The efficiency of European container ports: A cross-sectional data envelopment analysis

Transcription

1 International Journal of Logistics: Research and Applications ISSN: (Print) X (Online) Journal homepage: The efficiency of European container ports: A cross-sectional data envelopment analysis Kevin P. B. Cullinane & Teng-Fei Wang To cite this article: Kevin P. B. Cullinane & Teng-Fei Wang (2006) The efficiency of European container ports: A cross-sectional data envelopment analysis, International Journal of Logistics: Research and Applications, 9:1, To link to this article: Published online: 23 Jan Submit your article to this journal Article views: 2781 View related articles Citing articles: 61 View citing articles Full Terms & Conditions of access and use can be found at

2 International Journal of Logistics: Research and Applications Vol. 9, No. 1, March 2006, The efficiency of European container ports: a cross-sectional data envelopment analysis KEVIN P. B. CULLINANE* and TENG-FEI WANG School of Marine Science and Technology, Armstrong Building, University of Newcastle upon Tyne, Newcastle NE1 7RU, UK International Shipping and Logistics Group, Plymouth Business School, University of Plymouth, Drake Circus, Plymouth PL4 8AA, UK This paper focuses on measuring the efficiency of container terminals in Europe using the data envelopment analysis (DEA) approach. Container terminals in Europe play an important role in the region s economic development and, as the result of their geographic concentration, face fierce competition compared with the rest of the world. Despite this, studies of the efficiency of container terminals in the region are minimal. As a basis for further research, this paper serves to supplement existing studies by deriving estimates of relative efficiency for a sample comprising 69 of Europe s container terminals with annual throughput of over 10,000 TEUs. The scale properties of container terminal production are also considered as part of the study, as is the relationship of efficiency to geographical influence. Keywords: Efficiency; European container ports; Data envelopment analysis; Scale efficiency 1. Introduction Recent trends in international trade in an era characterised by the globalisation of production and consumption patterns have led to the increasing importance of container transportation. This is largely because of the numerous technical and economic advantages it possesses over traditional methods of transportation. Standing at the interface of sea and inland transportation, container ports play a pivotal role in the container transportation process. One distinctive feature of the contemporary container port industry is that competition between container ports is more intensive than has previously been the case. Port markets used to be perceived as monopolistic due to the exclusive and immovable geographical location of the port and the unavoidable concentration of port traffic. However, the rapid development of international container and intermodal transportation has drastically changed the market structure from one of monopoly to one where fierce competition is rampant in many parts of the world. Many container ports no longer enjoy the freedom yielded by a monopoly over the handling of cargoes from within their hinterland. Instead, they have to compete for cargo with their neighbouring ports. *Corresponding author. kevin.cullinane@ncl.ac.uk International Journal of Logistics: Research and Applications ISSN print/issn X online 2006 Taylor & Francis DOI: /

3 20 K. P. B. Cullinane and T.-F. Wang The above two characteristics of the contemporary container port industry are particularly true for the container ports in Europe. The paramount importance of the container port industry as the basis for the economic development of the EU (e.g. Winkelmans 2004, Martine Zarzoso et al. 2004) and the fierce competition between/among ports (e.g. Notteboom 1997, Wang and Cullinane 2004) have been variously discussed. It is the intense competition that characterises the container port industry that has stimulated an overt interest in the efficiency with which it utilises its resources. Indeed, the analysis of the performance of individual container ports or terminals is of great significance for the survival and competitiveness of the industry and its players. Not only can such an analysis provide a powerful management tool for port operators, but it also constitutes an important input for informing regional and national port planning and operations. It is imperative to note, however, that the analysis contained within this paper is aimed solely at deriving and comparing various estimates of the relative efficiency of the industry under study. Alluding to the significant level of competition within the industry provides merely a justification for doing so. It should not be inferred, therefore, that any direct causal relationship is being proposed to exist between changing levels of efficiency over time and the degree of competition within the container port industry. Although this is certainly a hypothesis that is worth testing, particularly because of the potential impact of market contestability (Baumol et al. 1982), it is one of such complexity and importance that it merits a separate study in its own right. Traditionally, the performance of ports has been variously evaluated by calculating cargohandling productivity at berth (Bendall and Stent 1987, Tabernacle 1995, Ashar 1997), by measuring a single factor productivity (De Monie 1987) or by comparing actual with optimum throughput over a specific time period (Talley 1998). In recent years, significant progress has been made concerning the measurement of efficiency in relation to productive activities. In this vein, two more complex, yet more appropriately holistic approaches, named data envelopment analysis (DEA) and stochastic frontier analysis (SFA), have been increasingly utilised to analyse port production and performance. The ensuing section of this paper provides a review of the major studies of port efficiency that have utilised either DEA or SFA approaches. This is followed in section 3 by an exposition of the DEA methodology that is to be employed within this paper for the analysis of a data set of variables derived from the European container handling industry. Section 4 presents a specification of the input and output variables that are to be encompassed within the model that is implicit to the analysis. This specification also dictates the nature, form and sources for the data collection exercise. The results of applying DEA to the sample data are presented in section 5 and an overview of the major results, conclusions and implications of the analysis are presented in section Literature review Among the relative dearth of previous applications, that of Roll and Hayuth (1993) is probably the first attempt in this respect. However, this work should be treated as a theoretical exploration of applying DEA to the port sector, rather than as a genuine application. This is because no actual data were collected or analysed. In Martinez-Budria et al. (1999), 26 ports were divided into three groups, namely: high complexity ports, medium complexity ports and low complexity ports. After examining the efficiency of these ports using DEA-BCC (due to Banker, Charnes and Cooper 1984) models, Martinez-Budria et al. (1999) concluded that the ports of high complexity are

