies provide direct services. In markets with up to four competitors freight rates appear to be higher than in markets with more market players. In fact, a similar result was found by Harding and Hoffmann (2003), also for Caribbean freight rates. Kent and Ashar (2001) report that in some national antimonopoly commissions, indices aimed at measuring market power are based on the market shares of the top four market players. It may be that up to four companies are more likely to get together to agree on prices than five or more competing market players. Finally, a third grouping is found for freight rates on routes with more than four direct shipping services suppliers. Once having clustered the freight rates within these three distinct categories, the impact of an individual additional company on the freight rate looks relatively weak. Instead of categorising the different groupings of NUMCAR, an alternative approach could have been the use of a non-linear functional form. The linear form used in our regression could in theory lead to a negative freight rate. Maritime Economics & Logistics


141 Table 3: Rotated component matrix Variables/component

1

2

CONNECT NUMCARij TEUij NUMVESij SHIPMAXij SHIPPOSij NUMSERVij

0.921 0.837 0.941 0.691 0.855 0.914

0.168 0.200 0.197 0.198 0.202 0.210

INFRAEXP EXPAREA EXPSTOR EXPLENGTH EXPMAXDRAF

0.128 0.159 0.100 0.003360

INFRAIMP IMPAREA IMPSTOR IMPLENGTH IMPMAXDRAF

0.009462 0.186 0.400 0.290

0.002634 0.004571 0.004543 0.0009626 0.928 0.925 0.771 0.598

3 0.112 0.137 0.125 0.002753 0.150 0.104 0.882 0.895 0.837 0.671 0.0002268 0.002022 0.004575 0.005458

Italic values are to highlight relevant values for each component. Source: Own elaboration

However, we believe that categorising allows for an interesting interpretation of our results, without the need to abandon the linear approach adopted in this paper. Models 2–6 aim at quantifying the freight rates for the three identified categories. By way of example, the results of Model 3 can be interpreted as follows: The freight rate for a route without any direct service, that is, involving at least one transhipment, is 1,725 USD, plus 28 cents for each km of direct maritime distance (the actual distance travelled is of course longer and we will attempt to capture the travel time later). If at least one company provides a direct service (the dummy variable DIRECT), the freight rate goes down by 516 USD. The underlying reasons may be the cost savings when no port handling for transhipment is required and less deviation distance and time. Finally, in case there are five or more competing carriers (the dummy variable FOURCAR) with direct services on a particular route, the freight rate decreases a further 306 USD. A note on global trends In view of the above described impact of competition on liner shipping freight rates, shippers and ports may be concerned about the effect of global consolidation on competition in a port or on a specific trade route. According to our research, up to the beginning of 2005, the number of Maritime Economics & Logistics


142

carriers offering services at individual ports from a global perspective had continued to increase despite the global process of concentration. Mergers and acquisitions meant that while there are fewer carriers today than 10 years ago, the same global carriers continued to expand into new markets. As a result, the number of carriers providing services to a specific port had actually increased for the majority of countries. Since mid-2005, however, the average number of carriers per country has started to decline. Table 4 compares the averages per country for the months of July 2004, July 2005 and July 2006. While the deployed TEU capacity per country and the average vessel sizes continue to increase, the average number of companies that provide services to a country’s ports has decreased from 21.5 (July 2005) to 20.3 (July 2006). Although the average reduction of around one company per country may not appear to be significant, at first sight, it can make a considerable difference for smaller markets – as appears to be the case of the Caribbean basin. Transhipments and competition with direct services In a next step, the analysis distinguishes between the services provided by Company A, that is, the company that reports the freight rate, and the services provided by the market as a whole. Although there may be companies in the market that provide a direct service on a given route, Company A may still only provides a service with transhipment. Table 4: Variance explained by components Component

Total of initial eigenvalues

% of variance explained

Cumulative % of variance explained

6.020 2.883 1.634 0.847 0.756 0.597 0.368 0.347 0.297 0.008984 0.006426 0.004459 0.004066 0.001167

43.000 20.590 11.675 6.047 5.399 4.265 2.628 2.481 2.124 0.642 0.459 0.318 0.290 0.002833

43.000 63.589 75.264 81.311 86.710 90.975 93.602 96.084 98.207 98.849 99.308 99.626 99.917 100.000

