Transpn. Rcs.-A. Vol. 29A, No. 3, pp.

245-252, 1995
Copyright 0 1995 ElsevierScienceLtd
Pergnmoa Printed in Great Britain.All ri8htrreserved
0965-8SW95$9.50 + .OO

0965-8564(94)00032-8

DETERMINANTS OF PORT PERFORMANCE AND

EFFICIENCY

JOSE L. TONGZON
Department of Economics and Statistics, National University of Singapore,
Kent Ridge Crescent, Singapore 05 11

(Received 4 March 1994; in revisedform 14 June 1994)

Abstract-Studies that have identified the various factors influencing a port’s performance and
efficiency are few. Furthermore, these studies have failed to quantify the relative contributions of
these factors to overall port performance and efficiency. This article attempts to fill this gap by
establishing a model of port performance and efficiency. Specifically, it aims to specify and
empirically test the various factors which influence the performance and efficiency of a port.
Based on a sample of 23 international ports, the study inter alia is able to provide an empirical
basis for the crucial role of terminal efficiency relative to other factors in the overall port perfor-
mance.

1. INTRODUCTION

1. I. Objectives and rationale

A number of studies on the subject of port performance and efficiency have compared
some selected ports using certain performance and efficiency criteria, for example, studies
by the Australian Bureau of Industry Economics (1993), Australian Transport Advisory
Council (1992), Marconsult ( 1991), and Australian Business Council (1988). But only
few have attempted to explain and identify the various factors underlying a port’s perfor-
mance and efficiency. Furthermore, these studies have failed to quantify the relative
contributions of these factors to overall port performance and efficiency.
This article attempts to fill this gap in the current literature by establishing a model
of port performance and efficiency. Specifically, the model seeks to specify and empiri-
cally test the underlying factors that influence port performance and efficiency.
This study is significant because there is scant literature on this subject of port
performance and efficiency, and studies carried out so far on port performance and
efficiency have overall been constrained partly by the lack of up-to-date and reliable data
(UNCTAD, 1987). ’ Second, as nations are becoming more “global” and their industries
more exposed to the pressures of international competition, there is a growing realisation
that services supplied to their industries must be provided on an internationally competi-
tive basis. Thus, there is a push amongst port authorities to improve their port perfor-
mance and efficiency due to increasing competition between ports and growing pressure
from shippers for lower port and shipping charges. Needless to say, ports form a vital link
in the overall trading chain and consequently, their level of efficiency and performance
determines to a large extent a nation’s international competitiveness. However, in order
to achieve and maintain a competitive edge in the international markets, a nation needs
to understand the underlying factors of port competitiveness, and continually assess its
performance relative to the rest of the world so that appropriate business strategies can
be devised.

1.2. Selection of ports

Given the multiplicity of ports and cargoes handled, it is necessary to restrict the
scope of the analysis to a limited number of ports and a specific type of cargo. This study
‘Moreover.due to the unavahabihty of data across ports, most studies have attempted to explain the
differences in port performance by observing the performance of a port over time using time-series data. gee,
for example, Tongzon (1993).
246 J. L. TONOZON

examines only the performance in handling containerised cargoes across selected ports. ’
A sample of 30 container ports was initially selected based on size, geographical location,
and data availability. Questionnaires requesting 1991 data on port performance and
efficiency were sent out to these selected ports. Due to confidentiality reasons and data
nonavailability, 23 sampled ports have responded. The survey method was supplemented
with data from secondary sources. Two major secondary sources deserve special mention:
Port of Melbourne inhouse study on Comparative Port Study (1992) and the newly
published report by Australia’s Bureau of Industry Economics on International Perfor-
mance Indicators in the Waterfront (1993). Publications such as Lloyd’s Ports of the
World (1993) and Containerisation International Yearbook ( 1992) also provided infor-
mation. 3 Appendix 1 presents the summary statistics for all the relevant variables consid-
ered in this study.
Given data availability, the final sample of ports analysed are listed in Table 1,
together with their corresponding countries.

