Journal of Productivity Analysis

, Volume 40, Issue 1, pp 83–98

Determinants and strategies for the development of container terminals

Authors

    • UCL QASER LaboratoryUniversity College London
  • Qianwen Liu
    • Drewry Shipping Consultants
Article

DOI: 10.1007/s11123-012-0291-1

Cite this article as:
Medda, F. & Liu, Q. J Prod Anal (2013) 40: 83. doi:10.1007/s11123-012-0291-1

Abstract

The paper examines an important player in the container industry: the container terminal. We evaluate how terminal efficiency operation is affected by the following factors: terminal type, operation type, scale efficiency and returns to scale. In so doing, we test how the typology and operation of terminals and the level of scale efficiency that a terminal can achieve, represent significant factors in the development and growth of the container terminal industry. The analysis is based on the assessment of 165 container terminals worldwide. We develop the estimation through the application of stochastic frontier analysis. We demonstrate that container terminals are more efficient than multi-purpose terminals, and that, compared with local operators, global terminal operators do not have a dominant position in international maritime trade in terms of productivity and efficiency. However, global terminal operators appear to be more dominant than local operators when we examine the Mediterranean Basin. In the final part of the paper we suggest how resource-constrained container terminals may improve their scale efficiency and identify general strategies related to container terminal investments.

Keywords

Stochastic analysisContainer terminalGlobal operatorScale and technical efficiency

JEL Classification

L91L99

1 Introduction

The ports of Antwerp and Rotterdam, throughout their long history of competition, have often invested significant resources in new terminal facilities in order to establish their roles as maritime leaders (Loyen et al. 2003). Over time, the progressive development of terminals towards specific and defined roles has largely shaped competition, selection and concentration in global maritime activity (Slack and Fremont 2005; Olivier 2005). In this context, due to the rapid economic growth of emerging markets in the last two decades, container terminals in particular have been recognized as the most dynamic and globalized segment of maritime operations (Steenken et al. 2004).

The collapse of global financial institutions, followed by the economic and financial crisis, has led to the present slowdown of global container traffic; according to Notteboom and Rodrigue (2011), “the year 2008 was a turning point for the container terminal operator, as the final quarter saw unprecedented volume decline.” Nevertheless, prior to the economic downturn of 2008, containerized traffic has accounted for an average annual growth rate of 9.2 % since 1985 (Lloyd’s Register-Fairplay Research 2008), and the development of containerized traffic has grown from 28.7 million Twenty-foot Equivalent Units (TEUs) in 1990 to 148.9 million TEUs in 2008 (Drewry Shipping Consultants Ltd 2008). Many drivers have contributed to this steady growth and have strongly influenced and accelerated the internationalization and consolidation of the container terminal industry (Fremont 2007). Several scholars (Bascombe 1998; Musso et al. 2001; Martin and Thomas 2001; Brennan 2002; Cullinane and Song 2002) note that the changing management structure in ports, through privatization and liberalization–and consequently the reduction of public finance subsidies–has strengthened and extended container terminal industry activity. Terminals have also gained the ability to provide reliable services and hinterland connections that fit into global alliance networks. Another key driver in the evolution of terminal operations has been due to the shipping industry itself; for instance, the adoption of large container ships, new long-distance trade corridors (De Souza et al. 2003), and the development of a specialized freight industry requiring longer turnaround times, advanced logistics and operational structures in relation to connectivity and synchronisation of services (Heaver et al. 2001; Alicke 2002). Altogether, these elements have posed an on-going challenge for container terminals in the aim to optimize their container traffic flows (Levinson 2008).

Given the significant evolution in the container terminal industry towards highly specialized global terminal operators, Slack and Fremont (2005) surmise that “important questions arise out of the transformation of the terminal. The economic efficiency of dedicated terminals versus common user facilities is a basic research question … The evidence is unclear.” Against this background, the objective of the paper is to examine the container terminal operation industry and assess how terminal efficiency is affected by three main factors: terminal type, operation type, scale efficiency and returns to scale. In so doing, we test how the typology and operation of terminals, and the level of scale efficiency that a terminal can achieve, are significant factors in the development and growth of the container terminal industry (Cullinane et al. 2005; Slack and Fremont 2005; Heaver 2006; Wiegmans et al. 2008).

In addition, within our context we provide another component to the analysis. Although various scholars (Wiegmans et al. 2002; Notteboom and Rodrigue 2005) remark that the location of a terminal has become progressively less sensitive to the shippers’ call decision, other studies (Airriess 2001; Haralambides et al. 2002; Slack and Fremont 2005) point out that terminal location is significant because of the uniqueness of interactions within a specific region. We therefore examine a set of container terminals in the Mediterranean Basin in order to verify our hypotheses in a specific regional area.

The paper proceeds as follows. Section 2 describes the model specification and data. In Sects. 3 and 4 we report on the model estimation and interpret the results. We summarize the top performance terminals in different efficiency indices in Sects. 5 and 6, and in Sect. 7 we draw conclusions and identify general strategies related to container terminal investments.

2 Hypotheses formulation and data description

Three main factors constitute the backbone of our analysis on terminal development: terminal type, operation type and scale efficiency and returns to scale. In order to address the objectives, we begin by examining these factors individually:

Factor 1: Terminal type

According to Steenken et al. (2004), a container terminal is “an open system of material flow with two external interfaces … the quayside with loading and unloading of ships, and the landside where containers are loaded and unloaded on/off trucks and trains”. We distinguish between multi-purpose terminal and container terminal in this paper. A multi-purpose terminal handles any type of cargo such as bulk and container, whereas a container terminal specializes in container handling only. Given this distinction of terminal typology, we assume that specialization in terminals leads to an increase in production; therefore, in relation to container operations we test the following hypothesis: Container terminals are more productive than multi-purpose terminals, ceteris paribus.