4 Efficiency of European container ports 21 associated with high efficiency, compared with the random mix of medium and low efficiency found in the other ports. Using both DEA-CCR (due to Charnes, Cooper and Rhodes 1978) and DEA-additive models, Tongzon (2001) studied the efficiency of four Australian and 12 other international container ports for the year Clearly plagued by poor data availability and a small sample size (only 16 observations), Tongzon (2001) identified more ports as efficient, rather than inefficient. Realising this serious drawback, Tongzon (2001) suggested that further work should be done in collecting more observations to enlarge the sample analysed. Valentine and Gray (2001) applied the DEA-CCR model to 31 container ports out of the world s top 100 container ports for the year 1998 to examine the relationship between port efficiency and particular types of ownership and organisational structures. Barros and Athanassiou (2004) applied DEA to the estimation of the relative efficiency of a sample of Portuguese and Greek seaports. The broad purpose of this exercise was to facilitate benchmarking so that areas for improvement to management practices and strategies could be identified and, within the context of European ports policy, improvements implemented within the seaport sectors of these respective countries. The results of the analysis pointed to the inefficiencies of particular ports and appropriate benchmarks by which ports could improve performance. Scale efficiency was recommended as the paramount objective for the ports under study and privatisation was advocated as the most appropriate method for achieving economic efficiency. In order to examine the pros and cons of applying alternative non-parametric approaches (including DEA) to the container port industry, Cullinane et al. (2006) applied both DEA and the free disposal hull (FDH) model to 57 of the world s leading container ports, and found that the available mathematical programming methodologies led to different conclusions and that the definition of input and output variables was a crucial element in meaningful applications of DEA and FDH. All the studies outlined above share the common property that only DEA approaches for analysing cross-sectional, rather than panel, data were used. This is despite the fact that panel data have occasionally been utilised within these studies (e.g. Martinez-Budria et al. 1999). In such cases, although panel data had been collected, in the ensuing analysis they were treated only as though they were actually cross-sectional data (i.e. dynamic time-based changes in relative efficiency levels have not been explicitly investigated or isolated). Realising that DEA cross-sectional data approaches might be inferior to those utilising panel data and given the characteristics of the container port industry and the random effects associated with each single measured value of production and the level of measured efficiency associated with it, Cullinane et al. (2004a, 2005) applied a range of alternative DEA panel data approaches to deriving the efficiency of container ports. In so doing, the development of the efficiency of each container port in the sample could be tracked over time and, in consequence, the efficiency results are ostensibly more convincing. Among the applications of SFA to the port industry, Liu (1995) set out to test the hypothesis that public sector ports are inherently less efficient than those in the private sector.a set of panel data relating to the outputs and inputs of 28 commercially important UK ports over the period was collected for analysis. The results failed to identify ownership as a significant factor in production and the evidence did not imply any clear-cut efficiency advantage for any particular form of ownership. Notteboom et al. (2000) applied a Bayesian approach based on Monte-Carlo approximation to the estimation of a stochastic frontier model aimed at assessing the productive efficiency of a sample of 36 European container terminals located in the Hamburg Le Havre range and in the western Mediterranean. The data analysed relate to the single year of The robustness and validity of the estimated model were tested by comparing the results with

