Technical Efficiency of Container Terminal Operations: a Dea Approach

: Nowadays, transporting cargoes via container are key indicator for every shipment. The movement of container involves multi modes to reach destination. The efficient transport networking systems are determinant attribute towards container terminal in providing excellent services to their client. The paper focuses on the metamorphosis of the terminal efficiency and container movements at 6 major container terminals in Peninsular Malaysia. The aim is to measure efficiency of container termi - nals that contributes significant economic development for a nation. Non parametric approach under frontier method is used to analyse panel data from 2003 to 2010 in relation with container terminal equipments and throughput. The result shows no significant relationship between container terminal size and efficiency. Thus, efficiency is determined from allocation of resources efficiently by terminal operators and not by size of terminals.


INTRODUCTION
Since the invention of container by Malcom Mclean late 1950s, and the first international shipment in 1966, the shipments of goods have changed drastically (Levinson, 2006).In addition, containerisation is applied to all modes of transport such as rail, container vessel and haulage.The handling process of moving of goods continuously improved, and it has benefited to all parties.Containerisation and the development of intermodal transports system have had a profound effect on the shipping industry, its structure, management and operation.The movement of goods in a single container by more than one mode of transportation was an important development in the transportation industry and all the elements involved for the international and domestic trade.Classically, the terms 'Through Transport', 'Combined transport', 'Intermodal transport', and 'Multimodal transport' are preferable for movement of goods.It started from the point of origin to point of destination.These four terms have very similar meaning, where the movements of goods are involved with more than one mode to ship the cargo (UNCTAD, 1993;2001).Multimodality or intermodality has given tremendous impact to the transport industry (Hayuth, 1987;Hariharan, 2002;Levinson, 2006).
Intermodality is defined as the movement of cargo from shipper to consignee by at least two different modes of transport under a single rate, throughbilling and through-liability (Hayuth, 1987).Multimodal transport refers to a transport system usually operated by one carrier with more than one mode of transport under the control or ownership of one operator.It involves the use of more than one means of transport i.e., truck, railcar, aeroplane or ship in succession to each other e.g. a container line which operates both a ship and a rail system of double stack trains (UNCTAD, 1993;2001).The objective of these concepts is to transport goods from point of origin to point of final destination in the most cost and time effective.Therefore, to achieve the objective of multimodalism, intensive cooperation and coordination among transportation modes are essential.The paper studies the container terminal efficiency from where transportation network systems generate container from and to container terminal.
The study covers 6 major container terminals in Peninsular Malaysia.The non parameter technique under frontier method called as data envelopment analysis (DEA) is used to analyse panel data from 2003 to 2010.The first section starts with introduction and follows with theoretical perspective on transportation systems in section 2. Under section 3, discussion on the efficiency technique and DEA model is developed for the research.Section 4 discusses DEA that has been applied at container terminal.Furthermore, the model is applied for this research to analyse the panel data.Section 5 represents results and discussion on the analysis from DEA-CCR and DEA-BCC output-oriented.In Section 6 represents conclusion on the research.