Factor 2: Operation type

Following the literature (Trotman-Dickenson 1996; Slack and Fremont 2005; Olivier et al. 2007; Rodrigue and Notteboom 2011), we can generally classify container terminal operators as either local terminal operators or global terminal operators. We define terminal operators that operate in more than one world region1 as global terminal operators, whereas terminals that operate in a single world region are known as local terminal operators. We assume that global terminal operators, because they can establish world connections with other terminal operators, share their experience among different terminals in order to achieve high levels of efficiency in their operations. Our second hypothesis in the study is thus: Global container terminal operators are more efficient than local terminal operators, ceteris paribus.

Factor 3: Scale efficiency and returns to scale

The container terminal industry is capital-intensive (Notteboom and Rodrigue 2011) and scale economies diminish if container terminal growth expands substantially. Hence, because increasing returns to scale may diminish in relation to terminal operations, we focus on how and the extent to which it does.

After having defined the research lines, we next examine the data sets under consideration. In the analysis we consider 165 container terminals worldwide which we divide into two sets. The first set in Table 1 lists the terminals belonging to the top 19 container ports in the world in relation to levels of throughput in 2006.2
Table 1

World’s top container ports ranked in terms of TEU throughput levels in 2006

Country

Port

No. of terminals studied

Singapore

Singapore

6

Hong Kong

Hong Kong

6

PRC

Shanghai

7

PRC

Shenzhen

9

S Korea

Busan

9

Taiwan

Kaohsiung

15

The Netherlands

Rotterdam

9

PRC

Qingdao

2

UAE

Dubai

2

Germany

Hamburg

6

USA–W

Los Angeles

7

USA–W

Long Beach

7

PRC

Ningbo

3

Belgium

Antwerp

11

PRC

Guangzhou

2

Malaysia

Port Klang

2

PRC

Tianjn/Xingang

4

USA–E

New York

7

Germany

Bremerhaven

4

In order to test location-specific terminals, the second set of terminals is represented by those belonging to 30 Mediterranean container ports (Table 2). The choice of ports is made in relation to the value of throughput in 2006 (over 10.000 TEU), but also in order to account for different institutional arrangements present in the Basin, that is, ports in European Union (EU) member countries (France, Italy, Slovenia, Greece, Malta, and Spain), ports belonging to EU-Candidate countries (Turkey and Croatia), and one port (Bar) in Montenegro, which is neither an EU nor an EU-Candidate country. The rapid expansion of the Mediterranean terminals, begun during the 1990s, is indicative of a market outcome which is mainly the result of the growth of globalization, the establishment of long-distance east–west freight traffic routes, and first-mover advantage within the constraints of location convenience. However, as observed in various studies dedicated to the Mediterranean Basin (Barros and Athanassiou 2004; Barros 2006; Trujillo and Tovar 2007; Barros and Peypoch 2007; Bergantino and Musso 2011), EU intervention may have facilitated port investment and thus be determining, on average, an increase in productivity.
Table 2

Mediterranean container ports in our data base ranked in terms of throughput levels in 2006

Country

Port

No. of terminals studied

Spain

Algeciras

2

Italy

Gioia Tauro

1

Spain

Valencia

4

Spain

Barcelona

5

Italy

Genoa

3

Malta

Marsaxlokk

1

Greece

Piraeus

2

Italy

La Spezia

2

France

Marseilles-Fos

2

Italy

Taranto

1

Turkey

Izmir

1

Italy

Cagliari

1

Turkey

Mersin

1

Italy

Naples

2

Greece

Thessaloniki

1

Italy

Salerno

2

Italy

Venice

2

Italy

Trieste

1

Slovenia

Koper (Capodistria)

1

France

Sete

1

Spain

Alicante

1

Italy

Ravenna

2

Spain

Cadiz

1

Spain

Seville

1

Croatia

Rijeka

1

Malta

Valetta

1

Spain

Cartagena (Spain)

1

Turkey

Antalya

1

Montenegro

Bar

1

Spain

Tarragona

1

Our data is derived from the Drewry Shipping Database (2006, 2007) for year 2006. The decision to focus on this specific period is due mainly to the global financial and economic crisis, starting in 2008, which has largely impacted on maritime and terminal flows. The crisis has exposed the vulnerability of the maritime industry and therefore prompted terminal operators to shift their strategies towards greater rationalization of the services and investment, and to a “more cautious assessment of the future prospects” (Notteboom and Rodrigue 2011). The same level of maritime traffic pre-crisis is nevertheless expected to return in the future (Vanelslander and Van de Voorde 2009); we have therefore chosen to analyze the results from year 2006 in order to provide insights into possible future trends in container terminal operations.

For each of the 165 terminals we consider as inputs the following physical characteristics: maximum berth depth, quay length, yard space, crane spacing, and number of gantry cranes. Maximum berth depth, quay length and yard space represent the infrastructure category, whereas the crane spacing and number of gantry cranes variables represent the equipment category. In addition to five physical inputs, we introduce two binary variables that indicate terminal type and operation type. The output of the terminal industry is identified as TEU. Table 3 provides a summary of the variables.
Table 3

Variable specification

Output

y

TEU

Number

 

Inputs

×1

Max berth depth

Meter

 

×2

Quay length

Meter

 

×3

Yard space

Hectare

 

×4

Number of Gantry cranes

Number

 

×5

Crane spacing

Meter

 

Binary variables

Z1

Terminal type

Binary

0 = container terminal

1 = multi-purpose terminal

Z2

Operation type

Binary

0 = global terminal operator

1 = local terminal operator

The descriptive statistics of the six continuous variables, one output and five inputs, is shown below in Table 4.
Table 4