5 22 K. P. B. Cullinane and T.-F. Wang those of four benchmark terminals in Asia (Singapore, Kaohsiung and Hong Kong s MTL and HIT terminals). Coto-Millan et al. (2000) applied a stochastic frontier model to estimate the economic efficiency of 27 Spanish ports. Panel data for the period were collected and analysed using the Cobb Douglas and Translog versions of the model. Cullinane et al. (2002) analysed the administrative and ownership structures of major container ports in Asia. The relative efficiency of these ports was then assessed using the cross-sectional and panel data versions of the stochastic frontier model. Finally, Cullinane and Song (2003) assessed the success achieved by Korea s port privatisation policies in increasing the productive efficiency of its container terminals. The stochastic frontier model was justified as the chosen methodology for estimating productive efficiency levels and was applied to cross-sectional data under a variety of distributional assumptions. A panel data model was also estimated. In line with the argument that comparable data can only be obtained at the level of the container terminal, rather than the port (Goss 1990, Heaver 1995, Alderton 1999, Heaver et al. 2000, 2001), this paper focuses on measuring the efficiency of container terminals in Europe using the DEA approach. As a basis for enhancing the literature in this sphere, this paper serves to supplement existing studies by deriving estimates of relative efficiency for a sample comprising 69 of Europe s container terminals with throughput of over 10,000 TEUs. The scale properties of container terminal production are also discussed. 3. Data envelopment analysis DEA can be roughly defined as a non-parametric method for measuring the efficiency of a decision-making unit (DMU) with multiple inputs and/or multiple outputs. This is achieved by constructing a single virtual output to a single virtual input without predefining a production function. The term DEA and the CCR model were first coined by Charnes et al. (1978), whose work has been followed by a phenomenal expansion of DEA in terms of its theory, methodology and application over the last few decades. The influence of the CCR paper is reflected in the fact that, by 1999, it had been cited over 700 times (Forsund and Sarafoglou 2002). Among other models in the context of DEA, the two most widely used DEA models, named DEA-CCR and DEA-BCC, deserve greater attention, especially since they are utilised later in this paper. The DEA-CCR model assumes constant returns to scale so that all observed production combinations can be scaled up or down proportionally. The DEA-BCC model, on the other hand, allows for variable returns to scale and is graphically represented by a piecewise linear convex frontier. The standard formal mathematical specification of the various DEA models can be found in Cooper et al. (2000). For the purposes of the analysis contained herein, however, it is sufficient to state that the output-oriented DEA model seeks to maximise the proportional increase in output while remaining within the production possibility set. An output-oriented efficiency measurement problem can be written as a series of K linear programming envelopment problems, with the constraints differentiating between the DEA-CCR and DEA-BCC models. This threshold was adopted in order to ensure an appropriate balance between the number of ports in the sample population and their representativeness with respect to the proportion of Europe s total container throughput for which they accounted. A twenty-foot equivalent unit (TEU) is a standard size of container, typically used for denoting the output or capacity of container ports or terminals or for defining the container carrying capacity or loading of ships.