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Italic values are to highlight relevant values for each component. Source: Own elaboration Note: Extraction method: Principal component analysis. Rotation method: Varimax with Kaiser normalisation. Rotation converged in five iterations Maritime Economics & Logistics


143 Table 5: Regression results Variable/model Observations CONSTANT DISTANCE NUMCAR

1

2

3

4

5

6

N ¼ 189 1,497 (15.0) 0.36 (4.37) 78.6 (8.7)

N ¼ 189 1,719 (17.4) 0.29 (3.83) 42.7 (4.13) 492 (5.9)

N ¼ 189 1,725 (17.5) 0.28 (3.79) 13.8 (0.75) 516 (6.2) 306 (1.9) 0.523 53

N ¼ 189 1,495 (15.0) 0.35 (4.36) 65.1 (3.60)

N ¼ 189 1,719 (17.5) 0.29 (3.87)

N ¼ 189 1,641 (16.2) 0.36 (4.63)

545 (7.3) 407 (4.50) 0.524 70

696 (9.9)

DIRECTA FOURCAR Adjusted R2 F

0.428 71

0.516 68

153 (0.87) 0.428 48

0.480 86

Number of carriers and competition. Notes: T-statistics in parentheses. The dependent variable is the freight rate for a standard 20 foot container from the exporting country i to the importing country j. Models were estimated by OLS

The results presented in Table 7 support the general thrust of those presented in Table 5. In addition, they allow for some interesting interpretations regarding the cost of transhipments and the impact of competition. If Company A’s service on a given route is not direct, but involves at least one transhipment, the freight rate charged to its client will be roughly 600 USD higher, reflecting the additional port costs and deviation to the transhipment port. However, if competing companies in the market do provide direct services on that same route, Company A seems to adjust its freight rate. Instead of charging an extra 619 USD (Model 9), it can only charge an extra 106 USD (619 minus 513). In fact, if there are four or more companies in the market with direct services, Company A’s estimated freight rate will go down by 612 USD (Model 11: the combined impact of DIRECTA and FOURCAR is 313+299 ¼ 612). Thus, also overcompensating the additional costs for transhipment, which results in a freight rate reduction of 21 USD (581612 ¼ 21) in comparison to a route with a direct service and no competition. Model 11 further supports the categorisation of routes based on the number of companies competing with direct services. Whereas DIRECTA and FOURCAR are statistically significant with strong impacts on the freight rate of Company A, the variable NUMCAR is no longer statistically significant. Transit time and distance Many routes in the sample are not served by direct services, and the discussion regarding the number of companies is not applicable for those routes served only by transhipment services. Consequently, the variable maritime DISTANCE Maritime Economics & Logistics


144

3000 2500

Freight rate

2000 1500 1000 y = 55.796x + 904.6 R2= 0.2979

500

0

5

10 15 20 Transit time (Days)

25

30

Figure 3: : Correlation between freight rate and transit time. Own elaboration

cannot be applied, because transhipment will inevitably require a deviation, which is not included in the variable values. In order to better capture the real distance for all routes, we have gathered and calculated the effective transit time between pairs of countries, based on the offered services (October 2006), that is, including deviations for transhipment. And in effect, transit time (Figure 3) is closer correlated with the freight rate than the direct maritime distance (cf Figure 1). Figure 3 illustrates that, on average, each additional day of transit will lead to an increase of the freight rate of around 56 USD. Connectivity, port infrastructure and transit time As a next step, the three component variables (CONNECT, INFRAIMP and INFRAEXP), derived from PCA (section Connectivity and Port Infrastructure Endowment), are included in the analysis as explanatory variables in the regressions. Further, DISTANCE and TRANSITTIME (Table 8) are added as independent variables to the model (Table 6). Maritime distance and transit time are closely correlated and therefore cannot be used simultaneously as estimators in our model due to multicollinearity. Comparing Models 12 and 13 (Table 8), TRANSITTIME has a stronger statistical impact on the freight rate than DISTANCE and allows controlling for deviation due to transhipment. The impact of TRANSITTIME might also reflect the elasticity of the cargo owner to pay for a more rapid delivery. Thus, in the subsequent regression transit time is used instead of distance. Maritime Economics & Logistics


145 Table 6: Fleet deployment and companies providing services per country, 2004–2006

2004

2005

Average TEU capacity 296,025 309,658 deployed per country Average vessel size (TEU) 1,212 1,254 Average # of companies 21.7 21.5 per country