Descriptive statistics of terminal-level data

Variables

y

x1

x2

x3

x4

x5

Throughput

Max berth depth

Quay length

Yard space

Number of Gantry

Crane spacing

Mean

1264682

13.24

1284.33

49.34

10.19

188.65

Standard deviation

1548528

2.609

1171.277

43.264

10.087

213.995

Skewness

2.155

−1.072

4.432

1.757

1.790

5.389

Range

7686825

11.2

11142

234.8

51.5

1811

Minimum

2840

6.8

150

1.2

0.5

39

Maximum

7689665

18

11292

236

52

1850

Confidence level (95.0 %)

238035.6

0.4

180.0

6.7

1.6

32.9

We can observe in Table 4 that max berth depth has a negative skewness and the rest of the continuous variables have a positive skewness, which implies that, in relation to the overall range exhibited by the sample, most terminals fall into the category of relatively small terminal size when measured in relation to quay length, yard space, number of Gantry, and crane spacing. In the next two sections we examine the model specification and results of the analysis, first by looking at the statistical estimation of our models and then by interpreting the results on terminals according to our hypotheses.

3 Model specification and estimation

Given the characteristics of the data and our objectives, Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis (SFA) are two appropriate methods for testing our hypotheses. Both techniques allow for the derivation of relative efficiency ratios within a group of analyzed units, therefore the efficiency of the units/terminals is compared through an efficient envelopment technique (a comprehensive survey can be found in Lovell 1993; Greene 1993; Hjalmarsson et al. 1996; Aigner et al. 1997; Kunbhakar and Lovell 2000; Coelli et al. 2005). DEA and SFA are well-established methodologies in maritime economics generally and port studies in particular (Notteboom et al. 2000; Cullinane and Song 2003; Barros and Athanassiou 2004; Tongzon and Heng 2005; Barros 2005; Cullinane et al. 2006; Trujillo and Tovar 2007).

Due to the heterogeneity of the data in our sample, and the longevity of the maritime technological characteristics under scrutiny, Stochastic Frontier Analysis is in our case the preferable methodology because, as suggested by Hjalmarsson et al. (1996), it allows for a rich specification of the production technology (functional form). The advantages of the SFA approach are that it reveals information about the production technique, it distinguishes between different variables’ roles affecting the output, and it assesses the relative efficiency of individual units, i.e., terminals (Battese and Coelli 1995). Moreover, with the introduction of statistical noise, SFA considers explicitly in the model all aspects that the functional form cannot explain (Coelli and Perelman 1999). It is thus possible to test the validity of our hypotheses and to model the effects of the two specific variables: terminal type and operation type. But we nevertheless need to impose an a priori structure when constructing the frontier functional form in SFA; and our assumptions concerning the distribution of the inefficiency term will have to be imposed in order to decompose the error. However, in our case the advantages of SFA still outweigh its disadvantages.

To test our hypotheses in relation to the three factors: terminal type, operation type, scale efficiency/degree of returns to scale, we structure 22 efficiency model specifications in Table 5. The four different factor-parameter specifications are listed in the left-hand column of the table. The first row identifies the models with five physical (continuous) inputs (basic model). In the second and third rows we specify the five physical inputs and one binary variable, that is, terminal type in row two, and operation type in row three. The fourth row identifies the specification of the model with five physical inputs plus terminal type and operation type.
Table 5

Summary of the models

Factor parameters

Model specification

Cobb-Douglas

Translog

Net effect model

Gross effect model

Net effect model

Gross effect model

Half Normal

Truncated Normal

Truncated Normal

Half Normal

Truncated Normal

Truncated Normal

5 continuous inputs (basic model)

Model 1.1

Model 1.2

 

Model 1.4

Model 1.5

 

5 continuous inputs, and terminal type

Model 2.1

Model 2.2

Model 2.3

Model 2.4

Model 2.5

Model 2.6

5 continuous inputs, and operation type

Model 3.1

Model 3.2

Model 3.3

Model 3.4

Model 3.5

Model 3.6

5 continuous inputs, and terminal type, operation type

Model 4.1

Model 4.2

Model 4.3

Model 4.4

Model 4.5

Model 4.6

The rows in Table 5 designate the variables used in each model, whereas the columns represent three assumption categories about our model specification. The first assumption category is the functional form used for the deterministic part of the model. We consider two forms: Cobb-Douglas and Translog. The second category relates to the net effect models and gross effect models; we apply the Battese and Coelli specifications (1992, 1995). In the net effect models the variables terminal type and operation type are in the deterministic part of the models, and in the gross effect models these variables are in the random part. The difference between net effect models and gross effect models shows how terminal type and operation type affect the production technique and efficiency. In other words, the net effect models account for the impact of the two variables in the production technique, and consequently the impact on production efficiency, while the gross effect models account only for the impact of these two variables on production efficiency, because they do not affect the production technique. The third assumption category in Table 5 illustrates the distribution assumption imposed on the random term (Half Normal or Truncated Normal). Given our cross-sectional data, we use the Half Normal and Truncated Normal distribution where the Half Normal distribution is a special case of Truncated Normal distribution.

