CHAPTER V: OPERATIONALISATION
V.3 Sampling Frame, Dataset and Variable Selection
V.3.2 Data and Variables
In this study, the choice of variables is based on a high-level aggregation of container-terminal operations with a view to utilizing available and reliable data on operational performance and ensuring homogeneity between observation units. Where relevant, a second set of key performance indicators, namely the STS-crane move per hour, the free yard storage time, the cargo dwell time, and the gate cut-off time, is added to the dataset. Micro-performance indicators such as those related to scheduling, allocation, routing, and stacking policies are too detailed and terminal-specific for inclusion in a benchmarking exercise of productive efficiency. Furthermore, such data are hardly available outside terminal management.
Earlier in Chapters III and IV, we pointed out the shortcomings of the port benchmarking literature in incorporating the operating typologies and configurations of container ports and terminals. A typical manifestation of the gap between container-port practice and theory is the rather subjective definition and selection of input and output variables. For instance, most researchers include the number of quay and yard cranes as input variables but each crane category depicts a different production technology and operating configuration. To incorporate these differences, we define structured sets of input variables that account for the variations in crane technology and cargo handling operation:
A. As shown from the discussion in the previous chapter, STS cranes depict different operating configurations such as the gauge, the outreach, the back-reach, the lift capacity and the height. These parameters are usually proportional to the type and size of vessels serviced but they operate on speedier cycle times (hoist and trolley speed) so that standard operational benchmarks of crane move per hour can be achieved. Because large vessels have an extended outreach, the average cycle time of STS cranes operating them must be increased substantially in order to achieve comparable productivity levels to those of STS cranes handing smaller vessels (see tables 14 and 15). In addition to the cycle time parameter, the lifting capability is another key performance indicator for STS cranes. Modern cranes have a higher load capacity and are equipped with several extendable spreaders, which allow them to handle multi-container picks (e.g. twin and tandem lifts) in a single move. Therefore, performance data on both cycle time and lifting capability must be included in the crane input variable in order to capture the productive technology of STS cranes.
For the cycle time, one can capture its performance directly from the rate of crane move per hour, the latter being an additional output variable used in this study. For the lifting capability, we use industry data provided by terminal operators. When such information is unavailable, we use data from industry surveys on STS-crane delivery (see for instance Cargo systems, 2007a; 2008a) as well as data on crane engineering standards as compiled from global crane manufacturers. Our index for capturing STS-crane input is therefore expressed as follows:
STS Crane’s index = Number of cranes * Lifting capacity Lifting Capability index (in TEU):
• Conventional Technology 20ft = 1
• Twin 20ft = 2
• Tandem 40ft = 2
• Two tandem = Two Twin 20ft = 4
• Triple 40ft = 6
Table 14: Relationship between STS-crane speed and productivity -data based on average productivity of 25-30 moves per hour- (Source: Bhimani and Sisson, 2002)
Crane Generation Outreach (meter)
Lift Height (meter)
Hoist speed Trolley speed
MPM Ratio MPM Ratio
Panamax 35 24 48 1 150 1
Post-Panamax 44 29 55 1.15 180 1.2
Super-post Panamax 50 33 61 1.14 245 1.35
Malacca-max (22 wide) 65 40 90 1.88 300 2
Table 15: Relationship between STS-crane productivity and vessel turnaround time (Source: Bhimani and Sisson, 2002)
Crane productivity (move per hour)
Turnaround time in hours per vessel size
6000 TEU 8000 TEU 10000 TEU 12000 TEU
25-30 60 64 72 85
35-40 45 48 52 66
50 35 38 44 51
60 30 32 36 45
B. For yard handling equipment, we refer to the handling configurations described in Chapter IV and construct an index for yard stacking equipment based on two operational features namely the ground storage capacity (in TEU) and the staking height. These are the main performance data used by industry for container yard stacking equipment (Cargo systems, 2007b; 2008b). Information on yard equipment operational features is usually provided by the websites of terminal operators but can also be sourced from trade journals or from the manufacturers’ reference list of yard crane deliveries.
