Vous êtes sur la page 1sur 17

International Journal of Operations & Production Management

Emerald Article: On using DEA for benchmarking container terminals M.B.M. de Koster, B.M. Balk, W.T.I. van Nus

Article information:
To cite this document: M.B.M. de Koster, B.M. Balk, W.T.I. van Nus, (2009),"On using DEA for benchmarking container terminals", International Journal of Operations & Production Management, Vol. 29 Iss: 11 pp. 1140 - 1155 Permanent link to this document: http://dx.doi.org/10.1108/01443570911000168 Downloaded on: 19-07-2012 References: This document contains references to 30 other documents To copy this document: permissions@emeraldinsight.com This document has been downloaded 1768 times since 2009. *

Users who downloaded this Article also downloaded: *


Shana Wagger, Randi Park, Denise Ann Dowding Bedford, (2010),"Lessons learned in content architecture harmonization and metadata models", Aslib Proceedings, Vol. 62 Iss: 4 pp. 387 - 405 http://dx.doi.org/10.1108/00012531011074645 James DeLisle, Terry Grissom, (2011),"Valuation procedure and cycles: an emphasis on down markets", Journal of Property Investment & Finance, Vol. 29 Iss: 4 pp. 384 - 427 http://dx.doi.org/10.1108/14635781111150312 Charles Inskip, Andy MacFarlane, Pauline Rafferty, (2010),"Organising music for movies", Aslib Proceedings, Vol. 62 Iss: 4 pp. 489 - 501 http://dx.doi.org/10.1108/00012531011074726

Access to this document was granted through an Emerald subscription provided by OXFORD BROOKES UNIVERSITY For Authors: If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service. Information about how to choose which publication to write for and submission guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information. About Emerald www.emeraldinsight.com With over forty years' experience, Emerald Group Publishing is a leading independent publisher of global research with impact in business, society, public policy and education. In total, Emerald publishes over 275 journals and more than 130 book series, as well as an extensive range of online products and services. Emerald is both COUNTER 3 and TRANSFER compliant. The organization is a partner of the Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.
*Related content and download information correct at time of download.

The current issue and full text archive of this journal is available at www.emeraldinsight.com/0144-3577.htm

IJOPM 29,11

On using DEA for benchmarking container terminals


M.B.M. de Koster
Rotterdam School of Management, Erasmus University, Rotterdam, The Netherlands

1140
Received 27 November 2008 Revised 15 April 2009 Accepted 27 July 2009

B.M. Balk
Rotterdam School of Management, Erasmus University, Rotterdam, The Netherlands and Statistics Netherlands, The Hague, The Netherlands, and

W.T.I. van Nus


Maritime & Transport Business Solutions, Rotterdam, The Netherlands
Abstract
Purpose The purpose of this paper is to compare efciency scores from the benchmarking exercise with those of previous studies and to discuss the reasons behind diverging results. Design/methodology/approach This paper uses data envelopment analysis (DEA) on primary data of large container terminals. Findings The results differ strongly from those available in the literature. Causes for these differences are: public (secondary) data are not always accurate; different terminal types are compared; terminals of different scale are compared; and terminals are mixed with ports. Practical implications DEA may be appropriate for container terminal benchmarking, but only if better quality and additional input and output data can be obtained. In its application, the analysis should be controlled for terminal types. Originality/value Summary of the state of play in the use of DEA methodologies for comparing the efciency of container terminals at ports. Keywords Data analysis, Benchmarking, Freight containers Paper type Research paper

International Journal of Operations & Production Management Vol. 29 No. 11, 2009 pp. 1140-1155 q Emerald Group Publishing Limited 0144-3577 DOI 10.1108/01443570911000168

Introduction Owing to strongly growing international trade, the container industry has seen an almost exponential growth in the last decades. As a result from mergers and takeovers large stevedoring companies and shipping lines have emerged. In 2006, about half of the total worlds container port capacity was controlled by a small number of stevedores and shipping lines who operate on a global scale (Drewry, 2007). The four largest players, Hutchinson Port Holdings (HPH), APM terminals (APMT), Port of Singapore Authority (PSA), and Dubai Ports (DP) World, each control a large number of terminals on all continents. Not only companies have consolidated and grown, but also the ships have grown in size. The largest container ships are able to carry about 14,000 twenty-feet equivalent unit (TEU; a standard container size). As the size
The views expressed in this paper are those of the authors and do not necessarily reect the views of their organizations.