The descriptive statistics of terminal-level data in Table 4 show that the distribution of the variable maximum berth depth has a negative skewness. Thus, in relation to the two considered functional forms, we can depict in Fig. 1 the histograms of the variables y (output), ×1 (max berth depth with negative skewed input) and ×4 (number of gantry with positive skewed input). The horizontal axes in Fig. 1 represent the value of the variable broken into intervals, and the vertical axes depict the number (frequency) of observations occurring within the fixed intervals. The logarithm values of the three variables are shown because we use the functional forms Cobb-Douglas and Translog in our models, and both employ the natural logarithm of the original value. We can see that the logarithm function changes positive skewness to negative but it does not change the negative skewness significantly. We know that by evaluating the logarithm of the continuous variables, a negatively skewed distribution is produced. Nevertheless, the distribution is satisfactory for the application of the Cobb-Douglas and Translog functional forms.
https://static-content.springer.com/image/art%3A10.1007%2Fs11123-012-0291-1/MediaObjects/11123_2012_291_Fig1_HTML.gif
Fig. 1

Histograms of selected variables and their logged values

As an additional clarification of the methodology, we can show the model specification examples for both Cobb-Douglas and Translog for the deterministic part along with the three different error distribution assumptions.

Model 4.1 in Table 5 is specified as
$$ \ln y = \alpha_{0} + \sum\limits_{n = 1}^{5} {\alpha_{n} \ln x_{n} } + \delta_{1} z_{1} + \delta_{2} z_{2} + v_{n} - u_{n} \quad {\text{n}} = 1, 2, \ldots, 5 $$
(1)
where y is output; xn are inputs, n = 1, 2‚…, 5; z1 and z2 are environmental variables; α0, αn, δ1, and δ2 are model parameters; vit are random errors and assumed i.i.d., N(0, σv2); un are non-negative random variables and assumed i.i.d., |N (0, σu2)|.
Model 4.5 in Table 5 is specified as
$$ \begin{aligned} \ln y_{t} & = \alpha_{0} + \sum\limits_{n = 1}^{5} {\alpha_{n} \ln x_{nt} } + \frac{1}{2}\sum\limits_{n = 1}^{5} {\sum\limits_{m = 1}^{5} {\alpha_{nm} } } \ln x_{nt} \ln x_{mt} + \delta_{1} z_{1} + \delta_{2} z_{2} + v_{nt} - u_{nt} \\ {\text{u}}_{\text{nt}} & = {\text{u}}_{\text{n}} {\text{exp}}(- \eta ( {\text{t}} - {\text{T))}} \\ {\text{n, m}} & = 1 , 2 ,\ldots , 5\\ {\text{t}} & = 1 , 2 ,\ldots ,{\text{T}} \\ \end{aligned} $$
(2)
where yt is output; xnt and xmt are inputs, n, m = 1, 2, …, 5; z1 and z2 are environmental variables; α0, αn, αnm, δ1, and δ2 are model parameters; vit are random errors and assumed i.i.d., N(0, σv2); un are non-negative random variables and assumed i.i.d., as truncations at zero of the N (μn, σu2); η is a scalar parameter; t = T = 1 since we consider cross-sectional data for year 2006.
Model 4.6 in Table 5 is specified as
$$ \begin{aligned} \ln y_{t} & = \alpha_{0} + \sum\limits_{n = 1}^{5} {\alpha_{n} \ln x_{nt} } + \frac{1}{2}\sum\limits_{n = 1}^{5} {\sum\limits_{m = 1}^{5} {\alpha_{nm} } } \ln x_{nt} \ln x_{mt} + v_{nt} - u_{nt} \\ u_{nt} & \sim {\text{N(m}}_{\text{nt}} , { }\sigma_{\text{u}}^{ 2} ) { } \\ m_{nt} & = \delta_{0} + z_{1} \delta_{1} + z_{2} \delta_{2} \\ n,m & = 1,2, \ldots ,5 \\ t & = 1,2, \ldots ,T \\ \end{aligned} $$
(3)
where yt is output; xnt and xmt are inputs, n, m = 1, 2, …, 5; z1 and z2 are environmental variables; α0, αn, αnm, δ0, δ1, and δ2 are model parameters; vit are random errors and assumed i.i.d., N(0, σv2); un are non-negative random variables and assumed i.i.d., as truncations at zero of the N (mnt, σu2); t = T = 1 since we consider cross-sectional data for year 2006.

The models are estimated using the Maximum Likelihood Estimation method. The Likelihood Ratio (LR) test is used as the decision rule to compare model performance.

4 Analysis of the results

In the evaluation of the models the Likelihood Ratio (LR) test suggests that Translog models are preferable to the Cobb-Douglas functional form (the parameter estimation for all the considered models can be found in Appendix 1: Translog models, and Appendix 2: Cobb-Douglas models). When we compare net effect models and gross effect models, the latter are preferable for both Cobb-Douglas and Translog models, according to the LR test. The estimation results confirm that terminal type and operation type variables influence terminal efficiency directly rather than through the production technique. Moreover, we can conclude that Model 2.6 (Table 5) is the best performing model. The parameter crane spacing has a negative sign in the Cobb-Douglas specification (Model 2.3) and a positive sign in the Translog model (Model 2.6). Crane spacing indicates the density of the container handling machines and reflects the usage of available space, where the higher the usage and the lower crane spacing is, the better; crane spacing also reflects the potential for extending handling capacity and thereby attracting future container traffic within a relatively short period of time. Translog allows for the calculation of interaction between variables, but Cobb-Douglas does not, so it is not clear whether the change of sign is due to the nature of the variable or to the choice of functional form. It is nevertheless reasonable to assume that both are significant and that the change of sign indicates that this variable requires more sophisticated modeling, which, however, is beyond the scope of this paper.