Percentage change 2005/2004 +4.6 +3.4 0.7

Percentage Percentage change change 2006/2004 2006/2005

2006 337,940 1,399 20.3

+14.2

+9.1

+15.4 6.2

+11.6 5.5

Source: Authors, based on ci-online.co.uk. See also UNCTAD Transport Newsletter, Fourth Quarter 2006. Data are global, for 161 countries

Improved connectivity (CONNECT) has a significant impact on the freight rate. An increase of CONNECT (composed of the number of carriers, level of deployed TEU, number of vessels, vessel sizes, number of shipping possibilities and the number of services) of one standard deviation implies a potential reduction in the freight rate of 287 USD (Model 12). Improved port infrastructure (INFRAIMP and INFRAEXP), too, has a significant impact on the freight rate. The order of magnitude is similar to that of CONNECT. Especially the port infrastructure in the importing country seems to have a strong bearing on the freight rate. An increase of INFRAIMP of one standard deviation reduces the freight rate of Company A by 225 USD (Model 12). The impact of INFRAEXP is weaker. Further determinants of the freight rate In Model 14, further variables are introduced, notably CONNECT, INFRAIMP and INFRAEXP (Table 9). If a country imports more than it exports, on a given route, the corresponding freight rate should increase, because it has to cover the cost of returning the empty container to its origin. The estimated parameter for BALANCEROUTE has the expected positive sign. It is statistically significant and stable in different models. The estimated value of the parameter would imply that an imbalance of 10, that is, a country imports 10 times more than what it exports, leads to an increase of the freight rate of 8.40 USD (Model 15). The impact of imbalances is difficult to measure for pairs of countries, because liner services usually cover trade between regions, or even continents. Also, it is questionable if a coefficient of imports/exports is the best way to measure imbalances, because it can reach very high absolute values if a country exports virtually nothing to another country. Nevertheless, considering all these restrictions, BALANCEROUTE helps to improve the econometric ‘fit’ of the econometric model. Maritime Economics & Logistics


146

The level of development of a country can be expected to be positively correlated with transport service quality. More use of information and communication technologies, better educated populations, fewer accidents and less corruption may all be correlated positively to the GDP per capita, which can thus be taken as a proxy for ‘development’. In the case of countries in the Caribbean basis (Models 16 and 17), it appears that effectively a higher GDP per capita is reflected in reduced freight rates. The difference between the poorest and richest country in the sample is about 10,000 USD. Based on Model 16, it can be expected that export freight rates from the richest country will be 410 USD lower than export freight rates from the poorest country – although there most likely also exists a different causality: Lower freight rates improve a country’s trade competitiveness, thus helping to increase its GDP. A higher service speed (knots) has a price, and SPEED leads to an increase of freight rate of around 39 USD. It is interesting to note that if SPEED is included in the regression together with TRANSITTIME, the estimated parameter for TRANSITTIME also increases (Models 17 and 18). At a slower speed, a longer transit time might actually lead to some fuel savings, reducing the expected incremental impact on the freight rate (Table 9). Finally, in Model 19, we re-incorporate the variables TRANSHIPA and DIRECTA, which we had already included in Models 9 and 11. The estimated parameter values and their statistical significance are in the same range as before. The issue of transhipment services is correlated with CONNECT and TRANSITTIME, and also with INFRAIMP and INFRAEXP, because the region’s

Table 7: Regression results: Transhipments and competing direct services Variable/model Observations CONSTANT DISTANCE TRANSHIPA

7

8

9

10

11

N ¼ 189 609 (6.2) 0.35 (4.3) 821 (8.2)

N ¼ 189 1,025 (9.2) 0.24 (3.1) 627 (6.5) 57 (6.3)

N ¼ 189 1,146 (10.3) 0.24 (3.2) 619 (6.8)

N ¼ 189 1,037 (10.1) 0.24 (3.3) 644 (7.1)

N ¼ 189 1,213 (10.9) 0.21 (2.9) 581 (6.5) 6 (0.4) 313 (3.7) 299 (2.5) 0.578 52

NUMCAR

513 (7.6)