of container vessels is still increasing (Ircha, 2001), the number of ports of call is decreasing (Gilman and Williams, 1976) and, moreover, due to the increasing demand for container transportation, the number of ships is still increasing. These trends increase the importance of the marine container terminals in the complete chain of international transport. This importance is reinforced by the effect that, although the number of ports-of-call is limited, the (super) post-Panamax container vessels experienced an increase of relative in-port times compared to smaller vessels (24 percent of overall voyage time for a 8,000 TEU vessel compared to 17 percent for a 4,000 TEU Panamax vessel; Miduro et al., 2005). Container shipping is a cyclical market where under and overcapacity succeed each other following the economic cycles. The cycles have a signicant impact on the overall competitiveness of the companies within this industry. All such companies make great efforts to identify these shipping cycles and predict future demand. While the industry is highly protable at the peak of the cycle, a depression may have serious consequences for protability. Owing to the costs and the long production time of vessels and port facilities, the shipping cycle resembles the pig cycle. When demand and prices are high more investments are made. However, it takes time for their effect to appear. By then the market becomes saturated, which leads to a price decline. As a result, capacity is reduced but the effects take time to be noticed, which then leads to increasing demand and again increasing prices. This procedure repeats itself cyclically. This justies a container terminals focus on inputs or outputs at different moments in the cycle, when attempting to improve performance. When the market is strong and the cycle is at a peak, terminals face under-capacity: they have insufcient capacity to cope with the supply of containers from the shipping lines. In this case a focus on maximizing the output of every terminal is expected. Large investments in terminal facilities, cranes and additional employment can be seen at this stage. In times of depression, the number of containers to be handled drops. Terminals try to minimize costs by using a minimum amount of inputs like employees, energy, and cranes. Decreasing capacity is very hard in this industry because the expected life of a terminal facility (30 years) is much longer than an average cycle. Shipping lines and ship builders face the same cycle, so the expected scarcity in terminal capacity will be mitigated by a scarcity in ship capacity. As container volumes have been growing rapidly, the main performance criteria of container terminals currently were to allocate scarce resources (handling equipment and labor) to maximize throughput, paired with a focus on speed of handling vessels and other transport modalities. At a terminal, different types of material handling equipment are used to transship containers from vessels to barges, trucks, and trains and vice versa. The containers can be transshipped directly from one mode of transportation to another. Alternatively, containers can be stored for a certain period in a stack, before they are transferred to another mode. Extensive overviews of various processes and decision making at terminals are given by Vis and de Koster (2003), Steenken et al. (2004), and Stahlbock and Voss (2008). The unloading and loading process at a typical modern import/export container terminal is shown in Figure 1. This process can be divided into different subprocesses. When a ship arrives at a port, quay cranes (QCs) take the import containers off the ships hold or off the deck. Next, the containers are transferred from the QCs to vehicles that

Benchmarking container terminals 1141

IJOPM 29,11

1142

travel between the ship and the stack: automated guided vehicles, tow trucks with trailers, or straddle carriers (SCs). This stack consists of a number of lanes, where containers can be stored for a certain period. The stack lanes are served by systems like cranes automated stacking crane (ASC) or SCs. A SC can both transport containers and store or retrieve them at the stack. If a transport vehicle arrives at the stack, it puts the load down or the stack crane takes the container off the vehicle and stores it in the stack. After a certain period the containers are retrieved from the stack by cranes and transported by vehicles to transportation modes like barges, deep sea ships, trucks, or trains. This process can also be executed in reverse order, to load export containers onto a ship. Transshipment terminals primarily serve as sea hubs and transship containers from one sea going vessel to other sea going vessels. The processes in these types of terminals differs: as import/export terminals handle multiple transport modalities they usually show a greater variety in processes. Quay space and materials handling equipment used at a terminal are very expensive, regardless of whether they are automated or manned: a QC costs e6-8 million. The investment in a large modern container terminal can therefore easily amount to e100 million. For the performance of container terminals the question is how a maximum throughput can be achieved, paired with a minimum turnaround time per ship and using minimum resources. In operational terms, managers want to know whether they achieve a proper performance at their terminal. Many studies have been carried out to measure performance of container terminals and to compare different terminals. However, we will argue in this paper that many of these studies are awed. Most studies use data envelopment analysis (DEA) and this method is very sensitive to data errors. DEA also presupposes that the decision-making units (DMUs) are comparable. We nd that several studies do not have consistent units of analysis: they compare container terminals with complete ports (which often contain multiple container terminals, owned by different companies and presumably using different technologies), or they compare small terminals with large ones. Small terminals that cannot handle large ships are less efcient than larger ones, as frequently changing ships on the quay takes much time and therefore leads to lower throughput (measured in TEUs handled per time unit) at the same level of input. In this paper, we use DEA to benchmark container terminals, using detailed, extensively cross-checked, information obtained from container terminal operators. We nd sub-stantial differences with previous benchmarking studies. The remainder of the paper is organized as follows: Section DEA modeling briey introduces DEA as a benchmarking method. Section Benchmarking container terminals gives a literature overview. Sections Methodology and Results present method and results, followed by conclusions and discussion.
QC ASC SC Truck train barge Landside transport

Figure 1. Typical process of loading and unloading a vessel at an import/export terminal

AGV Quayside transport

Vessel

DEA modeling The literature shows that different methods are used for benchmarking, i.e. cross-sectionally comparing, operations. Most companies use ratio indicators; see de Toni and Tonchia (2001) and de Koster and Warffemius (2005), for overviews of performance measurement and comparison methods in production and warehouse environments, respectively. Container terminal operators use for example QC productivity or container output per hectare of land used. The problem with such partial indicators is that they cannot measure overall technological efciency. Non-parametric methods, such as DEA, measure technological efciency by relating multiple inputs to multiple outputs. With other approaches, such as the stochastic frontier approach (SFA), it is possible to measure economic efciency, which is a broader concept than technological efciency, as it covers the units reactions to market prices. Economic efciency includes technical efciency and allocative efciency. DEA has been used widely for benchmarking DMUs, varying from university libraries (Reichmann and Sommersguter-Reichmann, 2006) to nursing homes (Duffy et al., 2006), third-party logistics providers (Min and Joo, 2006), airports (Graham, 2005), and warehouses (de Koster and Balk, 2008). A basic assumption of DEA is that, during the time period considered, all DMUs have access to the same technology. Such a technology is dened as the set of all feasible combinations of input quantities and output quantities; that is, all those combinations of input quantities and output quantities such that the output quantities are producible from the input quantities. Let there be N inputs, the quantities of which are measured by a vector x ; x1 ; . . . ; xN , and M outputs, the quantities of which are measured by a vector y ; y1 ; . . . ; yM . Then: S {x; yjx can produce y} 1