NETWORK
Back in 1955, delivery process has been changed when Malcom Mclean introduced standardised container box (UNCTAD, 1993;2001;Levinson, 2006).The first shipment by using container took place in Newark, New Jersey USA where shipment of cargoes to Puerto Rico of a Sea-Land vessel.However, Sea-Land international maiden only happened in 1966 because of confrontations with shipping lines (Talley, 2000;Levinson, 2006).First international called for Sea-Land was to Rotterdam, and since that time; the new era of shipping industry has emerged with the international trade via container.The container revolution has been improved with the general introduction of twenty footer and forty footer standardised container or International Organisation for Standardisation (ISO) container.The invention of containerised cargo means it is able to load and be secured on a truck chassis, a rail car, or in vessel's hole or deck.
Generally, intermodalism terminology is being since 1920s, however intermodalism freight transportation officially used in 1985 (UNCTAD, 1993;2001).In addition, multimodal transport was officially introduced in 1980 when United Nation sponsored Multimodal Convention, the term attained legal recognition on 1 st January 1992 when 1992 UNCTAD/ ICC Rules for Multimodal Transport was launched (UNCTAD, 1993;2001).Since then, the movement of container from point of origin to point of destination by using different type of mode became commercially feasible to the industry.
An efficient and good road networks are main catalyst for movement of good via road (World Bank,  Hayuth (1987Hayuth ( , 1994) ) emphasis that integrated logistic network is important for movement of container via road.Figure 1 depicts road network in Peninsular Malaysia.The road network consists of expressway, federal and state road.In Peninsular Malaysia, total road network systems are 82144 kilometre (PWD, 2009).The road breakdowns are 61420 km Page | 2 introduced in 1980 when United Nation sponsored Multimodal Convention, the term attained legal recognition on 1 st January 1992 when 1992 UNCTAD/ICC Rules for Multimodal Transport was launched (UNCTAD, 1993;2001).Since then, the movement of container from point of origin to point of destination by using different type of mode became commercially feasible to the industry.
An efficient and good road networks are main catalyst for movement of good via road (World Bank, 2005).The road networks are accessible throughout Peninsular Malaysia and contribute significant towards state economic development.The accessibility has spurred development of container terminal in Peninsular Malaysia.Its locations are in Penang (Penang Port), Selangor (Westport and Northport), Johor (Johor Port and Tanjung Pelepas Port) and Pahang (Kuantan Port).Hayuth (1987Hayuth ( , 1994) ) emphasis that integrated logistic network is important for movement of container via road.Figure 1 depicts road network in Peninsular Malaysia.The road network consists of expressway, federal and state road.In Peninsular Malaysia, total road network systems are 82144 kilometre (PWD, 2009).The road breakdowns are 61420 km under state and municipality roads, 18904 km under federal roads and 1820 km are toll highways (PWD, 2009;Levinson and Zhu, 2011).Figure 2 depicts the impact and transformation from conventional to container on the containerised port system.It was manifested into two impacts which are spatial and organisational.With the introduction of container system, the port process has been changed drastically from the equipments, manpower, port system and port's charges.This transformation has classified terminal more organised even though the process under state and municipality roads, 18904 km under federal roads and 1820 km are toll highways (PWD, 2009;Levinson and Zhu, 2011).becoming more complex.However, by having an organised structure container terminal operation is able to handle efficiently.
Figure 2. The impact of containerisation on the conventional general cargo port system (Hayuth, 1987) A terminal involves a lot of parties from government agencies, shipping agents, forwarding agents, carriers, ship owners, container terminal operators and clients.Even a console good inside container will through similar process for documentation, handling rate, custom clearance, and shipment.Containerisation is the largest form of unitisation.Containers are loaded with products at the shipper's premises and sealed, and then they are carried over to the consignee's premises intact, without the content being taken out or re-packed en route.This is the essence of container transport as well as multimodal transport, but containerisation is not synonymous with multimodal transport.Containerisation contributes to a higher efficiency in the development of multimodal transport operations.The focus, now, is more on the organisation of the transport industry and the synchronisation of the integrated logistical system (Hayuth, 1987;Carrese and Tatarelli, 2011;Kasypi et al, 2013).In order to achieve multimodal transport, intensive co-operation and co-ordination among transport modes are essential.Kasypi and Shah (2012) establish the integration model of container terminal by applying IDEF0 with supply chain.The model integrates gate, yard, wharf and vessel components at container terminal in enhancing the operational activity.The IDEF0 lean supply chain model is a mapping process for movement of containers from and to vessel as well as gate in and out.The mapping model is able to monitor and execute operational process in maximising efficiency and productivity.Figure 3  tainer system, the port process has been changed drastically from the equipments, manpower, port system and port's charges.This transformation has classified terminal more organised even though the process becoming more complex.However, by having an organised structure container terminal operation is able to handle efficiently.A terminal involves a lot of parties from government agencies, shipping agents, forwarding agents, carriers, ship owners, container terminal operators and clients.Even a console good inside container will through similar process for documentation, handling rate, custom clearance, and shipment.Containerisation is the largest form of unitisation.Containers are loaded with products at the shipper's premises and sealed, and then they are carried over to the consignee's premises intact, without the content being taken out or re-packed en route.
This is the essence of container transport as well as multimodal transport, but containerisation is not synonymous with multimodal transport.Containerisation contributes to a higher efficiency in the development of multimodal transport operations.The focus, now, is more on the organisation of the transport industry and the synchronisation of the integrated logistical system (Hayuth, 1987;Carrese and Tatarelli, 2011;Kasypi et al, 2013).In order to achieve multimodal transport, intensive co-operation and co-ordination among transport modes are essential.(Kasypi and Shah, 2012) Figure 4 shows the impact of intermodal transport on the containerized port system.During those days, there are two phases of transformation of containerized port system.The first phase of port containerization involved a period of technological change and a massive growth in the spatial dimensions of terminals.For the second phase, its focuses attention on organizational aspects of international transport and the port industry i.e., marketing strategies, participation by ports in the physical distribution of cargo.Thus, in this phase the containerized port system is an integrated transport system   (Kasypi and Shah, 2012) Figure 4 shows the impact of intermodal transport on the containerized port system.During those days, there are two phases of transformation of containerized port system.The first phase of port containerization involved a period of technological change and a massive growth in the spatial dimensions of terminals.For the second phase, its focuses attention on organizational aspects of international transport and the port industry i.e., marketing strategies, participation by ports in the physical distribution of cargo.Thus, in this phase the containerized port system is an integrated transport system Figure 4.The impact of intermodal transport on the containerized port system.(Hayuth, 1987)  there are two phases of transformation of containerized port system.The first phase of port containerization involved a period of technological change and a massive growth in the spatial dimensions of terminals.For the second phase, its focuses attention on organizational aspects of international transport and the port industry i.e., marketing strategies, participation by ports in the physical distribution of cargo.Thus, in this phase the containerized port system is an integrated transport system