6 Efficiency of European container ports 23 The technical efficiencies derived from the DEA-CCR and DEA-BCC models are frequently used to obtain a measure of scale efficiency for the kth DMU, given by SE k = U CCR_k /U BCC_k. A value of SE k = 1 indicates scale efficiency and SE k < 1 indicates scale inefficiency (Cooper et al. 2000). Scale inefficiency is due to either increasing or decreasing returns to scale, which can be determined by inspecting the sum of weights, eλ, under the specification of the CCR model. If this sum is equal to one, the law of constant returns to scale prevails, whereas increasing returns to scale and decreasing returns to scale prevail when the sum is, respectively, less than, or greater than, unity. 4. Definition of variables and data The scientific definitions of input and output variables are critical to the application of DEA. The specification of erroneous or ill-defined variables inevitably leads to the wrong conclusions emerging, however elaborate the models employed may be. Input and output variables should reflect the actual objectives and process of container port production as accurately as possible (Norman and Stoker 1991, Wang 2004). As far as the former is concerned, the observed performance of a port is closely related to its objective. In this paper, the main objective of a port is assumed to be the minimisation of the use of input(s) and maximisation of the output(s). Because of the extreme difficulty of obtaining confidential data that are often inconsistent between corporate entities, nations, etc. the respective prices of inputs or outputs are not taken into account in the empirical analysis contained herein. In consequence, this assumed objective may not be entirely consistent with that of profit maximisation. However, the argument can be made that the capital intensity of a container port, handling a largely homogeneous throughput with only small variation in the price charged per container, means that its profits do stem mainly from handling sufficiently large volumes to cover the significant fixed cost of investment. As far as the output variable of container terminal production is concerned, container throughput is unquestionably the most important and widely accepted indicator of container port or terminal output. Almost all previous studies have treated it as an output variable, because it relates closely to the need for cargo-related facilities and services and is the primary basis upon which container ports are compared, especially in assessing their relative size, investment magnitude or activity levels. Most importantly, it also forms the basis for the revenue generation of a container port or terminal. On the other hand, a container terminal depends crucially on the efficient use of labour, land and equipment. The total quay length, the terminal area, the number of gantry cranes, the number of yard gantry cranes and the number of straddle carriers have been deemed to be the most suitable factors to be incorporated into the models as input variables. Other input factors that possibly influence the efficiency estimates that may be derived from this analysis include aspects such as berth occupancy, berth accessibility, proximity to major trade lanes, crane operating hours, different handling speeds of yard and ship-to-shore cranes, equipment age and maintenance, the capital invested in a terminal and associated equipment, average container interchange per ship and quayside water depth. However, the practical problem of obtaining data on each of these variables across the whole sample is likely to prove insurmountable. In addition, with the vast number of potential input variables that may be hypothesised as influencing container port efficiency, the issue of multicollinearity becomes salient. In the light of the unavailability or unreliability of direct data, information on labour inputs is derived from a predetermined relationship to terminal facilities (De Neufville and Tsunokawa 1981, Notteboom et al. 2000). It is very important to note, however, that this predetermined

7 24 K. P. B. Cullinane and T.-F. Wang Table 1. Summary statistics for the sample. Output Input Container Terminal Terminal Equipment throughput (TEU) length (m) area (ha) (number) Mean 426, Standard error 69, Median 233, Mode 246, Standard deviation 580, Kurtosis Skewness Range 2,876,699 3, Minimum 20, Maximum 2,896,835 3, Sum 29,397,391 64,389 2,586 2,550 Count relationship is not applicable to all types of ports with different characteristics of production. It is also dangerous to apply this relationship to container ports of different production scale (throughput) because of the different equipment and labour arrangements employed. The final sample included in the analysis comprised 69 of Europe s container terminals distributed across 24 countries with throughput of over 10,000 TEUs. The required secondary data for a cross-sectional analysis for the year 2002 were taken mainly from various issues of both the Containerisation International Yearbook and Lloyd s Ports of the World. To solve the problem of data unavailability for some terminals, a questionnaire was sent to these container terminals and the data collected in this way were also incorporated into the final sample. Important statistics relating to the sample are summarised in table Empirical analysis The software DEA-Solver-PRO 3.0 (Cooper et al. 2000) was employed to solve the DEA models. Without precise information on the returns to scale of the port production function, two types of DEA models, namely the CCR and BCC models, were applied to analyse the efficiency of the sample container terminals. In line with formulae (1) (7) with the relationship (discussed in section 3) between scale inefficiency and the sum of weights, eλ, under the specification of the CCR model, table 2 reports the efficiency estimates, the scale efficiency and scale properties of each container terminal. It is clear from table 2 that, as one would expect, the DEA-BCC model yields higher average efficiency estimates than the DEA-CCR model, with respective average values of 0.48 and 0.42, and where an index value of 1.00 equates to perfect (or maximum) efficiency. Respectively, nine and four out of the 69 terminals (ports) included in the analysis were identified as efficient when the DEA-BCC and the DEA-CCR models were applied. The result that the DEA- BCC model yields more efficient container terminals is not surprising since a DEA model with an assumption of constant returns to scale provides information purely on technical and scale efficiency taken together, while a DEA model with the assumption of variable returns to scale identifies technical efficiency alone. An analysis of variance (ANOVA) of the efficiency for the DEA-BCC and DEA-CCR analyses (F = 1.95) indicated that the efficiency measures calculated using these two different approaches were not significantly different at