Stacking equipment index = Yard equipment * Ground storage capacity * Stacking capacity
The definition and selection of other variables follow the same reasoning. Variables should be practical and consistent with both the objectives of this research and the results of IDEF0 modelling. Variables selected for benchmarking container terminal operations consist of seven inputs and one output. The input variables are terminal area, maximum draft, length overall (LOA), STS-crane index, yard-stacking index, internal trucks and vehicles, and number of gates (or gate lanes). The output variable is terminal throughput in TEU. Additional variables used for benchmarking site and network efficiency are the free yard storage time and the gate cut-off time as inputs, and the STS-crane move per hour and the cargo dwell time as outputs.
Table 16: Input and output variables for container terminal operations
Variables Descriptions Units of
measurement Site INPUTS
Terminal area Total terminal area in square meters 1000 m2 Terminal
Maximum draft Maximum draft in the terminal Meter Quay
Length overall
(LOA) Total quay length in meter Meter Quay
Quay crane index STS crane index
= Lifting Capability * STS Cranes TEU Quay
Yard stacking index
Yard equipment stacking index
= staking height *storage capacity *Yard Equipment TEU /1000 m2 Yard Trucks &
Vehicles
Internal trucks, tractors and other supporting vehicles
Number of
vehicles Terminal Number of gates Number of gates, gate lanes, and/or railway tracks at
the gate Number Gate
OUTPUT Terminal
Throughput Annual total throughput 1000 TEU Terminal
The dataset consists of annual observations of sampled container terminals and spans the period from 2000 to 2006. This is because many container terminals have started implementing the new security regulations as early as 2004 and we wanted to select a reasonable observation period that would allow us assess productivity changes before and after the introduction of security measures. The collection of data observations over a 7-year time-span resulted in a panel data of 420 terminal-years. In a dynamic context, panel data prevail over times-series and cross-sectional data. On the one hand, because a DMU is observed only once in either the times-series or the cross-sectional analysis, its efficiency estimate would be subjected to a higher degree of randomness and, therefore, may be misleading. On the other hand, the increase of the sample size under panel data analysis (from 60 to 420) would reinforce analytical reliability and reduce statistical error. In a panel data analysis, a DMU is defined as a container terminal-year, for instance HIT-2003.
Regarding the data collection methods, we used both primary and secondary data sources, mainly the latter source:
- Primary data is sourced directly from the terminals under study using a standard on-line questionnaire as shown in Appendix 13. However, only 15 responses were received, and secondary data was used for the rest of terminals in the sample.
- Secondary data was sourced from the websites and annual reports of port and terminal operators in the sample as well as from subscribed databases of trade journals namely Containerisation International yearbooks for the period 2000-2006, Containerisation International On-line website, Cargo World, World Port Focus, and the Fairplay database of container ports and terminals.
- We also relied on the information reported on the websites of global carriers and shipping lines, particularly the information on free-time demurrage and detention at the yard, and gate procedures and cut-off time. We verified and crosschecked information from all these sources. Where inconsistency arises, we record information from primary sources if data is available, otherwise from the website of sampled ports and terminals.
The combination of 60 terminals, 8 variables, and a 7-year (2000-2006) timeframe has resulted into a container-terminal panel dataset of 420 DMUs and 3360 data points.
Table 17 depicts a summary of descriptive statistics relative to the aggregate container terminal dataset.
Table 17: Descriptive statistics of the aggregate container terminal dataset
Variable Minimum Maximum Mean Standard
Deviation
Terminal area (1000 m2) 105 2650 730 505
Maximum Draft 10 18 14 2
LOA 305 4875 1515 993
STS-crane index 2 390 55 57
Yard stacking index 6 212 35 35
Internal trucks and vehicles 2 390 55 57
Gates 3 37 10 7
Terminal throughput (1000 TEU) 123 8865 1526 1465