Benchmarking container terminals 1143

is the formal denition of the technology. The usual regularity assumptions, in particular free disposability of inputs and outputs, are supposed to hold. Free disposability means that producing less output quantities with the same input quantities is feasible, as well as producing the same output quantities with more input quantities. This technology must of course, be estimated. Given K observed DMUs, we have a set of data {x k ; y k ; k 1; . . . ; K }. The very least we know is that all these data are contained in the technology set S. The technique of DEA consists of enveloping all the linear combinations of these data as closely as possible by a piecewise linear contour, which is called the frontier. Put formally, the technology can be approximated by: ( ) K K X X k k Z k x # x; y # Z k y ; zk $ 0 k 1; . . . ; K 2 S < x; yj
k1 k1

This specication imposes globally constant returns to scale (CRS). Adding the P restriction: K k1 zk 1 would impose variable returns to scale (VRS). Adding further the restriction that all zk [ {0,1} leads to the free disposal hull (FDH) technology. Any DMU that is lying inside the set S is called (technically) inefcient. Such a DMU could, given its output quantities, reduce its input quantities without leaving the technology set; or, given its input quantities, it could expand its output quantities while staying in the technology set. Any DMU that is lying on the boundary (the frontier)

IJOPM 29,11

of the set S is called (technically) efcient. A numerical measure of technical efciency of a DMU can then be based on the magnitude of feasible input quantity reduction or output quantity expansion. In the literature several of those measures have been proposed. We choose here a simple radial measure. The input-oriented technical efciency of DMU k (k 1, . . . , K) is then dened by: ITE x k ; y k min{djdx k ; y k [ S } In view of equation (2), this efciency is calculated as: ITE x k ; y k mind;z d subject to
k X k0 1

1144

z k0 x # d x ; y #

k0

K X k0 1

zk0 y ; zk0 $ 0 k 1; . . . ; K

k0

30
0

As one sees, the calculation of ITE (x k, y k) reduces to the solution of a linear programming problem. For example, ITE (x k, y k) 0.6 means that all the input quantities of DMU k can be reduced by 40 percent, such that the output quantities remain producible. Put otherwise, the input-oriented technical efciency is an inverse measure of the distance of DMU k to the frontier of the technology. By looking at the LP-dual formulation of problem (30 ), alternative interpretations of the input-oriented efciency can be given; see various chapters of Fried et al. (2008). Output-oriented technical efciency is, similar to equation (3), dened by: OTE x k ; y k 1 max{fjx k ; fy k [ S} 4

Under CRS the two measures of technical efciency coincide, whereas under VRS they are different. The DMUs that are making out that part of the frontier to which DMU k is compared are called its efcient peers. Mathematically seen, they correspond to the positive solution values for the intensity variables zk 0 (k 0 1, . . . , K) in the linear programming problem (30 ). Benchmarking container terminals Many researchers have used DEA for benchmarking container terminals. Roll and Hayuth (1993) were among the rst to apply DEA to ports, albeit not based on real data. Martinez-Budria et al. (1999) compared the performance of 26 Spanish ports. Song et al. (2001) applied DEA to container terminals in Korea and the UK. In subsequent papers Wang et al. (2003) and Cullinane et al. (2005) considered a broad range of DEA variants. They compared results from DEA-CRS, DEA-VRS, and FDH models. Lin and Tseng (2005) compared DEA results with SFA results on 27 container ports. In a recent publication Wang and Cullinane (2006, p. 95) performed an analysis of 104 container terminals in Europe. They stressed that:
[. . .] it is extremely important to note that although the results derived from DEA provide important information on theoretically optimum production, such results should always be interpreted with a fair degree of caution in practice.

They referred to the large variation in port characteristics (natural situation, cargo hinterland, and inland transportation). Other publications that are worth mentioning deal with geographical efciency comparisons using DEA (Tongzon, 2001; Barros and Athanassiou, 2004; Ramos Rios and Gastaud Mac ada, 2006), and multistage DEA (Park and De, 2004). The unit of analysis (terminals, ports, or mixtures thereof), the type of data, the DEA model, the input and output variables in all these studies vary greatly. In most studies secondary data were used, in particular from various issues of the Containerization International Yearbook and Lloyds Ports of the World. These trade publications collect information directly from port authorities and container terminal operators. However, as we will show in the next section, the quality of these data are not guaranteed, which can easily lead to unreliable results. Although some publications are recent, the most recent data used has stemmed from 2003. See Table I for a summary. Looking at these papers some observations are: . Apparently, different units of analysis are used: container ports as well as container terminals. However, particularly the larger ports often contain multiple terminals. Some of them are small, while others may be very large. For example, Rotterdam is home to the very large HPH Europe Container Terminals (ECT) terminal and the very small Uniport terminal (and several others), both handling sea-going vessels. These terminals have different operators and different processes (for example, the Uniport terminal cannot handle large vessels). In DEA it is only reasonable to compare units with similar production functions. According to Alderton (1999) for maximal comparability of data one should concentrate on terminals, but, a distinction should also be made between import/export and transshipment terminals, as these have different production processes. . Different input and output variables are used. According to Wang et al. (2002):
[. . .] the variables containing information on human resources [such as how many stevedores and management staff ], natural resources and man-made resources [such as terminal area, number of cranes, number of container berths, number of tugs] should be built into DEA models as input variables. The output variables should include cargo ow variables [such as container throughput], the quality of customer service [such as the delay time of a ship at port].