The signs of parameters for other inputs are consistent between Cobb-Douglas and Translog models and between the net effect and gross effect models. All are positive in both Cobb-Douglas and Translog models. The variables quay length, yard space and number of gantry cranes, as expected, have a positive sign, since bigger terminals are generally assumed to handle greater amounts of throughput. In contrast, the parameter max berth depth was expected to exhibit an unstable parameter sign. The reason for this presumption is that when berth depth exceeds the requirement for ships, it is not important if the water happens to be deeper. Berth depth in our data set ranges between 6.8 and 18 meters; fully loaded super-Panamax container ships have a draft of 12 meters (Post-Panamax containerships were launched in 2006 and their draft is 15.2 meters) and would be likely to encounter problems only in shallow water. However, the parameter for berth depth is positive in all the models, implying that the deep water feature of the terminals contributes to the attraction of container traffic. We next examine the obtained results in light of our hypotheses in order to verify their validity.

Factor 1: Terminal type (container/multi-purpose terminal)

In the net effect models (Models 2.4 and 2.5, Table 5) the sign of the parameter terminal type is negative, while in the gross effect model (Model 2.6, Table 5) the sign is positive. This opposite parameter sign indicates a result consistent with our hypotheses; in the net effect models the factor terminal type is specified in the deterministic part as an ‘input’, while in the gross effect models the factor terminal type is specified in the random inefficiency term. Terminal type as an input in the net effect models contributes to the output positively, whereas in the gross effect model terminal type as part of the inefficiency term contributes to the output negatively. In our case, the result shows that container-only terminals are more productive than multi-purpose terminals and thus proves our hypothesis.

Factor 2: Operation type (global/local operator)

In both of the net effect models (Models 2.4 and 2.5, Table 5) and in the gross effect model (Model 2.6, Table 5) the sign of the parameter operation type is positive. As operation type has a positive sign in both net and gross models, we must reject our hypothesis that global container terminal operators are more efficient than local operators.

On the basis of the results we can conclude that the terminal type factor improves model performance significantly, whereas the operation type factor does not. In the next sections we examine the third factor of our analysis: scale efficiency and returns to scale.

5 Scale efficiency and returns to scale

Based on the data set, we analyze a selected number of terminals in relation to total, scale and technical efficiency (Tables 6, 7, 8). Although the operator type is insignificant in our data, it is still listed in the efficiency indices tables, because as emphasized earlier, the emergence of global terminal operators is a profound institutional change in the development of container terminals. We calculate scale efficiency by following Ray’s procedure (Ray 1998, 2004) for Translog functional form, a well-established method in the economic literature (Banker 1984; Fare and Grosskopf 1985; Fare et al. 1994; Balk 2001); it is also easily applicable to our data set and our objective. Scale efficiency, technical efficiency and overall efficiency have been estimated for Model 2.6 (Table 5). We obtain an average scale efficiency of 0.89, average technical efficiency of 0.57 and average overall efficiency of 0.51 for the 165-terminal data set. The overall degree of returns to scale is 0.42. As such, by being smaller than 1, this result indicates that the container terminal industry is experiencing decreasing returns to scale at the current scale level.
Table 6

Top 10 terminals in total efficiency in Model 2.6

No.

Port

Terminal

Scale effi.

Tech effi.

Total effi.

Rank

Operator

76

Shenzhen

Chiwan Nanshan development group

1.00

0.90

0.90

(1)

Local

69

Shanghai

Shanghai East container terminal (Waigaoqiao phase 4)

1.00

0.89

0.89

(2)

Global

73

Shanghai

SIPG Zhendong container terminal (phase 2)

0.97

0.92

0.89

(3)

Local

72

Shanghai

Shanghai Shengdong international cont. term. phase 1

0.99

0.88

0.87

(4)

Global

107

Kaohsiung

Terminal 5 (APM terminals)

0.98

0.87

0.86

(5)

Global

101

Kaohsiung

Terminal 3 (APL: 68/69)

1.00

0.86

0.86

(6)

Global

85

Tianjin/Xingang

Number 2 container terminal

0.94

0.91

0.85

(7)

Local

63

Ningbo

Beilun no. 2 container company

0.94

0.89

0.84

(8)

Local

70

Shanghai

Shanghai mindong cont. term. (Waigaoqiao phase 5)

0.98

0.85

0.83

(9)

Global

151

Singapore

Tanjong Pagar (PSA)

0.99

0.83

0.82

(10)

Local

Table 7

Top 10 terminals in scale efficiency in Model 2.6

No.

Port

Terminal

Scale Effi.

Rank

Tech Effi.

Total Effi.

Operator

84

Tianjin/Xingang

CSX orient (Tianjin) terminals

1.00

(1)

0.65

0.65

Global

58

Hong Kong

Modern terminals (Kwai Chung)

1.00

(2)

0.78

0.78

Global

95

Busan

Pusan East container terminal (PECT)

1.00

(3)

0.77

0.77

Local

1

Rijeka

Brajdica container terminal

1.00

(4)

0.17

0.17

Global

156

Long Beach

Pier J Berths J232–234 (Int. transport service, K Line)

1.00

(5)

0.55

0.55

Global

28

Valetta

Valetta gateway terminal

1.00

(6)

0.10

0.10

Local

153

Long Beach

Pier C Berths C60–C62 (Matson)

1.00

(7)

0.58

0.58

Global

69

Shanghai

Shanghai East container terminal (Waigaoqiao phase 4)

1.00

(8)

0.89

0.89

Global

88

Busan

Dongbu busan cont. term. (Singamman term, Evergreen)

1.00

(9)

0.80

0.80

Global

37

Cadiz

Reina Sofia

1.00

(10)

0.29

0.29

Local

Table 8

Top 10 terminals in technical efficiency in Model 2.6

No.

Port

Terminal

Scale effi.

Tech effi.

Rank

Total effi.