8 Efficiency of European container ports 25 Table 2. Terminal efficiency under the DEA-CCR and DEA-BCC models. Scale Returns Country Port Terminal DEA-CCR DEA-BCC efficiency to scale Austria Krems Weincont Krems Increasing Kombiterminal Vienna Weincont Krems Constant Kombiterminal Bulgaria Varna Varna Increasing Canary is Las Palmas La Luz Terminal Constant Leon y Castillo Dock Decreasing Denmark Aarhus Container terminal Increasing Estonia Tallinn Muuga Container Increasing Terminal Finland Hamina Container terminal Increasing Kotka Mussalo Container Increasing terminal France Dunkirk West harbour Constant Le Harve Europe atlantique terminal Decreasing Terminal Europe Decreasing Ameriques Lille Lille Increasing Mulhouse- Mulhouse-Ottmarsheim Increasing Ottmarsheim Rouen Rouen Increasing Strasbourg Strasbourg Increasing Germany Bremen/ Eurogate Container Decreasing Bremerhaven Terminal North Sea Terminal Constant Hamburg Eurogate Container Decreasing Terminal Unikai Contaier Terminal Decreasing Mannheim Mannheim Increasing Greece Piraeus Venizelos Container Decreasing terminal Thessaloniki Container terminal Constant Hungary Budapest Mahart Container terminal Increasing Ireland Cork Tivoli Container Terminal Increasing Italy Genoa Messina terminal Decreasing Southern European Constant Container hub terminal Voltri Terminal Constant Gioia Tauro Medcenter Container Constant terminal La Spezia La Spezia Container Decreasing terminal Leghorn Darsena Toscana terminal Decreasing Sintermar/sogeter Decreasing terminal porto Increasing commerciale Ravenna terminal Container Decreasing Ravenna spa Salerno terminal Container Decreasing Ravenna spa Taranto Container terminal Increasing Venice Container terminal Increasing Latvia Riga Baltic Container terminal Increasing Netherlands Born Born Increasing Rotterdam APM Constant HT Holland Terminal Increasing Poland Gdansk Gdanski Terminal Increasing Konntenerowy SA Gdynia Baltic Container terminal Increasing (continued)

9 26 K. P. B. Cullinane and T.-F. Wang Table 2. Continued. Scale Returns Country Port Terminal DEA-CCR DEA-BCC efficiency to scale Portugal Leixoes Leixoes Decreasing Lisbon Alcantara-sul terminal Decreasing Santos Quay Increasing Romania Constantza Socep container terminal Increasing Umex container terminal Increasing Russia St Petersburg First Container Terminal Decreasing Vostochiniy VICS terminal Increasing Slovenia Koper Koper Increasing Spain Algeciras TCB terminal Constant Bilbao Abra Terminal Maritimas Increasing TMB terminal Decreasing Tarragona Container terminal Increasing Valencia TCV Decreasing Terminales del turia Increasing Valencia Container Decreasing terminal Sweden Stockholm Stockholm Increasing Turkey Haydarpasa Haydarpasa Decreasing Izmir Izmir Increasing Kumport Kumport Decreasing UK Felixstowe Landguard and Trinity Decreasing Goole Aldam terminal Increasing Liverpool Liverpool Constant Southampton Southampton container Constant terminal Thamesport Thamesport Constant Tilbury Short Sea container Increasing terminal Ukraine Odessa Container terminal Decreasing Average Alternative efficiency results and returns to scale are calculated by equations (1) (7). The estimates presented here are based on output-orientated DEA-CCR and DEA-BCC models. 1 = efficient. the 5% level (with a critical value of 3.91). A Spearman s rank order correlation coefficient between the efficiency rankings derived from DEA-BCC and DEA-CCR analyses was The positive and high Spearman s rank order correlation coefficient indicated that the rank of each firm derived from applying the two different models was similar. A combination of ANOVA and Spearman s rank order correlation coefficient leads to the conclusion that the efficiency estimates yielded by the two approaches are similar and follow the same pattern across firms. Empirical results reveal that there exists substantial waste in the production of the container ports in the sample. For instance, the average efficiency of container ports derived from applying the DEA-CCR model amounts to This indicates that, in theory, the ports under study can, on average, dramatically increase the level of their outputs to 2.4 (=1/0.42) times as much as their current level while using the same inputs. This is dependent, however, on the appropriate approaches to production being implemented and the appropriate scale of production adopted. As far as individual container terminals are concerned, table 2 also reports the returns to scale properties of port production yielded by DEA. Of the 69 terminals (ports), 13 exhibit constant returns to scale, 33 exhibit increasing returns to scale and 23 have decreasing returns to scale. Among those large terminals (classified as having annual container throughput of more than 1 million TEU), all five scale-inefficient terminals show decreasing returns to scale.