Benchmarking container terminals 1145

Unfortunately, much of such data are treated by container terminal operators as strictly condential and only a few variables are in the public domain. Different models were used: input- and output-oriented DEA, constant and VRS, etc. The tendency in more recent papers is to use and compare multiple methods.

One of the big problems is the reliability of public data. Table II shows some striking differences between the different sources. For example, the number of QCs at the Rotterdam ECT Delta terminal in 2005 seemed to vary between 7 and 25. However, according to ECTs own web site the number was 28. By comparing the berth length for the APMT in Los Angeles as provided by the different sources we recognize a plausible cause of error: apparently data were entered incorrectly.

1146

IJOPM 29,11

Reference Hypothetical data of 20 ports 26 Spanish ports, with each VRS ve observations, 1993-1997 Four Australian and 12 other container ports, 1996 57 container terminals in the CRS, VRS, largest 28 container ports FDH 25 international container ports 1992-1999 CRS, VRS CRS Cargo throughput, ship working rate Container throughput Container throughput CRS Manpower, capital, cargo uniformity

Roll and Hayuth (1993) Rate efciency of ports

Martinez-Budria et al. (1999)

Tongzon (2001)

Wang et al. (2003) and Cullinane et al. (2005)

Cullinane et al. (2004)

Barros and Athanassiou (2004)

Park and De (2004)

Lin and Tseng (2005)

Wang and Cullinane (2006)

Cullinane et al. (2006)

Table I. Overview of DEA applications to container port and terminal environments


Data Cargo throughput, service level, user satisfaction, ship calls Labor cost, depreciation Total cargo throughput, charges, other expenditures port revenue DEA model Inputs Outputs Number of cranes, container berths, and tugs, terminal area, delay time, labor Quay length, terminal area, number of QCs, yard cranes, and SCs Quay length, terminal area, number of QCs, yard cranes, and SCs Labor, capital

Objectives

Examine relative efciency of ports and efciency development Test factors inuencing port performance and efciency

Efciency ranking of container ports

Efciency development

Compare port efciency and Six international seaports in CRS, VRS Ship throughput, freight nd best practices Greece and Portugal, 1998movement, cargo handled, 2000 container throughput Four-stage port efciency 11 Korean ports Multi-stage Berthing capacity, cargo handling capacity and measurement CRS, VRS throughput, ship throughput, revenue, customer satisfaction Comparing DEA and SFA 27 container ports, 1999CRS, VRS Number of gantry cranes, Container throughput rankings 2002 quay length, stevedoring equipment Ranking of terminals, 104 European terminals, CRS, VRS Quay length, terminal area, Container throughput relating this to location and with throughput . 10,000 equipment costs scale TEU in 2003 Comparing DEA and SFA 57 container terminals in the CRS, VRS Quay length, terminal area, Container throughput rankings largest 28 container ports, number of QCs, yard cranes, 2001 and SCs

Terminal 1,750,000 4,350,000 1,643,000 3,601,000 6,095,200 2,500,000 880,000 1,560,000 7 14 20 25 6 6 10 39 11 10 14 1,260 1,097 1,219 1,236 1,097 2,192 83 101 47 69 196 2,600 1,250 2,281 81 70 82 102 55 69 139 101 55 67 196 8 22 14 19 39 5 6 10 9 25 23 24 39 12 10 12 11 28 12 1,250 3,400 1,250 2,281 2,384 836 1,097 2,191 1,250 3,600 1,250 2,281 3,024 1,236 1,097 1,970 1,600 3,600 1,250 70 236 70 236 80 55 97 54 69 196 93 248 155

CI

Throughput (TEU) LL Dr Web CI CI

Number of cranes CI LL Dr Web

Berth length (m) LL Dr Web

Yard area (ha) LL Dr Web

Rotterdam-NL-APMT-2006 Rotterdam-NL-HPH-ECT Delta-2006 Shanghai-PRC-APMT-2006 Shanghai-PRC-HPH-2005 Dubai-UAE-DPW-Jebel Ali-2005 Raysut/Salalah-Oman-APMT-2006 Long Beach-USA-W-APMT-2005 Los Angeles-USA-W-APMT-2006

Notes: CI, Containerization International Yearbook; LL, Lloyds list; Dr, Drewrys shipping consultants; web, terminals own web site; a blank cell means not available