2. THE MODEL

Port performance is measured in terms of the number of containers moved through

a port (throughput) on the assumption that ports are throughput maximisers.’
There are several ways to measure port efficiency depending on which aspects of the
port operation are being evaluated. This article looks at the terminal operation aspect
which is measured in terms of the number of containers loaded and unloaded while a ship
is at berth. This aspect of terminal operation constitutes the largest component of the
total ship turnaround time. And to improve efficiency in this area is also consistent with
port authorities’ intention to maximize berth utilization. From the viewpoint of ship
operators and shippers in several countries, berth rentals have increasingly become a

Table 1. List of sampled ports

Name of Ports Country

1. Port of Adelaide Australia

2. Port of Auckland New Zealand
3. Port of Baltimore United States
4. Port of Bangkok Thailand
5. Port of Bombay India
6. Port of Brisbane Australia
7. Port of Charleston united states
8. Port of Felix&owe United Kingdom
9. Port of Fremantle Australia
10. Port of Hong Kong Hong Kong
11. Port of Jakarta Indonesia
12. Port of Kaohsiung Taiwan
13. Port of Klang Malaysia
14. Port of Le Havrc France
15. Port of Manila Philippines
16. Port of Melbourne Australia
17. Port of Montreal Canada
18. Port of Rotterdam Netherlands
19. Port of Singapore Singapore
20. Port of Sydney Australia
21. Port of Tacoma united states
22. Port of Wellington New Zealand
23. Port of Zeebruggc Belgium

qhere are generally four categories of cargoes that are handled in ports: dry bulk, liquid bulk, container-
iscdcargo, and noncontainerlscd nonbulk cargo. Each of these types require a certain type of ships and certain
Cargo-handllng quipment .
‘Data are available on request from the author.
‘It may also be postulated that the objective function of a port is to maximise throughput on a per unit
time basis subject to a revenue constraint in terms of a minimum amount of dividend required from government-
run ports. In the case of private ports. the constraint could be imposed by the stakeholders in the form of a
minimum profit level.
 Port performance and efficiency 241

significant item in port charges and thus, it is vital to keep this type of cost to a mini-
mum.5 The speed of moving cargoes into and out of ships at berth will influence the
overall port performance via the charges imposed on shipowners and the actual through-
put handled.

2.1. Determinants of throughput

Cargo size or throughput is determined by the following factors in the following
ways:
2.Z.I. Location. The geographical location of a port has an impact on cargo sire.
Tram-shipment ports such as Singapore, Hong Kong, Rotterdam, and Felixstowe are
different from the Ports of Melbourne and Sydney which are driven by an isolated and
small economy. The Ports of Singapore and Hong Kong are city states which have
entrepot trade as a basis for their economies. There is also a reasonable amount of such
“free trade” around Rotterdam, i.e., trans-shipment land-based cargo travel “tax free” till
they reach the country of final destination. The presence of a free trade zone around a
port also encourages its performance such the Port of Colombo, similar to Hong Kong
and Singapore.
2.1.2. Frequency of ship calls. The number of ship calls is important as it influences
the volume of cargo which can be moved through a port. Furthermore, increased fre-
quency of ship calls is attractive to importers and exporters. A study by Slack (1985)
provided interesting information on port choice. Slack surveyed port end users and
freight forwarders engaged in trans-Atlantic container trade between the American mid-
West and Europe to identify port selection criteria. Number of sailings was ranked first
among the list of port choice criteria. A study by Bird (1988) on the perceptions of
European freight forwarders again indicated that frequency of ship service is the main
reason for port choice. This study also confiied that time is of the essence in freight
forwarding and that frequency of shipping service is a major determinant of that time.
There are numerous published studies of the freight transport choice process. With
the notable exceptions of Bardi (1973), Brooks (1984; 1985), Saleh and Lalonde (1972)
and Merick and D’Este (1989), they have centered on modal choice. rather than address-
ing the more specific question of choice between competing ports. However, it seems
reasonable to assume that modal choice and the question of port choice are ultimately
related and that the results from modal choice studies are directly applicable.
Investigations of decision factors, e.g., Gilmour (1976), McGinnis (1979), Ogden
and Rattray (1982), Brooks (1985), Wilson et al. (1986), and Meyrick and D’Este (1989)
have identified three categories of factors: route factors (which include frequency, capac-
ity, and convenience directness, flexibility and transit time), cost factors (freight rate and
other costs), and service factors (delays, reliability and urgency, avoidance of damage,
loss and theft, fast response to problems, cooperation between shipper and carrier, docu-
mentation and tracing capability). They have found that shippers are generally risk-averse
and thus have limited options. Service factors, particularly service frequency, take prece-
dence over price. The predominance of service factors is not unusual and has been
previously reported by Cook (1967), Bayliss and Edwards ( 1970), Brooks (1984), Wilson
et al. (1986), and Meyrick and D’Este (1989).
2.1.3. Port charges. Port charges are undoubtedly an important factor, but shippers
are more concerned with indirect costs associated with delays, loss of markets/market
share, loss of customer confidence, and opportunities foregone due to inefficient service.
Port charges account for an extremely low proportion of overall costs of international
trading.
2.1.4. Economic activity. The demand for port services is a derived demand in that
it only occurs because of the interaction between individuals or sectors within an economy
or across countries for the exchange of goods that are produced and consumed at differ-
ent locations, respectively. The level of demand for port services is, therefore, dependent