DIRECTA FOURCAR Adjusted R2 F

0.412 67

0.512 67

0.550 78

540 (7.6) 0.549 77

Notes: T-statistics in parentheses. The dependent variable is the freight rate for a standard 20 foot container from the exporting country i to the importing country j. Models were estimated by OLS Maritime Economics & Logistics


147 Table 8: Regression results: Connectivity, port infrastructure and transit time Variable/model Observations CONSTANT DISTANCE CONNECT INFRAIMP INFRAEXP

12

13

14

N ¼ 173 1,295 (16.2) 0.28 (3.8) 287 (8.9) 225 (7.1) 64 (2.0)

N ¼ 164 1,237 (15.5)

N ¼ 173 1,573 (49.2)

240 (6.8) 210 (6.6) 84 (2.7) 28 (4.4) 0.531 47

323 (10.1) 253 (7.9) 95 (3.0)

TRANSITTIME Adjusted R2 F

0.532 50

0.496 57

Notes: T-statistics in parentheses. The dependent variable is the freight rate for a standard 20 foot container from the exporting country i to the importing country j. Models were estimated by OLS. N varies somewhat due to unavailable data for some variables on some routes

main transhipment ports are also those with the best port infrastructure. Hence, the estimated impacts of CONNECT, TRANSITTIME, INFRAIMP and INFRAEXP are somewhat reduced in Model 19. INFRAEXP is in fact no longer statistically significant. Nevertheless, the overall magnitude and direction of impacts remains as before. Model 19 statistically explains more than 60% of the variance of Company A’s freight rates for the Caribbean in July 2006, using just six explanatory variables, not including DISTANCE.

SUMMARY OF FINDINGS Distance is usually assumed to be among the main determinants of transport costs and thus also of the trade competitiveness of countries. For a sample of 189 liner shipping freight rates of one company in the Caribbean, reported in July 2006, the general positive correlation between distance and freight rates is confirmed in principle. However, statistically, distance explains only one-fifth of the variance of the published freight rate and its explanatory value has to be questioned in transhipment markets. The number of liner shipping companies providing direct services between pairs of countries appears to have a stronger impact on the freight rate than does distance. For routes where there is no company providing direct service, that is, where all containerised maritime trade involves at least one Maritime Economics & Logistics


148 Table 9: Regression results: Connectivity, port infrastructure, transit time and additional variables Variable/model Observations CONSTANT CONNECT INFRAIMP INFRAEXP TRANSITTIME

141

15

16

17

18

19

N ¼ 164 1,237 (15.5) 240 (6.8) 210 (6.6) 84 (2.7) 28 (4.4)

N ¼ 164 1,207 (15.3) 226 (6.5) 214 (6.8) 94 (3.1) 28 (4.5) 0.84 (2.8)

N ¼ 164 1,356 (14.3) 204 (5.8) 212 (6.9) 105 (3.4) 33 (5.0) 0.81 (2.7) 0.01755 (1.4) 0.04101 (2.5)

N ¼ 164 1,057 (8.3) 174 (4.9) 196 (6.5) 88 (2.9) 48 (6.2) 0.86 (3.0) 0.02419 (2.0) 0.04505 (2.8) 39 (3.4)

N ¼ 164 927 (7.5) 204 (5.8) 201 (6.5) 78 (2.5) 40 (5.5) 0.91 (3.1)

N ¼ 164 948 (9.7) 130 (3.3) 128 (4.0) 16 (0.5) 13 (2.0)

BALANCEROUTE GDPIMP GDPEXP SPEED

34 (2.9)

TRANSHIPA DIRECTA Adjusted R2 F

0.531 47

0.551 41

0.566 31

0.593 31

0.571 37

708 (5.7) 308 (3.5) 0.608 43

Model 14 was already presented above. It is included again in this table for ease of reference. Notes: T-statistics in parentheses. The dependent variable is the freight rate for a standard 20 foot container from the exporting country i to the importing country j. Models were estimated by OLS

transhipment in a third country’s port, freight rates in our sample range from 1,170 to 3,290 USD, with an average of 2,056 USD. For routes with one to four carriers providing direct services the reported freight rates range from 650 USD to 2,250 USD with an average of 1,449 USD. If five or more competing carriers provide direct services, the freight rate ranges from 650 to 1,730 USD, averaging 973 USD. Statistically, the number of carriers explains around two fifths of the variance of the freight rate. More detailed analysis suggests that the following variables have a statistically significant impact on liner shipping freight rates in the Caribbean:

transhipment versus direct services; the number of competing carriers; an index of liner shipping connectivity; transit time; and port infrastructure endowment in the importing and exporting countries. A model that incorporates the above variables statistically explains threefifths of the variance of the freight rate. The empirical results support the Maritime Economics & Logistics