EFFICIENCY TECHNIQUE: DATA ENVELOPMENT ANALYSIS
Efficiency is derived and part of productivity, where it is a ratio of actual output attained to standard output expected (Sumanth, 1984).Mali (1978) express together the terms productivity, effectiveness and efficiency as follows: Productivity index = output obtained performance achieved effectiveness input expected resources consumed efficiency = = (1-0) Therefore, Sumanth (1984) and Ramanathan (2003) express efficiency as follow: The (2-0) equation is applicable for evaluation of simple data.The entity of output and input are diverse significantly.Therefore, equation (2-0) is not suitable for complex relationship between outputs and inputs.The weight cost approach is the solution for complexities of outputs and inputs as follows: weighted of outputs Efficiency weighted of inputs By assuming all weights are uniform, mathematically equation is expressed as follows: An efficient is denote = 1, therefore, to classify unit of efficiency is set as 0 < Efficiency ≤ 1.

Technical Efficiency: Data Envelopment Analysis
Technical efficiency (TE) is described as the conversion of physical inputs (such as the services of employees and machines) into outputs relative to best practice.In other words, given current technology, there is no wastage of inputs whatsoever in producing the given quantity of output.An organization operating at best practice is said to be 100% technically efficient.If operating below best practice levels, then the organization's technical efficiency is expressed as a percentage of best practice.Managerial practices and the scale or size of operations affect technical efficiency, which is based on engineering relationships but not on prices and costs.Data Envelopment Analysis (DEA), first introduced by Charnes, Cooper and Rhodes (CCR) in 1978 (Charnes et al, 1978), extended Farrell's (1957) idea of estimating technical efficiency with respect to a production frontier.The definition of efficiency is referred from the "Extended Pareto-Koopmans" and "Relative Efficiency" The CCR is able to calculate the relative technical efficiency of similar Decision Making Units (DMU) through the analysis, with the constant returns to scale basis.This is achieved by constructing the ratio of a weighted sum of outputs to a weighted sum of inputs, where the weights for both the inputs and outputs are selected so that the relative efficiencies of the DMUs are maximized with the constraint that no DMU can have a relative efficiency score greater than one.On the other hand, the DEA-BCC model (Banker et al., 1984) extend from DEA-CCR by assuming variable returns to scale where performance is bounded by a piece-wise linear frontier.There are other DEA models in the literature, but DEA-CCR and DEA-BCC are the most commonly used models.
There are numerous articles, journals and books published about DEA since 1978, with numerous extensions of the methodology and many novel applications (Seiford andThrall, 1990 andSeiford, 1994).Since the CCR (1978), the development has introduced the BCC model that is Banker, Charnes andCooper in 1984 (Barnes et al, 1984).The BCC model relaxes the convexity constraint imposed in the CCR model which allows for the efficiency measurement of DMUs on a variable returns to scale basis.The BCC model results in an aggregate measure of technical and scale efficiency, the CCR model is only capable of measuring technical efficiency.This allows for the separation of the two efficiency measures.
The scale efficiency measurement indicates whether a DMU is operating at the most efficient scale, while technical efficiency is a measure of how well the DMU is allocating its resources to maximize its output generation.It is important to note that the BCC model is both scale and translation invariant, while the CCR model is only scale variant.The development of the Additive model, which involves reduction of inputs with a simultaneous increase in outputs, and Multiplicative models note worthy ad-vances which, along with further explanations of the DEA technique and its extensions, are outlined in (Ali and Seiford, 1993, Charnes et al, 1994a, Charnes et al, 1994band Lovell, 1993).Since the first application of DEA for measuring the efficiency of business student to schools (Chanrnes et al, 1978) the technique has been applied in over 50 industries i.e., healthcare, transportation, hotel, education, computer industry etc.