‘This occurred when the port pricing structures in many countries have shifted from wharfage-based to
ship-based charges.
248 J.L.TONOZON

on the level of economic activity within a country and between countries. A change in this
level of activity will have a direct effect on the level of demand for port services.
2.1.5. Terminal efficiency. As defined earlier, terminal efficiency is measured in
terms of the number of containers loaded and unloaded per berth hour. This reflects the
labour and capital productivity levels in ports.

2.2. Determinants of efficiency

Terminal efficiency is in turn determined by the following factors:
2.2.1. Container mix. The composition of trade in relation to the proportions of
40-foot and 20-foot containers is an important determinant. Although a 40-foot container
is equivalent to 2 TEUs, it takes approximately the same time to handle as a 20-foot one.
2.2.2. Work practices: Delays in commencing and during stevedoring. Delays in
commencing and during stevedoring can cause inefficiency. The effect of these delays can
be measured in terms of differences between berth time and gross working time, and
between gross and net working time, respectively. Gross working time is calculated from
the time stevedoring labour goes on board to the time labour leaves the vessel on comple-
tion of cargo handling. Net working time takes account of delays that interrupt the
stevedoring operation. These delays could be due to meal breaks, equipment breakdown,
ship problems, weather, etc. The difference between berth time and gross working time
could be due to delays caused by industrial disputation and award conditions such as high
rates of penalty and shift allowances.
2.2.3. Crane efficiency. Crane efficiency is an indicator of how well working time is
being used. There are two areas of crane efficiency: crane hours per working hour and
effectiveness of crane operation. Crane hours depend on the number of cranes used to
load/unload a vessel as well as on the hours worked per day. In very high productivity
ports up to three cranes may be used whereas in some ports only one or two cranes are
used.
Effectiveness of crane operation refers to crane productivity measured in terms of
number of lifts per crane hour. Crane productivity may be related to the type and age of
cranes, related work practices, terminal layout, and management.
2.2.4. Vessel size and cargo exchange (economies of scale). Vessel size and cargo
exchange (containers loaded plus containers unloaded per ship) is also an important
determinant of terminal efficiency. A greater number of cranes can normally work a
larger vessel, and a larger cargo exchange allows better container selectivity in the vessel
hold and results in a lowering of berthing time as a percentage of total service time.
Formally, the functional relationship may take the following form:

TH = f (LOC, FS, EA, CH, E)

++ +-+ (1)

E = g (CONMIX, BRLWT, GWLN, CHWH, TEUCH, CE)

+ - + + + (2)

where:
TH = number of containers (TEUs) in a year;
LOC = location represented by a dummy variable;
FS = frequency of ship calls (all);
CH = average government and port charges;
EA = level of economic activity measured by respective countries’ GDP;
E = terminal efficiency (i.e., average number of containers per berth hour);
CONMIX = average container mix represented by the proportion of 40-foot containers;
BRLWT = average delays in commencing stevedoring represented by the difference
between the berth time and gross working time;
GWLN = average delays during stevedoring represented by the difference between
gross working and net working time;
CHWH = average crane hours per working hour;
 Port performance andefficiency 249
TEUCH = average crane productivity represented by the number of containers lifted
per crane hour;
CE = average vessel size and cargo exchange.

The algebraic signs underneath the equations indicate the expected direction of the
effect of the independent variables.
There are other qualitative factors that can influence terminal efficiency. The more
significant of these factors are as follows: balance or lack of balance between various
subsystems at a terminal; motivation and quality of container terminal personnel; stow-
age distribution pattern over the bays of the vessel; stowage position in the bays; lashing
systems utilised for on-deck containers; one-way or two-way handling and allocation of
handling equipment on terminal.

3. ESTIMATION
The structural equations for the determination of port performance and efficiency
have the following mathematical forms’?