Benchmarking container terminals 1147

Table II. Example of data on container terminals from various sources

IJOPM 29,11

Methodology In this section, we discuss the details of the application of DEA to container terminals, in particular the units of analysis, the data, the way the analysis has been executed, and the variables used. Unit of analysis As container terminals should not be mixed with container ports, we only include container terminals. Furthermore, we do not mix small and large terminals in the sample. Large terminals are by nature more efcient than small ones. We therefore included only terminals with an annual throughput of at least 500,000 TEU. Data We have used primary data related to 2006 (some to 2005), obtained in cooperation with two large container operators, APMT (39 terminals) and European terminals of PSA (seven terminals). The data of APMT extends to terminals of major competitors. We further used data from Drewry (2006, 2007), which was veried with knowledgeable persons of APMT and cross-checked with public information on the companys own web sites. In addition, we occasionally used Google maps to verify the number of QCs. Only terminals were included of which data could be established unambiguously. In total 38 terminals were included. Although our data can still be awed, the fact that we used multiple sources of information strongly reduces the probability of major errors (like those listed in Table II). Table III lists the data used in this study. Execution There is no agreement among researchers about whether to use a CRS or a VRS production model. Most publications use both models and we followed that practice. Since a terminal operator has a fairly stable customer base and is therefore able to predict important changes in demand, Cullinane et al. (2006) recommend an input-oriented approach for minimizing resources needed for this demand. When capacity utilization and expansion plans are the issue-at-stake they recommend an output-oriented approach. We have applied both approaches. Variables We only included input and output variables with publicly available data, as used by many other authors. The output variable is annual throughput in TEU. This variable (sometimes indicated as cargo throughput) was used in all the papers we studied. The input variables are: number of quay gantry cranes (which is a measure for all the capital invested), total quay length (in meters), terminal area (in hectare). These variables are commonly used by various authors. The number of annual direct man hours is important, but unavailable. Notteboom et al. (2000) discovered that there exists a fairly stable and close relationship between the number of gantry cranes and the number of dock workers in a container terminal, which suggests that this variable can be dropped. Analysis showed that all the input variables correlate strongly with throughput and also with each other. Results DEA-CRS efciency scores recall that under CRS the orientation does not matter can be found in the second and third column of Table IV. It appears from results not

1148

Terminal 19 8 11 10 10 39 9 24 10 18 21 20 9 49 2 5 6 15 4 10 12 12 7 11 25 12 9 25 23 24 29 37 36 27 27 22 6 5 1,494 1,180 1,200 1,240 1,128 3,024 1,402 2,084 1,036 3,011 1,400 2,850 995 3,292 610 640 675 1,431 300 1,097 1,970 1,300 1,200 1,378 2,400 1,236 1,250 3,600 1,250 2,281 2,630 3,220 2,319 8,300 3,600 1,635 650 750 68.60 66.46 62.00 69.50 78.50 96.70 93.10 108.87 37.00 130.00 55.20 160.00 40.00 92.00 29.00 25.80 20.90 59.50 10.50 68.80 195.90 37.00 65.56 106.40 225.00 54.00 70.00 236.00 80.00 55.00 79.00 96.00 84.00 80.00 110.00 100.00 21.80 35.60 2,937,000 1,134,635 1,090,840 2,067,000 1,273,430 6,095,200 1,523,800 2,175,000 1,115,000 3,160,981 1,800,000 2,600,000 8,70,000 5,969,000 5,22,087 800,000 1,200,000 1,510,800 590,746 880,000 1,560,000 1,213,000 853,000 900,000 5,443,086 2,500,000 1,750,000 4,350,000 1,643,000 3,601,000 5,477,526 6,100,000 4,841,561 5,250,000 4,170,000 1,725,000 564,000 580,000 EUR EUR EUR EUR US ME ME EUR CAR EUR EUR EUR EUR SEA US EA EA CAR SEA US US EA US US EA ME EUR EUR EA EA SEA SEA SEA SEA SEA SEA SEA EA T I/E I/E I/E I/E T T I/E T T I/E I/E I/E I/E I/E I/E I/E T T I/E I/E T I/E I/E I/E T I/E I/E I/E I/E T T T T T I/E I/E I/E ASC SC SC SC ASC ASC SC ASC SC SC ASC SC SC ASC ASC ASC SC SC ASC ASC ASC ASC ASC ASC ASC ASC SC ASC ASC ASC ASC ASC ASC ASC ASC ASC ASC ASC

Gantry cranes

Inputs Quay length (m) Yard area (ha) Location

Output TEU

Characteristics Operation Material handling system

Algeciras Spain-APMT Antwerp Belgium-PSA-Europa Antwerp Belgium-PSA-Noordzee Bremerhaven Germany-APMT Charleston USA-E-SSA-Wano Welch Dubai UAE-DPW-Jebel Ali Dubai UAE-DPW-Port Rashid Felixstowe UK-HPH-Trinity Freeport Bahamas-HPH Gioia Tauro Italy-APMT Hamburg Germany-HHLA-Altenwerder Hamburg Germany-HHLA-Burchardkai Hamburg Germany-HHLA-Tollerort Hong Kong-HPH-HIT Houston USA-G-APMT Kaohsiung Taiwan-APMT-Terminal 4 Kaohsiung Taiwan-APMT-Terminal 5 Kingston Jamaica-APMT300 Laem Chabang Thailand-APMT Long Beach USA-W-SSA-Pier A Los Angeles USA-W-APMT Manila Philipines-ICTSI-MICT Newark USA-E-DPW Port Elisabeth USA-E-APMT Qingdao PRC-APMT Raysut/Salalah Oman-APMT Rotterdam The Netherlands-APMT Rotterdam The Netherlands-HPH-ECT Delta Shanghai PRC-APMT Shanghai PRC-HPH Singapore-PSA-Brani Singapore-PSA-Keppel Singapore-PSA-Pasir Panjang Singapore-PSA-Tanjong Pagar Tanjung Pelepas Malaysia-APMT Tanjung Priok Indonesia-HPH-JICT Tanjung Priok Indonesia-HPH-Koja Yokohama Japan-APMT