Model Development
The model is developed from the extension of the ratio technique used in traditional efficiency approaches.The measurement is obtained from DMU as the maximum of a ratio weighted output to weighted input.The numbers of DMUs are not determined outputs and inputs, however, larger DMUs are able to capture higher performance.This would determine the efficiency frontier (Golany and Roll, 1989).In addition, the number of DMUs should be at least twice the number of inputs and outputs (Golany and Roll, 1989).The scale efficiency measurement indicates whether a DMU is operating at the most efficient scale, while technical efficiency is a measure of how well the DMU is allocating its resources to maximize its output generation.It is important to note that the BCC model is both scale and translation invariant, while the CCR model is only scale variant.The development of the Additive model, which involves reduction of inputs with a simultaneous increase in outputs, and Multiplicative models note worthy advances which, along with further explanations of the DEA technique and its extensions, are outlined in (Ali and Seiford, 1993, Charnes et al, 1994a, Charnes et al, 1994band Lovell, 1993).Since the first application of DEA for measuring the efficiency of business student to schools (Chanrnes et al, 1978) the technique has been applied in over 50 industries i.e., healthcare, transportation, hotel, education, computer industry etc.

Model Development
The model is developed from the extension of the ratio technique used in traditional efficiency approaches.The measurement is obtained from DMU as the maximum of a ratio weighted output to weighted input.The numbers of DMUs are not determined outputs and inputs, however, larger DMUs are able to capture higher performance.This would determine the efficiency frontier (Golany and Roll, 1989).In addition, the number of DMUs should be at least twice the number of inputs and outputs (Golany and Roll, 1989).
The parameters and variables are needed in developing the model.Therefore, the model is based on the following parameters and variables: N = number of DMU {j = 1,2,...n} y = number of outputs {y = 1,2,...R} x = number of inputs {x = 1,2,...S} y i = Quantity of output r th of output of j th DMU x i = Quantity of input s th of input of j th DMU u r = weight of r th output v s = weight of s th input . . Figure 5: DMU and Homogeneous units Golany and Roll (1989) describe that homogenous unit is important in choosing DMUs to be compared and identifying the factors affecting DMUs.Therefore, homogenous group of units need to perform similar task and objectives, under same set of market conditions and the factors (inputs and outputs).Figure 5 depicts the DMU and homogeneous units.This concept is using linear programming (LP) formulation to compare the relative efficiency of a set of decision making units (DMUs).Farrell (1957) has developed similar approach to compare the relative efficiency of a cross-section sample of agricultural farms.
The efficiency measures under constant returns to scale (CRS) are obtained by N linear programming problems under Charnes et al. 1978 as below:  Golany and Roll (1989) describe that homogenous unit is important in choosing DMUs to be compared and identifying the factors affecting DMUs.Therefore, homogenous group of units need to perform similar task and objectives, under same set of market conditions and the factors (inputs and outputs).
Figure 5 depicts the DMU and homogeneous units.
This concept is using linear programming (LP) formulation to compare the relative efficiency of a set of decision making units (DMUs).Farrell (1957) has developed similar approach to compare the relative efficiency of a cross-section sample of agricultural farms.
The efficiency measures under constant returns to scale (CRS) are obtained by N linear programming problems under Charnes et al. 1978 as below: ∑ thus, BCC is able to distinguish between technical and scale inefficiencies by (i) estimating pure technical efficiency at the given scale of operation and (ii) identifying whether increasing, decreasing or constant return to scale possibilities are present for further exploitation.It is called as variable return to scale.Therefore, for CCR efficient is required both scale and technical efficient, BCC efficient is only required technically efficient.