10 Efficiency of European container ports 27 Among those medium size terminals (classified as having annual container throughput of less than 1 million TEU and more than 100,000 TEU), scale-inefficient terminals (ports) exhibit a reasonably even mix of both increasing and decreasing returns to scale (16:17). Finally, most small terminals (classified as having annual container throughput of less than 100,000 TEUs), evaluated as scale-inefficient, exhibit increasing returns to scale. In seeking to interpret or explain these results on economies of scale, a certain degree of caution must be exercised. This is because of the lumpiness of investment in port infrastructure. Investments of significant capital sums are made very infrequently and, in an effort to cater for future growth in demand, often have the impact of expanding capacity to levels well in excess of what may be currently required. It is an obvious tautology to suggest that this is more likely to be the case for large, successful and expanding ports, rather than for smaller ports. As such, this is a potential shortcoming of any cross-sectional analysis and provides support for an approach based on panel data that may capture the dynamics associated with this feature of the industry. In figure 1, the average tendency for the relationship between efficiency scores and production scale (as measured by container throughput) is plotted. It reveals that a large scale of production is more likely to be associated with high efficiency scores. Since the correlation coefficients of the mean efficiency scores against container throughput are, respectively, 0.60 and 0.58 for the DEA-BCC and DEA-CCR models, it would appear that the efficiency of a terminal is significantly influenced by its production scale and that there is evidence to support the existence of economies of scale in the sector. This is not surprising considering the fact that large terminals are more likely to utilise more state-of-the-art equipment and sophisticated management than their smaller counterparts. This finding is consistent with that of the pioneering work conducted by De Neufville and Figure 1. Relationship between efficiency and production scale.

11 28 K. P. B. Cullinane and T.-F. Wang Figure 2. Efficiency of container ports in different regions in Europe. Note that the countries where the ports under study are located are roughly divided into seven regions. They are Scandinavia (Denmark, Finland and Sweden), the British Isles (Ireland and the UK), West Europe (France, Netherlands), South Europe (Portugal, Spain, Canary Isles and Italy), Central Europe (Austria, Germany, Hungary and Poland), Southeast Europe (Bulgaria, Greece, Romania, Slovenia and Turkey) and East Europe (Estonia, Latvia, Ukraine and Russia). Tsunokawa (1981), who analysed a sample of a mere five container ports in the USA over the time period and found that the production of container ports had a tendency to follow the law of increasing returns to scale. The average efficiency of container terminals located in different regions of Europe is found to differ to a large or small extent. Figure 2 shows the average efficiency of container ports in different regions and reveals that the average efficiency of ports in the British Isles is higher than that of the other regions, while container terminals in Scandinavian and eastern European countries have the lowest average efficiency. This result should be viewed with great caution, bearing in mind that some important container terminals were not included in the sample due to data unavailability and that these missing terminals are likely to exert either a positive or negative influence on the efficiency estimates of those that remain in the sample. Nevertheless, this finding is interesting in that such a difference might indicate that more complicated factors exert a significant influence over the efficiency of a region s ports. Such region-specific influences might include the level of general economic development (Miyashita 2005), the comparative competence of management, organisational style/culture, accessibility to/from major trade lanes (Tongzon 2005), or simply the level of competition or geographical proximity between ports within a particular region (Cullinane 2000, Cullinane et al. 2004b). 6. Conclusions In this paper, DEA analysis has been applied to determine the relative efficiency of Europe s leading container terminals. To this end, the fundamental properties of DEA and the definitions of input and output variables with reference to the characteristics of container terminal production have been addressed. Data for the year 2002 on 69 leading container terminals with an annual container throughput over 10,000 TEUs distributed across 24 European