Operator

73

Shanghai

SIPG Zhendong container terminal (phase 2)

0.97

0.92

(1)

0.89

Local

46

Izmir

Container Berths (13–16/17–19)

0.79

0.91

(2)

0.73

Local

85

Tianjin/Xingang

Number 2 container terminal

0.94

0.91

(3)

0.85

Local

76

Shenzhen

Chiwan Nanshan development group

1.00

0.90

(4)

0.90

Local

63

Ningbo

Beilun no. 2 container company

0.94

0.89

(5)

0.84

Local

69

Shanghai

Shanghai East container terminal (Waigaoqiao phase 4)

1.00

0.89

(6)

0.89

Global

72

Shanghai

Shanghai Shengdong international cont. term. phase 1

0.99

0.88

(7)

0.87

Global

82

Shenzhen

Yantian international container term (phase 1, 2 & 3)

0.78

0.88

(8)

0.69

Global

107

Kaohsiung

Terminal 5 (APM terminals)

0.98

0.87

(9)

0.86

Global

91

Busan

Gamman Hutchison cont. term. (ex Gamman Hyundai BGCT)

0.86

0.87

(10)

0.75

Global

In the total efficiency index shown in Table 6, five of the top 10 terminals are operated by global terminal operators and five are run by local operators. We can observe that in the scale efficiency index (Table 7), seven of the top 10 terminals are operated by global terminal operators; and in the technical efficiency index (Table 8), five terminals are operated by global operators.

The emergence of the global container terminal operator is indeed a phenomenon that cannot be overlooked (Fremont 2007). Over the past decade the top global terminal operators have increased their market share significantly, and for instance by 2005, the five biggest operators were handling 28 % of the world containers (Table 9). Nevertheless, we can also notice through our analysis that global terminal operators do not out-perform their local counterparts in the global context, and we therefore conclude that when considering the best performing terminals in relation to efficiency, we cannot state categorically that global terminals are more efficient than local ones. Terminal efficiency is affected by many contextual factors, e.g., regulation and market trends, to name two. Consequently, the emergence of the global container terminal operators has been driven by a variety of factors other than efficiency, such as improved access to capital, greater bargaining power and reputation with shippers, and the influence of principal investment funds seeking to acquire and consolidate their assets.
Table 9

Top 10 global terminal operators by 2005 throughput

Ranking

Operator

Million TEU

% Share

1

Hutchison Port Holdings (HPH)

33.2

8.3

2

PSA—Singapore Port Authority

32.4

8.1

3

APM terminals

24.1

6.0

4

P&O ports

21.9

3.3

5

DP world

13.3

2.5

6

Evergreen

11.5

1.7

7

Eurogate

11.4

1.6

8

COSCO

8.1

1.5

9

SSA marine

6.7

1.4

10

HHLA

5.7

1.3

Source Drewry shipping consultants, annual review of global terminal operators (2006)

When we consider the location of the selected terminals in the total efficiency index (Table 6), the top 10 terminals are located in the Far East; in the scale efficiency index (Table 7) among the top 10 terminals, five are situated in the Far East, three are located in the Mediterranean Basin and two are in North America; and in the technical efficiency index (Table 8), nine of the top 10 terminals are located in the Far East and one is situated in the Mediterranean Basin.

Terminals in the Far East (mainly P.R. China) dominate the class of the best performing terminals in relation to efficiency indices; this implies that geographic location of terminals plays an important role. As we have remarked in the introduction, some literature shows that the location of a terminal has become less important but our analysis reaches the conclusion that terminal location is correlated with the efficiency of the terminals. The reasons for our results may be twofold, both of which correspond to the significant scale of throughput movement in Asian terminals: firstly, the Far East and especially China, is a major regional production hub as well as a primary driver of international merchandise trade. Secondly, the Far East deploys logistics services and hinterland connection capabilities which are integrated into the global supply chain network.

In the case of the specific region of the Mediterranean Basin (Tables 10, 11, 12), global terminal operators are dominant in the area. In fact, in the total efficiency index, seven of the top 10 Mediterranean terminals are operated by global terminal operators; in scale and technical efficiency indices (Tables 11, 12) six terminals in their respective rankings are operated by global terminal operators.
Table 10

Top 10 Mediterranean terminals in total efficiency in Model 2.6

No.

Port

Terminal

Scale effi.

Tech effi.

Total effi.

Rank

Operator

29

Algeciras

Terminal 2000 (APM terminals)

0.99

0.76

0.75

(35)

Global

46

Izmir

Container Berths (13–16/17–19)

0.79

0.91

0.73

(40)

Local

47

Mersin

2 container quays

0.94

0.75

0.71

(44)

Global

22

Salerno

Salerno container terminal (SCT)

1.00

0.67

0.67

(56)

Global

44

Valencia

Valencia container terminal (Principe Felipe quay)

0.97

0.69

0.66

(58)

Global

34

Barcelona

TerCat

0.89

0.69

0.62

(68)

Global

16

La Spezia

Terminal de Golfo

0.98

0.63

0.62

(70)

Local

14

Gioia Tauro

Medcenter container terminal

0.78

0.77

0.60

(74)

Global

9

Thessaloniki

Pier 6

0.99

0.59

0.59

(76)

Local

15

La Spezia

La Spezia cont. term. (Molo Fornelli Berths 13–15/17–18)

1.00

0.57

0.57

(81)

Global

Table 11

Top 10 Mediterranean terminals in scale efficiency in Model 2.6

No.

Port

Terminal

Scale effi.

Rank

Tech effi.

Total effi.