Notes: Location: EUR, Europe; US, USA; CAR, Caribbean; SEA, South-East Asia; EA, East Asia; Operation: T, transshipment; I/E, import/export; dominant material handling system for storage operations: ASC, automated stacking crane; SC, straddle carrier

Benchmarking container terminals 1149

Table III. Data

IJOPM 29,11
Terminal This study Score Rank 1 1 1 1 1 1 0.99 0.97 0.95 0.94 0.94 0.94 0.92 0.92 0.88 0.79 0.77 0.76 0.75 0.74 0.63 0.61 0.60 0.59 0.59 0.56 0.56 0.53 0.52 0.52 0.51 0.49 0.49 0.47 0.47 0.47 0.41 0.37 1 1 1 1 1 1 7 8 9 10 10 10 13 13 15 16 17 18 19 20 21 22 23 24 24 26 26 28 29 29 31 32 32 34 34 34 37 38

Cullinane et al. (2006) and Wang et al. (2003) Score Rank 1 0.50 0.90 0.90 0.90 0.44 0.61 0.75 1 0.83 0.90 0.90 0.46 1 0.74 1 17 7 7 7 20 15 10 1 9 7 7 19 1 11

Lin and Tseng (2005)a Score Rank 0.62 0.80 0.80 0.80 0.68 0.42 0.44 0.50 1 0.77 0.80 0.71 0.46 0.01 13 4 4 4 11 19 18 15 1 8 4 9 17 21

1150

Houston USA-G-APMT Kaohsiung Taiwan-APMT-Terminal 5 Qingdao PRC-APMT Raysut/Salalah Oman-APMT Singapore-PSA-Brani Singapore-PSA-Tanjong Pagar Singapore-PSA-Pasir Panjang Dubai UAE-DPW-Jebel Ali Bremerhaven Germany-APMT Shanghai PRC-HPH Laem Chabang Thailand-APMT Hong Kong-HPH-HIT Algeciras Spain-APMT Singapore-PSA-Keppel Rotterdam The Netherlands-APMT Gioia Tauro Italy-APMT Rotterdam The Netherlands-HPH-ECT Delta Kaohsiung Taiwan-APMT-Terminal 4 Tanjung Pelepas Malaysia-APMT Dubai UAE-DPW-Port Rashid Antwerp Belgium-PSA-Europa Hamburg Germany-HHLA-Altenwerder Shanghai PRC-APMT Charleston USA-E-SSA-Wano Welch Hamburg Germany-HHLA-Burchardkai Los Angeles USA-W-APMT Freeport Bahamas-HPH Newark USA-E-DPW Yokohama Japan-APMT Manila Philipines-ICTSI-MICT Kingston Jamaica-APMT Felixstowe UK-HPH-Trinity Tanjung Priok Indonesia-HPH-JICT Tanjung Priok Indonesia-HPH-Koja Antwerp Belgium-PSA-Noordzee Hamburg Germany-HHLA-Tollerort Long Beach USA-W-SSA-Pier A Port Elisabeth USA-E-APMT

0.64

14

0.71 1 0.35 0.25 0.38 0.55 1 0.48 1 0.68

12 1 22 23 21 16 1 18 1 13

0.95 0.48 0.66 0.62 1

3 16 12 13 1

0.70 0.24

10 20

Table IV. Efciency scores and ranks

Notes: aFor ports with multiple terminals, we attributed the Lin and Tseng port efciency score to the largest terminal (Rotterdam, Tanjung Priok, Antwerp) or, for terminals with comparable size (Singapore), to each of the terminals

shown here that under VRS in both orientations precisely the same ranking is obtained. The number of efcient DMUs then increases to 10 and the efciency scores of inefcient DMUs become higher.

The same table compares our scores and ranks with those obtained by Cullinane et al. (2006) (the same results were also reported by Wang et al., 2003), and Lin and Tseng (2005), these being the two most recent comparable publications. The table shows a number of striking differences in ranking, which are most likely partly due to errors in the data used (Table II). Lin and Tseng (2005) have a population consisting of both ports and terminals, which makes it hard to distinguish each type and to compare outcomes. When ranking DMUs it is important to consider the optimal weights for the different variables. As terminals need all input variables to deliver a sufcient throughput, efcient DMUs with optimal weights equal to zero for some input or output variables need to be treated with caution. As an example, Table V lists terminals with optimal weights smaller than 0.01 for three of the four variables. Houston is an efcient terminal, but has a zero weight for two of the three input variables. It is therefore questionable whether this terminal can serve as peer for other terminals. On further inspection there appears to be a measurement problem. This terminal reports that two quay gantry cranes are used but, according to APMT sources, it actually operates ve cranes, which are partly leased from the local port authority. Including this makes the terminal no longer efcient. This again shows the unreliability resulting from using reported data. The top three efcient terminals reported most frequently as peers to others, both under CRS and VRS are, in sequence, Quingdao-APMT, Raysut/Salah-APMT, and Singapore-PSA-Brani, all in Asia. Houston-APMT would be number four but, as mentioned, this score is awed. For the relationship between the type of operations and the efciency score, and the type of material handling systems and the efciency score we executed t-tests. For the relationship between geographical location and efciency, and between terminal owner-ship and efciency we carried out ANOVA analysis. We clustered the efciencies by these criteria in Table VI. Our results lead to the following conclusions: . American terminals have relatively low efciencies. This can be caused by their inefcient use of land (wheeled stacking operations) and/or congestion at the ports due to a lack of depth. Terminals in the Middle East are, although not signicantly, the most efcient, followed by terminals in East Asia and South-East Asia. The relatively high-efciency of terminals in South-East Asia can be attributed partly to large-scale operations on very small areas using
Weight Quay length Yard area 0.00017159 0.00016263 0.00003808 0.00005855 0.00008364 0.00003365 0 0.00271322 0.00257153 0 0.00092575 0.00132250 0.00053204 0