149

hypothesis that competition between shipping lines makes shipping services less expensive for the shipper, that is, oligopolistic market structures imply higher costs to shippers. Transit time is a more precise determinant of transport costs than distance, especially for bilateral trade routes that are not connected by direct services. At the same time trade imbalance is an important determinant of transport costs, implying higher costs for exporters from container deficit regions. A good level of port infrastructure endowment implies a reduction of transport costs. Moreover, transport costs for trade between ‘richer’ countries are potentially lower. These results suggest that there exists a potentially virtuous circle, where higher trade volumes and economic development help to reduce transport costs, which in turn helps to promote trade and development. This circular causality could also be among the explanations why the standard gravity model tends to provide biased estimates, overpredicting trade between low-volume traders and to underpredicting trade between high-volume traders (Wall, 2000).

CONCLUSIONS The analysis of determinants for freight rates of a single company provides new insights on the impact of market structure on maritime freight rates. A less concentrated liner shipping market reduces freight rates for shippers. In a disperse market with low trade volumes on many routes, like the Caribbean, the number of carriers offering direct services in many cases exhibits diseconomies of scale and oligopolistic market structures, which in return induces higher transport costs for trade on the respective routes. Combining information on the service of the shipping company itself with information on the market structure allowed us to analyse how the freight rates of a single company appear to influence strategic behaviour towards market competitors. Since these strategies are private sector decisions the potential influence of public policies to reduce maritime freight rates as part of trade costs is restricted. Further, the results show that trade routes with only indirect services (ie including transhipments) induce higher transport costs. The analysis suggests that transhipment has the equivalent impact on freight rates as an increase in distance between two countries of 2,612 km. In the case of the Caribbean this implies that in many cases intra-regional trade between small islands is not competitive as compared to trade for example with the United States, because of the lack of direct services. These findings support arguments that aim at promoting new innovative short sea shipping concepts in the Caribbean, which offer direct services especially between smaller islands, to strengthen intraregional trade. Maritime Economics & Logistics


150

Incorporating effective transit times into the analysis allows to better estimate transport costs for trade flows that include transhipments. This can be seen as an important improvement of previous models, where direct maritime distance was used as a proxy of transit times. The importance of a high degree of connectivity, that is, being relatively central in the Caribbean maritime network, is indicated by the results for the connectivity variables and the impact of network structures should be included in future research. These results also underline the potential benefits for importers and exporters of being based near a transhipment hub. The level of port infrastructure, such as berth length, storage capacities, maximum draft and port areas, appear to have a significant reducing impact on freight rates. This is important for policy makers, as most of the other variables that determine freight rates are beyond their control. In ports, however, the public sector can make a difference, reducing transport costs and attracting shipping services, thus further improving transport connectivity and trade competitiveness.

Acknowledgements We acknowledge important contributions in putting in the data gathering and initiation of this research by Lester Hernandez, during his internship with UNCTAD in summer 2006, and Omar Salgado, S&W Maritime Knowledge Network, for gathering data on the structure of liner services in the Caribbean. ENDNOTE 1

Including the following countries: Antigua and Barbuda, Barbados, Belize, Colombia, Costa Rica, Dominica, Dominican Republic, Grenada, Guatemala, Guyana, Haiti, Honduras, Jamaica, Mexico, Panama, St Kitts and Nevis, St Lucia, St Vincent and the Grenadines, Suriname, Trinidad and Tobago, Venezuela, RB.