Operator

1

Rijeka

Brajdica container terminal

1.00

(4)

0.17

0.17

Global

28

Valetta

Valetta gateway terminal

1.00

(6)

0.10

0.10

Local

37

Cadiz

Reina Sofia

1.00

(10)

0.29

0.29

Local

4

Marseilles-Fos

Fos container terminal–Seayard

1.00

(15)

0.15

0.15

Global

36

Barcelona

UTE Llevant

1.00

(17)

0.22

0.22

Global

19

Ravenna

Setramar terminal

1.00

(21)

0.03

0.03

Local

22

Salerno

Salerno container terminal (SCT)

1.00

(22)

0.67

0.67

Global

40

Tarragona

Tarragona container terminal (Moll D’ Andalusia)

1.00

(26)

0.05

0.05

Global

15

La Spezia

La Spezia cont. term. (Molo Fornelli Berths 13–15/17–18)

1.00

(31)

0.57

0.57

Global

21

Salerno

Other Berths

0.99

(37)

0.30

0.30

Local

Table 12

Top 10 Mediterranean terminals in technical efficiency in Model 2.6

No.

Port

Terminal

Scale Effi.

Tech effi.

Rank

Total effi.

Operator

46

Izmir

Container Berths (13–16/17–19)

0.79

0.91

(2)

0.73

Local

14

Gioia Tauro

Medcenter container terminal

0.78

0.77

(49)

0.60

Global

29

Algeciras

Terminal 2000 (APM terminals)

0.99

0.76

(51)

0.75

Global

8

Piraeus

Venizelos container terminal (Pier II)

0.61

0.75

(53)

0.46

Local

47

Mersin

2 container quays

0.94

0.75

(56)

0.71

Global

39

Seville

Muelle de Centenario

0.71

0.74

(62)

0.52

Local

18

Naples

Molo Bausan terminal (CoNaTeCo)

0.80

0.70

(70)

0.56

Global

34

Barcelona

TerCat

0.89

0.69

(72)

0.62

Global

17

Naples

Flavio Gioia terminal

0.40

0.69

(73)

0.28

Local

44

Valencia

Valencia container terminal (Principe Felipe quay)

0.97

0.69

(75)

0.66

Global

Global terminal operators are predominant in the Mediterranean Basin and this is certainly related to the European context and to the EU framework (EU policies, agreements and regulations), since interaction among terminal operators is highly incentive-driven. The EU has implemented different EU-wide maritime policies in the past two decades to stimulate the competitive provision of services in EU ports. For instance, Motorways of the Sea, TEN-T and Marco Polo I and II programs (European Commission 2009) have added significantly to the prospect of increasing container traffic volumes in the Basin, thus prompting global operators, driven by the potential for long-term growth, to acquire terminal operations and thereby serve a burgeoning market demand. Moreover, individual EU states have accelerated this process through their pursuit of privatization programs in order to receive a share of the profitability attainable through private sector ownership and investment in terminal expansion (Notteboom and Winkelmans 2001).

Difficulties have arisen, however, in relation to labor unions, port authorities and terminal service providers which have at times impeded the full implementation of some of these policies on container terminals. As Barros and Peypoch (2007) notice, “investment is not matched by upgraded managerial practices,” thus showing a certain level of rigidity in the management and organization of the ports and, in certain cases, of the terminals.

Nonetheless, the container handling business has demonstrated that private operators have been keen to capture the benefits of consolidation across the EU. For example, global terminal operators such as Hong Kong’s Hutchison Port Holdings (HPH), PSA corporation and P&O ports have developed an EU-wide network to include their presence in the Mediterranean, Hamburg-Le Havre range and the UK; their strategic decisions have played an essential role due to the importance of the Mediterranean as both transshipment and gateway zone on the Europe-Asia trade route. It is therefore unsurprising in our analysis to find high concentrations of global terminal operators in the Mediterranean Basin.

6 Improving scale efficiency by input level

Scale efficiency provides the measurement of the relative production ratio of a terminal’s current input level compared to its optimal level. However, it neither shows whether a terminal is oversized or undersized, nor does it indicate how much the input level should be adjusted to obtain the optimal scale. The degree of returns to scale for a particular terminal allows us to determine the level of the terminal size and show how to obtain its optimal scale (Fare et al. 1994; Battese and Coelli 1995). The objective in this section is to identify general strategies that correspond to terminal container investments.

In order to demonstrate how to improve scale efficiency (SE) by changing the input level, it is necessary to consider the concept of scale factor t*. The SE is calculated through the comparison of the current size with the Most Productive Scale Size (MPSS); scale factor t* is used in order to estimate the MPSS. The current observed size (input level) of the terminal is set to 1, and the optimal size (MPSS) is represented by t*. Using our data set, we assume that the terminal input mix is constant and we examine three cases to show how SE may be improved by the information given by t*. Each example represents a typical scale status, increasing, decreasing and constant returns to scale.

6.1 Increasing returns to scale (IRTS) case

The terminal of the Port ANTALYA in Turkey exhibits t* = 1.91, SE = 0.50, TE = 0.31, and overall efficiency of 0.15. In Fig. 2 the vertical axis represents output in TEU and the horizontal axis represents size of terminal, depicting 1 as its current size. Point A on the vertical line is the actual observation point for Antalya, and the darker curve is the production technique frontier that Antalya could achieve for its particular input mix (combination). Point B, where the vertical line and production frontier meet, represents the technical optimal that Antalya could achieve at its current input level.
https://static-content.springer.com/image/art%3A10.1007%2Fs11123-012-0291-1/MediaObjects/11123_2012_291_Fig2_HTML.gif
Fig. 2

Terminal at increasing returns to scale level

Tangent point C, where the tangent meets the frontier, represents the optimal scale that Antalya could achieve for its current input mix. For this case the technical optimal point and the scale optimal point are not the same. Therefore, two sources of inefficiency exist for this terminal, namely, technical inefficiency and scale inefficiency. The technical efficiency of 0.31 is measured as the relative distance between observation point A and technical optimal point B. Scale efficiency is 0.50 and is measured by the difference between the slope of technical optimal point B to origin point and the slope of scale optimal point C to origin point. In this case the terminal is experiencing increasing returns to scale; the implication here is that terminal Antalya needs to expand from its current operating size (or input level) to 1.9, the level at point C, in order to obtain its optimal scale.