TH = A LOC”’ FS& EAd CH” E”’ (3)

E = A CONMI??’ BRLWP’ GWLNB3CHWHB4TEUCHB5CEB6 (4)
where A refers to the constant term and the error terms are assumed to be normally
distributed with constant variances. Two econometric issues can be noted from the afore-
mentioned structural model. First, both equations indicate nonlinear relationships. Sec-
ond, one of independent variables, (E), in equation (3) is also a dependent variable in
equation (4).
To allow parameter estimation by multiple linear regression, both equations are
linearised by taking their respective logs. With respect to the second issue, the two-stage
feast squares (TSLS) estimation procedure seems appropriate for estimating equation
(3). Under the TSLS procedure, variable (E) is first regressed on its exogenous variables
and the estimated equation is used to generate a new independent variable, (E). The
values of E are then substituted for E in equation (3) before OLS estimation is applied.
Another way is by simply deriving the reduced form of the equations by substitution. To
highlight the role of the efficiency (E) variable in the determination of port performance,
the fust method is used (i.e., estimating the two structural equations separately). More-
over, because equation (3) has an error term independent from the error term in equation
(4), equation (3) can be estimated by OLS, without simultaneity bias.’ The OLS esti-
mates, based on a sample of 23 ports and under normality assumptions, for (3) and (4)
are presented in Table 2. * The first column ( 1) of Table 2 reports the estimated coeffici-
ents for equation (3). and the third column (3) refers to the estimation results for
equation (4).
The Goldfeld-Quandt Test points to an absence of heteroskedasticity in the afore-
mentioned estimates. The estimates are consistent with the postulates described in Section
2 and are highly significant particularly for equation (3) at the 1% level. Equation (4)
estimates are not as highly significant but their signs support the postulates. The Durbin-
Watson Test for equation (4) is inconclusive due to the omission of the two important
variables, GWLN and CHWH, as data on these variables are not available. 9
6AIIpossiblemathematicalforms were tried out. Butthisformis chosenbasedon economicandstatistical
criteria. To confirm that the functional form of the model is correctly specified, a testof functional
formdueto
Rv (1974). known as RESET,islaterused.
‘Thisis a case of recursive models in which the errors from the different equations are independent. Under
these conditions OLS estimation yields consistent estimates. See Maddala (1992, pp. 387-389).
“No data were available for variables, “GWLN” and “CHWH.” Because we are not dealing with time-series
data, it is not neassPry to test for stationarity and for the existence of a stable-state relationship between the
variables. The hypothesis of the exogeneity of the right-hand-side variables was tested using the Wu-Hausman
testing procedure.
9Although the data are cross-sectional, the Durbin-Watson statistic can be seen to provide some USefUl
(although not particularly powerful) evidence concerning model misspecif’ication (Bemdt, 1991. P. 93%
250 J.L. TONOZON
Table 2. Port performance and efficiency: Estimation results

In TH In TH’ In E

Variables (1) (2) (3)

Constant (::&** 6.20 -4.60

(2.91)*+’ ( - 1.99).
In PS 0.34 0.32
(2.34)** (2.70)**
In EA 0.37 0.31
(2.71)+* (2.41)+*
In E 0.75 0.80
(4.18)+** (5.25)***
LOC 0.36
(0.69)
In CH -0.11 -0.21
(-0.30) ( -0.72)
In CONMIX 0.43
(2.30)+*
In BRLWT -0.25
(-2.2)++
In TEUCH 1.25
(2.0).
In CE 0.37
(1.97).
Adj R2 0.83 0.86 0.40
D.W. 2.5 2.2 1.63
Ftest 18.0 35.4 10.00
RESET(l.17) 3.42 1.52 3.11
NORM (2) 1.61 1.25 0.48

‘Equation (3) without the variable (LOC). t ratios of coefficients

are given in brackets. Approximate critical values for the t ratios are:
10% = 1.74(.), 5% = 2.11(**), and 1% = 2.89(***). RESET =
Ramsey’s RESET test for functional form mis-specification. NORM
= Jarque-Bera test for the normality of the residuals. The degrees of
freedom for these tests are given in parentheses. In = natural loga-
rithm.