CONTAINER TERMINAL EFFICIENCY US-ING DATA ENVELOPMENT ANALYSIS
A firm's productivity is usually measured by comparing its actual production volume with a production frontier.Wang et al. (2005), productivity measurement can be classified into using a parametric frontier approach or a non-parametric frontier approach.In the parametric frontier approach, the productivity frontier is estimated in a particular functional form with constant parameters.Liu (1995) uses a stochastic parametric frontier approach on 25 world ports, whereas Estache et al. (2001) studies 14 Mexican ports in order to investigate the efficiencies gained after port reform.Other studies on port performance with a stochastic parametric frontier approach are Tongzon and Heng (2005), Cullinane and Song (2003), Cullinane et al. (2002) and Notteboom et al. (2000).Besides this, Coto-Millan et al. (2000) uses a stochastic cost function approach on 27 Spanish ports.De and Ghosh (2002) examined 12 Indian ports using a time-varying production function approach.On the other hand, the non-parametric frontier approach assumes no particular functional form for the frontier.The most commonly used non-parametric frontier technique is DEA.

Discussion of Input and Output
The research is using 6 container terminals in Peninsular Malaysia as DMU.The data used in this research is from the year 2003 to 2010.The presentation of results are base on general output oriented DEA-CCR and DEA BCC in obtaining efficiency score.The research is used DEA-Solver Pro 7 version for analysis of data for the model.Golany and Roll (1989) highlight that the number of DMUs should be at least twice the number of inputs and outputs for the homogeneity reason.In container terminal industry, the handling equipments for operation are varies from each others.In this case, it is the index approach is used for certain inputs to avoid homogeneity i.e., for quay crane;

Quay Crane's index = Number of quay cranes x average lifting capacity
We

Description of Slack
When a unit DMU is most efficient, the Performance Targets for inefficient can be set to ensure DMU reach 100% relative efficiency in comparison with DMU i .DMU can be set as benchmark, Input Target for DMU i is describe as follow

Input Target = Actual Input * Efficiency
However, for inefficient DMU, input target will be less than actual input.Hence the difference between actual input and input target is called input slack (Ramanathan, 2003;Mishra, 2012) Input Slack = Actual Input -Input Target.
In percentage;

RESULT AND DISCUSSION
Table 6 and 7 represent ranking score for efficient and inefficient DMUs.There are 19 DMU that represent efficient = 1, the other 29 DMUs are inefficient for DEA-CCR.The most inefficient DMU is FK03, in which represent inefficient of 0.607.In general, the bottom 3 of inefficient DMUs are FK04 (0.689) and FK05 (0.668).Rank 20 (FK10), 21(CP08) and EPP07 (0.976) are represent closely efficient for DMUs.The efficient DMUs are i.e., EPP10, AW03, CP10 etc.On the other hand, efficient DMUs for DEA-BCC are 25 and 15 are inefficient i.e., EPP10(1) and FK03(0.6).The inefficient DMUs means that between inputs and output, the utilisation of resources are not as maximum as possible, where there are improvement can be done by container terminal operators in achieving an efficient container terminal.Kasypi and Shah (2012) develop IDEF0 model for lean supply chain to expedite the terminal process flow.The lean supply chain process by using IDEF0 are able to evaluate and execute operational process.The IDEF0 model also used by NATO and Pentagon.(efficient) in which utilisation of all inputs and output are = 1.It shows that utilisation between inputs and output significantly = 1.The projection score is also efficient when technical efficient =1.It means, all resources allocated for that time are at maximum with the output that produces by container terminal.However, when technical efficient score is inefficient < 1, the projection score is greater than 1, when some of the inputs are not utilised (BN03-1.24).
On the other hand, Table 9 depicts technical efficiency and projection score DEA-BCC.The technical efficiency efficient for AW03(1).However, BN03 (0.80) inefficient for technical efficient and projection score is better than DEA-CCR at 1.12.The reason is DEA-BCC only requires technical efficient in determining the efficiency level rather than DEA-CCR in which, require both scale and technical efficiency to be efficient.