6.2 Decreasing returns to scale (DRTS) case

Medcenter Container Terminal at Port GIOIA TAURO in Italy expresses t* = 0.68, SE = 0.78, TE = 0.77 and overall efficiency is equal to 0.60. Point A in Fig. 3 on the vertical line is the actual observation point for Medcenter. The darker curve depicts the production technique frontier that Medcenter could achieve for its particular input mix. Point B, where the vertical line and the production frontier meet, represents the technical optimal that Medcenter could achieve at its current input level.
https://static-content.springer.com/image/art%3A10.1007%2Fs11123-012-0291-1/MediaObjects/11123_2012_291_Fig3_HTML.gif
Fig. 3

Terminal at decreasing returns to scale level

Tangent point C, where the tangent meets the frontier, represents the optimal scale that Medcenter could achieve for its current input mix. In this example the technical optimal point and the scale optimal point are also not the same, so in this case the terminal shows two sources of inefficiency. Technical efficiency, measured as the relative distance between observation point A and the technical optimal point (B) is 0.77. Scale efficiency is 0.80 and is measured by the slope difference between technical optimal point B and scale optimal point C. In this case Medcenter is experiencing decreasing returns to scale and needs to reduce its size to the level 0.68 from its current level in order to obtain optimal scale size C.

6.3 Constant returns to scale (CRTS) case

The Brajdica Container Terminal at the Port RIJEKA in Croatia exhibits t* = 1.01, SE = 1.00, TE = 17 and an overall efficiency of 0.17. Also in this case, as in the previous two cases, the horizontal axis represents size of terminal and the vertical axis represents output in TEU. Point A in Fig. 4 on the vertical line is the actual observation point for this terminal; the darker curve is the production technique frontier that Brajdica could achieve for its particular input mix. Point B, where the vertical line and the frontier meet, represents the technical optimal that Brajdica could achieve at its current input level.
https://static-content.springer.com/image/art%3A10.1007%2Fs11123-012-0291-1/MediaObjects/11123_2012_291_Fig4_HTML.gif
Fig. 4

Terminal at constant returns to scale level

Tangent point C, where the tangent meets the frontier, denotes the optimal scale that Brajdica could achieve for its current input mix. In this case the vertical line, tangent and frontier meet at the same point (B), indicating that Brajdica is already at its optimal scale, with a scale efficiency score of 100 %. Moreover, Brajdica is experiencing constant returns to scale. The technical efficiency is measured by the relative distance of the observation point to its technical optimal point on the frontier. In this case technical efficiency is quite low. Brajdica therefore needs to improve its production technique in order to improve efficiency.

An implicit condition is applied to all the efficiency analyses in this section, that is, the input mix is constant. We have demonstrated that by using the information given by t*, and given a particular input mix, we can improve the SE of terminals. More generally, it is noteworthy that, although this type of analysis is useful for policy makers because it signals trends for future development, our results must be examined and interpreted with caution. As observed by Banker and Thrall (1992), scale efficiency information should be studied together with the size elasticity of the scale efficiency in order to clarify how to effectively change the input level of a terminal to obtain its optimal scale. Moreover, since we examine the variables infrastructure and machinery of container terminals in this study, it is necessary to account for the long-term cycle of this type of investment. We may consider improving scale efficiency by means of other strategies, for example, by adjusting input mix (combination); although, this is not within the scope of the present analysis, it nevertheless represents an interesting prospect for further research.

7 Conclusion

This paper has provided a study of the main developments of maritime container terminal operations. We have analyzed two operational characteristics: terminal type and operation type, which we assume to influence the technical efficiency of terminals. We have demonstrated that container terminals are more efficient than multi-purpose terminals, and in our second finding we have shown that, compared with local operators, global terminal operators do not have a dominant position in international maritime trade in terms of productivity and efficiency. However, global terminal operators appear to be more dominant than local operators when we examine the Mediterranean Basin. This observation gains further relevance in light of the discussion on the competition among Mediterranean terminals (Gonzalez and Trujillo 2009; Bergantino and Musso 2011). Based on the results pertaining to terminal operations, we have shown that both terminal type and operation type impact on terminal efficiency directly, rather than through the production technique. As a consequence of this finding, future terminal development and strategies may be to focus on the implementation of advanced logistical and operational information technologies in terminal operations.

We have evaluated the infrastructure efficiency for 165 terminals worldwide by carrying out a terminal-specific assessment of technical and scale efficiency in which we have concluded how much efficiency a terminal may gain, given its current production technique and input (infrastructure) level. We have demonstrated how and whether a given terminal may increase or decrease its current input level, and shown where to target investment in order for it to achieve its optimal size. This result is significant in relation to future trends in terminal operations because not only have we identified an analytical tool that allows us to assess infrastructure investment, but we have also provided a way to benchmark investment sustainability against demand (UNCTAD 2010).

Footnotes
1

The classification of the world region is the following: North America, North Europe, South Europe, Far East, South East Asia, Middle East, Caribbean, Central America, South America, Oceania, South Asia, Africa, and Eastern Europe (Drewry Shipping Consultants 2008).

 
2

The port of Tanjung Pelepas in Indonesia, which is usually included in the set of the top container ports worldwide, is not considered in this analysis because we do not have satisfactory data on its terminals.

 

Copyright information

© Springer Science+Business Media, LLC 2012