REFERENCES Anderson, JE and van Wincoop, E. 2004: Trade costs. Journal of Economic Literature XLII: 691–751. Angeloudis, P, Bichou, K, Bell, M and Fisk, D. 2006: Security and reliability of the liner container shipping network: Analysis of robustness using a complex network framework. Presented to IAME 2006 Conference, Melbourne. Bichou, K. 2004: The ISPS code and the cost of port compliance: An initial logistics and supply chain framework for port security assessment and management. Maritime Economics and Logistics 6: 322– 348. Clark, X, Dollar, D and Micco, A. 2004: Port efficiency, maritime transport costs, and bilateral trade. Journal of Development Economics 75: 417–450. Containerization International Online. www.ci-online.co.uk. Fuchsluger, J. 1999: Analysis of maritime transport costs in South America. Unpublished Master’s Thesis. University of Karlsruhe, Germany. Maritime Economics & Logistics


151 Harding, A and Hoffmann, J. 2003: Trade between Caribbean Community (CARICOM) and Central American Common Market (CACM) countries: The role to play for ports and shipping services. Serie de la CEPAL 52, Santiago de Chile, May. Hernańdez Cruz, L. 2006: Intraregional trade and liner shipping services in the association of Caribbean states. Unpublished Masters Thesis, Erasmus University, Rotterdam. Hoffmann, J. 2002: El costo del transporte internacional, y la integracioń y competitividad de Ame´rica Latina y el Caribe. Boletıń Fal No. 191, July 2002. United Nations ECLAC, Santiago. Hummels, D. 1999: Toward a geography of trade costs Mimeo, University of Chicago, Chicago, United States. Hummels, D. 2000: Time as a trade barrier Mimeo, Purdue University, October. Hummels, D. 2001: Have international transport costs declined? Journal of International Economics 54: 75–96. Kent, P and Ashar, A. 2001: Port competition regulation: A tool for monitoring for anti-competitive behaviour. International Journal of Maritime Economics 3: 27–51. Kumar, S and Hoffmann, J. 2002: Globalization, the maritime Nexus. In: Grammenos, C.Th. (ed.). Handbook of Maritime Economics and Business. LLP: London. Limao, N and Venables, A. 2000: Infrastructure, geographical disadvantage, and transport costs and trade. World Bank Policy Research Paper 2257, Washington, DC: The World Bank. Linnemann, H. 1966: An econometric study of international trade flows. North Holland Publishing Company: Amsterdam. Marquez-Ramos, L, Martıńez-Zarzoso, I, Pe´rez-Garcıá, E and Wilmsmeier, G. 2006: Determinants of maritime transport costs. Importance of connectivity measures Presented to International Trade and Logistics Conference, Le Havre. Martıńez-Zarzoso, I, Garcıá-Meneńdez, L and Sua´rez-Burguet, C. 2003: Impact of transport costs on international trade: The case of Spanish ceramic exports. Maritime Economics & Logistics 5: 179–198. McCalla, R, Slack, B and Comtois, P. 2005: The Caribbean basin: Adjusting to global trends in containerization. Maritime Policy and Management 32: 245–261. Notteboom, T. 2006a: The time factor in liner shipping services. Maritime Economics and Logistics 8: 19–39. Notteboom, T. 2006b: Traffic inequality in seaport systems revisited. Journal of Transport Geography 14: 95–108. Poÿho¨nen, P. 1963: A tentative model for the volume of trade between countries. Weltwirtschaftliches Archiv 90: 93–99. Radelet, S and Sachs, J. 1998: Shipping costs, manufactured exports, and economic growth. Mimeo, Paper presented at the American Economic Association Meetings, Harvard University. Sańchez, R, Hoffmann, J, Micro, A, Pizzolitto, G, Sgut, M and Wilmsmeier, G. 2003: Port efficiency and international trade: Port efficiency as a determinant of maritime transport cost. Maritime Economics and Logistics 5: 199–218. Tinbergen, J. 1962: Shaping the world economy Suggestions for an international economic policy, Twentieth Century Fund, New York. UNCTAD. 2006 Transport Newsletter, various issues: http://extranet.unctad.org/transportnews. Wilmsmeier, G. 2003: Modal choice in South American freight transport: Analysis of constraint variables and a perspective for diversified modal participation in South America Unpublished Master’s Thesis, Technische Universita¨t Dresden. Wilmsmeier, G, Hoffmann, J and Sańchez, R. 2006: The impact of port characteristics on international maritime transport costs. In: Cullinane, K and Talley, W (eds). Port Economics, Research in Transportation Economics Vol. 16 Elsevier: Amsterdam. Wall, HJ. 2000: Gravity model specification and the effects of the Canada-U.S. border Working Paper 2000-024A, Federal Reserve Bank of St. Louis. Wilson, J, Mann, C and Otsuki, T. 2003: Trade facilitation and economic development World Bank Policy Research Working Paper 2988, Washington.