3
2005).The road networks are accessible throughout Peninsular Malaysia and contribute significant towards state economic development.The accessibility has spurred development of container terminal in Peninsular Malaysia.Its locations are in Penang (Penang Port), Selangor (Westport and Northport), Johor (Johor Port and Tanjung Pelepas Port) and Pahang (Kuantan Port).

Figure 1 .
Figure 1.Major Road Network in Peninsular Malaysia (PWD, 2009 Figure 2 depicts the impact and transformation from conventional to container on the containerised port system.It was manifested into two impacts which are spatial and organisational.With the introduction of con-

FigureFigure 4 Figure 3 .
Figure 3. IDEF0 Model for Container Terminal (Kasypi and Shah, 2012)Figure4shows the impact of intermodal transport on the containerized port system.During those days,

Figure 4 .
Figure 4.The impact of intermodal transport on the containerized port system.(Hayuth, 1987) r = quantity of output r u r = weight attached to output r x s = quantity of input s v s = weight attached to input s The parameters and variables are needed in developing the model.Therefore, the model is based on the following parameters and variables:N = number of DMU {j = 1,2,...n} y = number of outputs {y = 1,2,...R} for the efficiency measurement of DMUs on a variable returns to scale basis.The BCC model results in an aggregate measure of technical and scale efficiency, the CCR model is only capable of measuring technical efficiency.This allows for the separation of the two efficiency measures.

Figure 5 :
Figure 5: DMU and Homogeneous units vector.Solving above equation for each one of the N container terminals of the sample, N weights and N optimum solution found.Each optimum solution j ψ * is the efficiency indicator of container terminal j and, by construcare ef- ficient.Charnes et al. (1978) model constant returns to scale (CRS) was modified byBanker et al (1984) by adding the restrictionN i i =1 ë =1∑ , this has generalising model to variable returns to scale (VRS) as below;Charnes et al. (1978) from DEA-CCR discover the objective evaluation of overall efficiency and identify the resources and estimates the amounts of the identified inefficiencies.Thus it is called constant return to scale (CRS).Albeit,Banker et al, (1984), DEA-BCC remove the constraint from the CCR model by adding

Figure
Figure 1-0: Container Terminal Yearly Efficiency(Output-oriented Efficiency Rating) depicts the IDEF0 model for container terminal.

Table 1 . Capacity of Quay Crane
use average lifting capacity to indicate average lifting of quay crane at wharf.By using this, we are able to average maximum lifting capacity of quay crane.The lifting capacity of quay cranes are different according to it series i.e., Table 1 depicts Westport Malaysia container terminal informs its quay crane specification and Table 2 represents acronym for input and output.

Table 2 . Input and Output
Table 3 depicts descriptive statistics analysis which represent maximum, minimum, average and standard deviation of inputs and output.The maximum and minimum of TTA are 1800 and 27.28 m 2 respectively.The average and standard deviation for TTA are 723.876and 535.758 m 2 respectively.Maximun and minimum quay crane index are 1980 and 120 respectively with the average and standard deviation at 724.73 and 508.79.as for output, the maximum and minimum T (million) Teus at 5988.066 and108.108respectively with the average and standard deviation at 2189.48 and 1776.94.

Table 3 : Descriptive Statistics on input/output data
The descriptive statistics shows the varies in result as the container terminals in Peninsular Malaysia are different in size, equipment and throughput.In addition, correlation between variables is shown in Table4.Ideally, there is no weak correlation, the lowest at medium correlate (0.607) yet signif-icant.The highest correlations are 0.946 and 0.944 between BL and T, also YS and T. It means all variables are accepted as there are no strong correlations among variables with positive correlation.Table5(Apendix) depicts raw data for analysis.The data is used to tabulate the result accordingly.

Table 6 : DEA-CCR Ranking Score (Output-oriented)
Table 8 and 9 represent efficiency and projection score inputs and output for DEA-CCR and DEA-BCC.The analysis for DEA-CCR efficiency i.e., AW03