12 Efficiency of European container ports 29 countries were finally collected for the estimation of individual efficiency scores for each port/terminal. The primary finding of this paper includes the significant inefficiency that generally pervades most of the terminals under study. The average efficiency of container terminals under study amounts to 0.48 (assuming constant returns to scale) and 0.42 (assuming variable returns to scale). The latter, more realistic, figure indicates that, on average, terminals can dramatically improve the level of their output by up to 2.4 times as much as current levels while using the same inputs. The comparatively large sample analysed in this study has, it is ventured, yielded low relative efficiency scores compared with the findings of previous studies. However, as sample size increases this phenomenon is to be wholly expected. One major benefit of large sample studies, however, is that the efficiency estimates derived from them are likely to be more consistent and robust. Therefore, efficiency estimates are less sensitive to change (subject to maintaining a similar sample size) when the constituency of the sample is changed. There may be other reasons that contribute to explaining this phenomenon, but this question remains subject to further investigation and provides an interesting avenue for further research. For the purposes of this study, the scale properties of container terminal production have also been analysed. It was found that, while some container terminals are scale-efficient, in general, most of the container terminals under study exhibit increasing returns to scale. However, most of the container terminals that are large in production scale are more likely already to be associated with higher efficiency scores. These findings are particularly informative for policy-makers and corporate decision-makers. For example, these findings provide some theoretical support for the increasing tendency towards the construction of large-scale container ports (mega-ports) that is progressing world-wide, These findings, however, also suggest that not every individual container terminal (even ones that are currently small) follows the law of increasing returns to scale. Decision-makers, both commercial and political, will need to study carefully, therefore, their own particular set of circumstances and general situation. It has also been found that the average efficiency of container terminals located in different regions differs, either to a large or to a small extent. The container terminals in the British Isles were found to be the most efficient compared with their counterparts, with the lowest efficiency in Scandinavian and eastern European countries. A further investigation into the reasons behind the relative (in)efficiency of these terminals will be both interesting and complicated. On the one hand, to find the reasons behind (in)efficiency is a prerequisite for obtaining any further efficiency improvement. On the other hand, the possible reasons behind (in)efficiency (such as the level of economic growth within the region under study, the port governance structure, extent of port competition and private sector participation, etc.) can be so wideranging, difficult to quantify and interdependent that an investigation along these lines might prove to be extremely challenging. From this point of view, the derivation of the efficiency estimates for the sample of container terminals analysed in this paper constitutes merely a beginning, rather than an end in itself. Acknowledgements The authors are grateful to the Chartered Institute of Logistics & Transport for providing partial funding for this project in the form of the 2004/05 Logistics and Transport Seed- Corn Research Grant. The very helpful comments on an earlier version of this paper by three anonymous referees are also very much appreciated.