Benchmarking container terminals 1151

Terminal Newark USA-E-DPW Port Elisabeth USA-E-APMT Qingdao PRC-APMT Raysut/Salalah Oman-APMT Rotterdam The Netherlands-APMT Rotterdam The Netherlands-HPH-ECT Delta Houston USA-G-APMT

Score 0.53 0.37 1 1 0.88 0.77 1

Gantry cranes 0.21435252 0.20315856 0.03634431 0.07313708 0.10448154 0.04203280 0.50000000

TEU 0.00000117 0.00000111 0.00000018 0.00000040 0.00000057 0.00000023 0.00000192

Table V. Terminals with small optimal weights

IJOPM 29,11

Characteristic Geography

Cluster Europe US Caribbean Middle East South-East Asia East Asia Transshipment Import/export Gantry crane SC , 1 million TEU 1-2 million TEU 2-5 million TEU . 5 million TEU APMT HPH PSA DPW SSA ICTSI HHLA

Number of units 11 6 2 3 9 7 14 24 27 11 9 13 10 6 16 7 6 3 2 1 3

Mean 0.69 0.58 0.53 0.90 0.79 0.82 0.85 0.67 0.74 0.69 0.61 0.63 0.82 0.97 0.78 0.66 0.84 0.74 0.50 0.52 0.55

SD 0.18 0.25 0.03 0.14 0.23 0.20 0.18 0.21 0.23 0.19 0.23 0.16 0.18 0.04 0.21 0.22 0.23 0.22 0.12 0.00 0.08

1152
Operations Handling system Size

Owner

Table VI. Efciencies per cluster

gantry cranes and overhead bridge cranes for stacking. Such cranes can go up to nine levels high (Pasir Panjang), which reduces the land requirements. Although this may imply longer shufing times in the stack, this can be compensated partly by using more (manual) transport vehicles. Since labor in this region is relatively cheap, this apparently does not decrease efciency. Transshipment terminals are signicantly more efcient than import/export terminals. This is explained by the simpler processes at transshipment terminals. Transshipment terminals primarily serve as sea hubs and move containers from one sea-going vessel to other sea-going vessels. Import/export terminals move loads between deep-sea vessels and other transport modes and usually show a greater variety in processes. All the large Middle East and Singaporean terminals in the sample are primarily transshipment terminals which explains their high-efciency scores. Terminals with gantry cranes for stacking are on average slightly more efcient than terminals that operate SCs for stacking. The difference is, however, not signicant. Terminals of PSA are on average the most efcient, followed by terminals of APMT and DP World. The difference is however, not signicant. Size matters. The larger the throughput, the higher the efciency. This is not surprising, since economies of scale are prevalent in this industry.

Conclusions and discussion DEA is commonly used to benchmark container terminals by using public data. Many studies have appeared in the last decade. The method requires no prior information

on the data and is a powerful tool to relate multiple inputs to multiple outputs. Yet the method also has some drawbacks. As it is an extreme value method it is sensitive to data errors. Unfortunately, much detailed data are regarded as condential by the companies and therefore not available. All studies rely on public data, which appears to be highly unreliable. This leads to major errors in the results. On top of this, researchers tend to mix terminals with ports, very small with very large terminals, and import/export with transshipment terminals, which make their results difcult to compare and interpret. Using primary, veried data of large terminals only we are able to compare terminals of different types (import/export versus transshipment). Our results show that larger terminals are more efcient and transshipment terminals are signicantly more efcient than import/export terminals. Differences between terminals in different parts of the world, between terminals of different operators, and with different material handling systems for stacking are not signicant. Implications for management practice When we confronted the management of APMT with the results of our study they were not really surprised: our results conrmed their intuitive insights about the relative position of APMT, particularly as compared to non-APMT. Though performance differences between terminals are intuitively known, particularly for third-party terminals the data are often incomplete or unreliable. Both high-quality data and DEA analysis are helpful for quantifying managerial conclusions. Still, using publicly available data, albeit of high quality, does not tell the full story. At real-life container terminals other variables also play a role in measuring performance. Output variables of great importance to the terminals are, for example, the turnaround time of sea-going vessels and other transport modes (barges, trucks, and trains), and safety aspects, measured by number of yearly accidents. An important input variable is the number of operators used. Data on such variables are however, not publicly available. DEA may therefore be appropriate for container terminal benchmarking, but the value for terminal operators and policy decision makers will be limited, unless better quality and additional input and output data can be obtained. In its application, the analysis should be controlled for terminal types.
References Alderton, P. (1999), Port Management and Operation, LLP, London. Barros, C.P. and Athanassiou, M. (2004), Efciency in European seaports with DEA: evidence from Greece and Portugal, Maritime Economics and Logistics, Vol. 6, pp. 122-40. Cullinane, K., Song, D.-W. and Wang, T. (2005), The application of mathematical programming approaches to estimating container port production efciency, Journal of Productivity Analysis, Vol. 24, pp. 73-92. Cullinane, K., Wang, T.F., Song, D.-W. and Ji, P. (2006), The technical efciency of container ports: comparing data envelopment analysis and stochastic frontier analysis, Transportation Research Part A, Vol. 40, pp. 354-74. de Koster, M.B.M. and Balk, B.M. (2008), Benchmarking and monitoring international warehouse operations in Europe, Production and Operations Management, Vol. 17 No. 2, pp. 1-10.