The relatively high coefficient for the terminal efficiency variable (E) indicates its
high importance in the determination of port performance. The variable location (LOC)
has the right sign but is statistically insignificant.
The effect of location, represented by a binary variable, on port throughput must
have been reflected in the frequency of shipcalls (FS) and thus, did not emerge statisti-
cally significant. lo A diagnostic test for collinearity, using the variance-decomposition
proportions and condition indexes, also confirms a high correlation between the FS and
LOC variables. Variable (LOC) was later dropped out of the original equation to deal
with the multicollinearity problem. The estimation results without the variable (LOC), as
reported in column (2) of Table 2, did not significantly differ from the original results.
The government and port charges (CH) are not statistically significant although it
has the correct sign. Port charges are a small proportion of total transport costs so that it
is not surprising that their overall impact on port choice decisions could have been offset
by other more significant indirect costs of transport. Note that port charges are also the
feedback effects of throughput changes via the economies of scope argument. Ports that
handle more traffic can lower their charges as a result of the following: overhead costs
associated with the provision of berths, dredging and breakwater, which must be incurred
irrespective of the volume of cargo involved, will be smaller the more throughput is
handled; given the stochastic nature of ship arrivals, the more berths a port has the lower
the percentage of berth idle time needed to help queueing to an acceptable level; some
“A better proxy for the variable, location, is volume of trans-shipped containers. However, not all of the
responding ports reported this type of information.
 Port performance and efficiency 25.1

elements of port infrastructure such as container terminals and loading facilities operate
at lower average costs as throughput increases, and greater concentration of cargoes
makes the use of larger vessels feasible which in turn lowers unit costs.
The aforementioned findings provide empirical support for the argument that port
performance is influenced by several factors, some of which are beyond the port authori-
ties’ control such as the level of economic activity, geographical location, and frequency
of shipcalls. There are, however, two variables which ports might be able to control
directly depending on the role ports play on the waterfront. One is terminal efficiency
(which has emerged as the single most significant contributor to port performance) if
port authorities take a more pro-active role.
On the other hand, if a port authority plays a “landlord” rather than a pro-active
role, then it only provides the basic port infrastructure and leaves the cargo-handling
operation and management to private stevedores. The other is port charges. The impact
of port charges on port performance is contingent on how significant these charges are in
the overall transport costs incurred by the port users. If the ports’ role on the waterfront
diminishes, its impact through its charges will certainly be reduced.

4. SUMMARY AND POLICY CONCLUSIONS

An important issue addressed in this study concerns the determinants of port perfor-
mance and efficiency. So far only few in-house studies have informally investigated this
issue. In this study, an attempt is made to integrate and empirically test the various
hypotheses of port performance and efficiency. The model has also quantified the contri-
bution of terminal efficiency to the overall determination of port performance. This is an
important finding as it provides empirical support for the notion that terminal efficiency
is a vital component of any waterfront reform aimed at improving port performance and
efficiency. The stronger influence of terminal efficiency relative to other factors in the
determination of port performance provides an empirical justification for giving top
priority to improving terminal performance in the overall process of waterfront reform.
The dominant contribution of crane productivity (TEUCH) to terminal efficiency justi-
fies the need to put more emphasis on enhancements of crane productivity. ‘I
The unavailability of data on the two variables related to delays during stevedoring
(GWLN) and crane utilization (CHWH) rates has constrained the efficiency analysis. It
will be interesting to see the extent to which these variables can contribute to terminal
efficiency once they are incorporated in the model. Finally, it is also interesting to see the
outcome if the economies of scope effects are neutralised by restricting the analysis to
ports of similar size. ”
Clearly, there is a lot to do. The evidence, provided in this article, that port perfor-
mance and efficiency can be modelled, suggests many directions that future research can
follow.

Acknowledgement-Department of Economics and Statistics, National University of Singapore. I was pre-

viously Chief Economist for the Port of Melbourne Authority, Melbourne, Australia. I am grateful to the
anonymous referees of this journal for their valuable comments on this article.

REFERENCES
Australian Bureau of Industry Economics (1993). International performance indicators in the waterfront.
Research Report 47.
Australian Business Council (1988). Efficiency in Australian Ports. Business Council Bulletin, 8-l 1.
Australian Transport Advisory Council ( 1992). Port performance indicators.
Bardi, E. J. (1973). Carrier selection from one mode. TransportationJournal, 13,23-29.
~ybs, B. T., & Edwards, S. L. (1970). Industrialdemandfor transport. Ministryof Transport, London.
Berndt, E. R. ( 1991). Thepractice of econometrics: Classicand contemporary. Reading. MA: Addison-Wesley.