13 30 K. P. B. Cullinane and T.-F. Wang References Alderton, P., Port Management and Operation, 1999 (LLP Reference Publication: London). Ashar, A., Counting the moves. Port Development Int., 1997, Nov., Banker, R.D., Charnes, A. and Cooper, W.W., Some models for estimating technical and scale inefficiencies in data envelopment analysis. Mgmnt Sci., 1984, 30, Barros, C.P. and Athanassiou, M., Efficiency in European seaports with DEA: evidence from Greece and Portugal. Mar. Econ. Logistics, 2004, 6, Baumol, W., Panzar, J. and Willig, R., Contestable Markets and the Theory of Industry Structure, 1982 (Harcourt, Brace, Jovanovich: New York). Bendall, H. and Stent, A., On measuring cargo handling productivity. Mar. Policy Mgmnt, 1987, 14, Charnes, A., Cooper, W.W. and Rhodes, E., Measuring the efficiency of decision making units. Eur. J. Operational Res., 1978, 2, Cooper, W.W., Seiford, L.M. and Tone, K., Data Envelopment Analysis: A Comprehensive Text with Models, Applications, References and DEA-Solver Software, 2000 (Kluwer Acaemic: Boston). Coto-Millan, P., Banos-Pino, J. and Rodriguez-Alvarez, A., Economic efficiency in Spanish ports: some empirical evidence. Mar. Policy Mgmnt, 2000, 27, Cullinane, K.P.B., The competitive position of the port of Hong Kong. J. Korean Ass. Shipping Stud., 2000, 31, Cullinane, K.P.B., Ji, P. and Wang, T., The relationship between privatization and DEA estimates of efficiency in the container port industry. J. Econ. Business, 2005, 57, Cullinane, K.P.B. and Song, D.W., A stochastic frontier model of the productive efficiency of Korean container terminals. Appl. Econ., 2003, 35, Cullinane, K.P.B., Song, D.W. and Gray, R., A stochastic frontier model of the efficiency of major container terminals inasia: assessing the influence of administrative and ownership structures. Transport. Res. A, 2002, 36, Cullinane, K.P.B., Song, D.-W., Ji, P. and Wang, T., An application of DEA Windows analysis to container port production efficiency. Rev. Net. Econ., 2004a, 3, Cullinane, K.P.B., Song, D.-W. and Wang, T., A comparison of mathematical programming approaches to estimating container port production efficiency. J. Prod. Anal., 2006, 24, Cullinane, K.P.B., Wang, T. and Cullinane, S.L., Container terminal development in Mainland China and its impact on the competitiveness of the port of Hong Kong. Transport Rev., 2004b, 24, De Monie, G., Measuring and evaluating port performance and productivity. In UNCTAD Monographs on Port Management No. 6 on Port Management, 1987 (UNCTAD: Geneva). De Neufville, R. and Tsunokawa, K., Productivity and returns to scale of container ports. Mar. Policy Mgmnt, 1981, 8, Forsund, F.R. and Sarafoglou, N., On the origins of data envelopment analysis. J. Prod. Anal., 2002, 17, Goss, R.O., Economic policies and seaports part 2: the diversity of port policies. Mar. Policy Mgmnt, 1990, 17, Heaver, T., The implications of increased competition among ports for port policy and management. Mar. Policy Mgmnt, 1995, 22, Heaver, T., Meersman, H., Moglia, F. and van de Voorde, E., Do mergers and alliances influence European shipping and port competition? Mar. Policy Mgmnt, 2000, 27, Heaver, T., Meersman, H. and van de Voorde, E., Co-operation and competition in international container transport: strategies for ports. Mar. Policy Mgmnt, 2001, 28, Liu, Z., The comparative performance of public and private enterprises: the case of British ports. J. Transport Econ. Policy, 1995, 29, Martine Zarzoso, I., Perez Garcia, E. San Juan Lucas, E. and Suarez Burguet, C., How important are transport costs for international trade? An empirical study for Spanish exporting sectors, in International Association of Maritime Economists Annual Conference, Martinez-Budria, E., Diaz-Armas, R., Navarro-Ibanez, M. and Ravelo-Mesa, T., A study of the efficiency of Spanish port authorities using data envelopment analysis. Int. J. Transport Econ., 1999, XXVI, Miyashita, K., Logistics strategy of Japanese ports: the case of Kobe and Osaka. In World Shipping and Port Development, edited by T.-W. Lee and K.P.B. Cullinane, pp , 2005 (Palgrave-Macmillan: Basingstoke). Norman, M. and Stoker, B., Data Envelopment Analysis, The Assessment of Performance, 1991 (John Wiley & Sons: Chichester). Notteboom, T., Concentration and load centre development in the European container port system. J. Transport Geog., 1997, 5, Notteboom, T., Coeck, C. and van den Broeck, J., Measuring and explaining the relative efficiency of container terminals by means of Bayesian stochastic frontier models. Int. J. Mar. Econ., 2000, 2, Roll,Y. and Hayuth,Y., Port performance comparison applying data envelopment analysis (DEA). Mar. Policy Mgmnt, 1993, 20, Tabernacle, J.B., A study of the changes in performance of quayside container cranes. Mar. Policy Mgmnt, 1995, 22, Talley, W.K., Optimum throughput and performance evaluation of marine terminals. Mar. Policy Mgmnt, 1998, 15, Tongzon, J., Efficiency measurement of selected Australian and other international ports using data envelopment analysis. Transport. Res. Part A, 2001, 35,

14 Efficiency of European container ports 31 Tongzon, J., Key success factors for transhipment hubs: the case of the port of Singapore, in World Shipping and Port Development, edited by T.-W. Lee and K.P.B. Cullinane, pp , 2005 (Palgrave-Macmillan: Basingstoke). Valentine, V.F. and Gray, R., The measurement of port efficiency using data envelopment analysis, in Proceedings of the 9th World Conference on Transport Research, Wang, T., Analysis of the container port industry using efficiency measurement: a comparison of China with its international counterparts. PhD thesis, The Hong Kong Polytechnic University, Wang, T. and Cullinane, K., Industrial concentration in container ports, in International Association of Maritime Economists Annual Conference, Winkelmans, W., The socio-economic value of seaports in the European Union, in The European Sea Ports Conference 2004: European Sea Ports in a Dynamic Market Ports and the EU Agenda, 2004.