Benchmarking container terminals 1153

IJOPM 29,11

1154

de Koster, M.B.M. and Warffemius, P.M.J. (2005), American, Asian and third-party international warehouse operations in Europe: a comparison, International Journal of Operations & Production Management, Vol. 25 No. 8, pp. 762-80. de Toni, A. and Tonchia, S. (2001), Performance measurement systems. Models, characteristics and measures, International Journal of Operations & Production Management, Vol. 21 Nos 1/2, pp. 46-70. Drewry (2006, 2007), Annual Review of Global Container Terminal Operators, Drewry, London. Duffy, J., Fitzsimmons, J. and Jain, N. (2006), Identifying and studying best-performing services; an application of DEA to long-term care, Benchmarking: An International Journal, Vol. 13 No. 3, pp. 232-51. Fried, H.O., Lovell, C.A.K. and Schmidt, S.S. (Eds) (2008), The Measurement of Productive Efciency and Productivity Growth, Oxford University Press, New York, NY. Gilman, S. and Williams, G.F. (1976), The economics of multi-port itineraries for large container ships, Journal of Transport Economics and Policy, Vol. 10 No. 2, p. 137. Graham, A. (2005), Airport benchmarking: a review of the current situation, Benchmarking: An International Journal, Vol. 12 No. 2, pp. 99-111. Ircha, M.C. (2001), Serving tomorrows mega-size containerships: the Canadian solution, International Journal of Maritime Economics, Vol. 3, pp. 318-32. Lin, L.-C. and Tseng, L.A. (2005), Application of DEA and SFA on the measurement of operating efciencies for 27 international container ports, Proceedings of the Eastern Asia Society for Transportation Studies, Vol. 5, pp. 592-607. ez, M. and Ravelo-Mesa, T. (1999), A study of n Martinez-Budria, E., Diaz-Armas, R., Navarro-Iba the efciency of Spanish port authorities using data envelopment analysis, International Journal of Transport Economics, Vol. 26 No. 2, pp. 237-53. Miduro, R., Musso, E. and Parola, F. (2005), Maritime liner shipping and the stevedoring industry: market structure and competition strategies, Maritime Policy and Management, Vol. 32, pp. 89-106. Min, H. and Joo, S. (2006), Benchmarking the operational efciency of third party logistics providers using data envelopment analysis, Supply Chain Management: An International Journal, Vol. 11 No. 3, pp. 259-65. Notteboom, T.C., Coeck, C. and van den Broeck, J. (2000), Measuring and explaining the relative efciency of container terminals by means of Bayesian stochastic frontier models, International Journal of Maritime Economics, Vol. 2, pp. 83-106. Park, R.-K. and De, P. (2004), An alternative approach to efciency measurement of seaports, Maritime Economics and Logistics, Vol. 6, pp. 53-69. Ramos Rios, L. and Gastaud Mac ada, A.C. (2006), Analysing the relative efciency of container terminals of mercosur using DEA, Maritime Economics and Logistics, Vol. 8, pp. 331-46. Reichmann, G. and Sommersguter-Reichmann, M. (2006), University library benchmarking: an international comparison using DEA, International Journal of Production Economics, Vol. 100 No. 1, pp. 131-47. Roll, Y. and Hayuth, Y. (1993), Port performance comparison applying data envelopment analysis, Maritime Policy & Management, Vol. 20, pp. 153-61. Song, D.-W., Cullinane, K. and Roe, M. (2001), The Productive Efciency of Container Terminals: An Application to Korea and the UK, Ashgate, Aldershot. Stahlbock, R. and Voss, S. (2008), Operations research at container terminals: a literature update, OR Spectrum, Vol. 30 No. 1, pp. 1-52.

Steenken, D., Vo, S. and Stahlbock, R. (2004), Container terminal operation and operations research a classication and literature review, OR Spectrum, Vol. 26, pp. 3-49. Tongzon, J.L. (2001), Efciency measurement of selected Australian and other international ports using data envelopment analysis, Transportation Research Part A, Vol. 35, pp. 107-22. Vis, I.F.A. and de Koster, M.B.M. (2003), Transhipment of containers at a container terminal: an overview, European Journal of Operational Research, Vol. 147, pp. 1-16. Wang, T.-F. and Cullinane, K. (2006), The efciency of European container terminals and implications for supply chain management, Maritime Economics and Logistics, Vol. 8, pp. 82-99. Wang, T.F., Song, D.-W. and Cullinane, K. (2002), The applicability of data envelopment analysis to efciency measurement of container ports, paper presented at Annual Conference and Meeting of the International Association of Maritime Economists, Panama City. Wang, T.F., Song, D.-W. and Cullinane, K. (2003), Container port production efciency: a comparative study of DEA and FDH approaches, Journal of Eastern Asia Society for Transportation Studies, Vol. 5, pp. 698-713. Corresponding author M.B.M. de Koster can be contacted at: rkoster@rsm.nl

Benchmarking container terminals 1155

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com Or visit our web site for further details: www.emeraldinsight.com/reprints

Vous aimerez peut-être aussi