“This particular result is consistent with my earlier study on the role of crane productivity in the determina-
tion of the overall level of efficiency at the Port of Melbourne, Australia. See Tongzon (1993 ).
“Ports can be classified into various clusters in terms of size and facilities. Only ports belonging to the
same clusters are included in the comparative analysis.
252 J. L. TONGZON

Bird, J. (1988). Freight forwarders speak: The perception of route competition via seaports in the European
communities research project. Maritime Policy and Management, 15( 1).
Brooks, M. R. (1984). An alternative theoretical approach to the evaluation of liner shipping (Part 1. Situa-
tional factors). Maritime Policy and Management, II, 35-43.
Brooks, M. R. (1985). An alternative theoretical approach to the evaluation of liner shipping (Part 2. Choice/
Criteria). Maritime Policy and Management, 12, 145-155.
ContainerisationInternational (various issues).
Cook, W. R. (1%7). Transport decisions of certain firms in the black country. Journalof TransrporfEconomics
and Policy, 1,325-344.
Gilmour, P. (1976). Some policy implications of subjective factors in the modal choice for freight movements.
The Logistics and Transportation Review, 12,39-57.
Jansson, J. O., & Shneerson, D. (1982). Port economics. Cambridge, MA: Massachusetts Institute of Tech-
nology.
Lloyd’s Ports of the World ( 1993). Lloyd’s of London Press.
Maddala, G. S. (1992). Introduction to econometrics (2nd ed.). New York: Ma&Iii.
Marconsult (1991). Mqior European container terminals structure andperfomtance. &nova.
McGinnis, M. A. (1979). Shipper attitudes towards freight transport choice. A factor analytic study. Intema-
tional Journal of Physical Distribution and Materials Management, IO, 25-34.
Meyrick, S., & D’Este, G. (1989, September). More than the bottom lime-How users select a shipping service.
Proceedings of the 14th Australasian TransFortResearch Forum, Perth, Australia (pp. 65-81).
Ogden, K. W., & Rattray, A. L. (1982). Analysis of freight mode choice. Proceedings of the 7th Australian
Transport Research Forum (pp. 249-276). Hobart, Australia.
Port of Melbourne Authority (1992). Comparativeport study (monograph).
Ramsey, J. B. (1974). Classical model selection through specification error tests. In P. Zarembka (ed.),
Frontiers in econometrics. New York: Academic Press.
Rees, R. (1984). Public entetprim economics (2nd ed.). London: Weidenfeld & Nicholson.
Robinson, R., & Reyes, J. M. (1988). ASEAN and Australian ports: Some aspects of efficiency and productivity
in the early 1980s. In K. Trace et al. (eds.), Handmaiden of trade (pp. 113-144). Singapore: Singapore
University Press. 113-144.
Saleh, F.. & LaLonde, B. J. (1972). Industrial buying behaviour and the motor carrier selection decision.
Journal of Purchasing, 8,18-33.
Slack, B. (1985). Containerisation, inter-port competition and port selection. Maritime Policy and Manage-
ment, 12(4).
Tongzon, J. (1989). The impact of wharfage costs on Victoria’s export-oriented industries. Economic PaPer&
8.58-64.
Tongzon, J. (1993). Port of Melbourne Authority’s new pricing policy: Its efficiency and distributional imptica-
tions. Maritime Policy and Management. 2q 3), 197-205.
United Nations Conference on Trade and Development (1987). Meusuring and eva/uating Fort performance
andproductlvity. UNCTAD Monographs on Port Management, 6, New York: United Nations.
Wilson, F. R., Bisson, 8. J., & Kobia, K. B. (1986). Factors that determine mode choice in the transportation
of general freight. 7?ansporration Research Record, 1061, X-3 1.

Appendix 1. Descriptive statistics (based on the 23 sampled ports)

Coefficient of
Variables Maximum Minimum Mean SD Variation

E (TEUs per berth hour) 50.0 2.0 14.4 13.6 0.95

CONMIX (%) 67.0 4.0 28.0 17.7 0.63
BRLWT (hours) 31.0 0.0 4.8 6.7 1.38
TEUCH (TEUs per Crane hour) 44.0 13.1 24.8 8.6 0.35
EA (USSM) 449414 17976 149600 119026 0.79
FS (number of ship calls) 129303 560 10706 26783 2.50
CH (As) 151.1 30.68 90.40 33.25 0.37
TH (number of TEUs) 6350000 43450 1360947 1852359 1.36
CE (average number of TEUs
per shipcall) 667 25 232 186 0.80

Statistics are based on the data provided by the sampled ports and other secondary sources
discovered in Section 1.1 of this article.