Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

This article was downloaded by: [Oregon State University]

On: 08 January 2015, At: 23:30


Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Applied Economics
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/raec20

Port choice in a competitive environment: from the


shipping lines' perspective
a a
Jose L. Tongzon & Lavina Sawant
a
Department of Economics , National University of Singapore , Kent Ridge Crescent,
Singapore 119260
Published online: 30 Oct 2009.

To cite this article: Jose L. Tongzon & Lavina Sawant (2007) Port choice in a competitive environment: from the shipping
lines' perspective, Applied Economics, 39:4, 477-492, DOI: 10.1080/00036840500438871

To link to this article: http://dx.doi.org/10.1080/00036840500438871

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Applied Economics, 2007, 39, 477–492

Port choice in a competitive


environment: from the shipping
lines’ perspective
Jose L. Tongzon* and Lavina Sawant
Department of Economics, National University of Singapore, Kent Ridge
Crescent, Singapore 119260
Downloaded by [Oregon State University] at 23:30 08 January 2015

The present era of globalization and regional competition has witnessed


the increasing influence of the global operations of the shipping lines on
port development and the declining monopoly of ports. This warrants the
need for an identification of the port selection criteria from the perspective
of the shipping lines. Hence, an attempt is made to determine the port
choice from the perspective of the shipping lines. Moreover, most port
operators have designed their strategies based on the ‘stated preference’ of
the shipping lines. However, the stated preference approach can have
several shortcomings such as the shipping lines would tend to overstate
their demands for services at the port. This directs for an examination of
the port choice of the shipping lines based on a revealed preference
approach. The ‘revealed preference’ approach aims at judging the port
choice of shipping lines purely from their ‘actions’ rather than what they
actually ‘state’. The empirical study in this article is based on a survey
conducted among major shipping lines operating in Singapore and
Malaysia. The findings have shown port charges and wide range of port
services to be the only significant factors in their port choice. Moreover,
the results show no consistency between the stated and revealed
preferences of shipping lines.

I. Introduction decreasing monopoly of the ports and the major


influence of the global operations of the shipping
Ports form an important link in the overall trading lines on port development. Also, with the mergers
chain and consequently, their level of performance and acquisitions among the shipping lines, the greater
and efficiency determines to a large extent a nation’s volumes that are now controlled by a single line or
international competitiveness. However, with globa- alliance implies that the capacity of a shipping line to
lization and the growing regional competition,1 port seriously affect the business of a port is much greater
operators are seen to be having less and less control than it has been in the past. This warrants a need for
over their destinies. This is evident from the the port operators to understand the underlying

*Corresponding author. E-mail: josetongzon@yahoo.com


1
Some examples of severe competition between ports are the case of the Port of Singapore and the Malaysian ports such as
Port of Tanjung Pelepas (PTP) and Port Klang, the case of the South China ports and Hong Kong and also the case of
Antwerp and Rotterdam.
Applied Economics ISSN 0003–6846 print/ISSN 1466–4283 online ß 2007 Taylor & Francis 477
http://www.tandf.co.uk/journals
DOI: 10.1080/00036840500438871
478 J. L. Tongzon and L. Sawant
factors of port competitiveness from the perspective Baird (1996) has stated that faster turn around time
of the shipping lines. within the port is an important factor in port
Port operators have difficulty quantifying the advancement. Tabernacle (1995) has illustrated the
weightage given by the shipping lines to the importance of crane productivity in enhancing port
importance of factors at the port such as efficiency, performance with the help of the learning curve
strategic location, adequate infrastructure, port concept. Fung (2001) has tried to measure the
charges, connectivity, cargo size and provision of competition between the Port of Singapore (PSA)
wide range of port services. Hence, they have and the Hong Kong Port by looking at factors like the
established a valuation for the importance of these volume of trade passing through these two terminals,
factors through a ‘stated preference approach’. A especially to countries like China and the rest of Asia.
stated preference approach refers to a method of With the help of this, he has confirmed the trade
valuation of the importance of the various factors interdependency between the two countries and its
involved in determining port choice based on asking impact on their port performance. Further, he has also
the shipping lines to state their preferences through proven that, when the markets of the two ports
some type of survey or questionnaire.2 To date, most overlap, their market shares will become a function of
port operators have designed policies based on the the prices they charge and how well they meet the
stated preference of the shipping lines. However, the needs of the shippers and shipping lines. These
Downloaded by [Oregon State University] at 23:30 08 January 2015

stated preference approach tends to generate publications are, however, focusing on partial pro-
responses stating a much higher level of importance ductivity factors which might not prove to be helpful
for the factors at the port than actually being to the port authorities/operators for policy making in
observed by the shipping lines3 i.e. the shipping the long run. Also most shipping lines are concerned
lines tend to overstate the demands for the factors at with the overall productivity of the port rather than
the port by a large amount (Schaller, 1999). Hence partial productivity.
this creates a need to examine the determinants of Tongzon (1994) has come close to determining the
port choice through a ‘revealed preference approach’. overall port productivity and performance by looking
A revealed preference approach refers to a method of at such factors as location, frequency of ship calls,
valuation of the importance of the various factors economic activity within the sector, labour and capital
involved in determining port choice based on observ- productivity and work practices within a port. Besides
ing the behaviour of the shipping lines and the there is an extensive literature on freight transport
choices made by them.4 Perhaps the strategy for the choice of shippers centring on modal choice and
port operators to succeed in the competitive environ- carrier selection (see e.g. Jerman et al., 1978; Marr,
ment would be to see if there is a consistency in the 1980; McGinnis, 1980; Gray, 1982; Meyrick and
stated preference and revealed preference of the D’este, 1989). These articles have listed the ‘route
shipping lines and draw significant policy implica- factors’ such as frequency, capacity, convenience and
tions from the revealed preference. transit time; ‘cost factors’ such as freight rates and
In this light, the objective of this article is to assess ‘service factors’ such as delays, reliability, avoidance
the various factors of port choice from the shipping of damage, loss and theft as some of the important
lines’ perspective and to see if there is a consistency factors influencing the decision of the shippers in their
between the stated preference and the revealed choice of a carrier. However, no such study has been
preference of the shipping lines for the factors conducted investigating the port choice from the
influencing their port choice. shipping lines’ perspective. Moreover, a study based
Existing literature has attempted to analyse the on carrier choice by Brooks (1985) has found that
fierce competition for market share among ports shippers’ actual behaviour may not accord with their
located within the same region and serving approxi- stated preferences.5 But, no such attempt has so far
mately the same hinterland, using a wide range of been made to examine if there is a consistency between
approaches and key determinants in measuring the stated preference and the revealed preference of
port performance and efficiency. According to the shipping lines.
Fleming and Hayuth (1994), geographical location The rest of the article is structured as follows:
is vital to explaining a port’s competitive success. Section II explains briefly the methodology adopted;
2
http://www.schallerconsult.com/pub/enhancef.htm
3
This difference is called hypothetical bias.
4
http://www.schallerconsult.com/pub/enhancef.htm
5
Brooks (1985) found that when asked to rank the importance of the various factors, shippers rated cost more highly. But an
analysis of their actual decisions revealed that service factors were more important to them.
Port choice in a competitive environment: from the shipping lines’ perspective 479
Sections III and IV present the results followed by The second part of the study identifies the revealed
conclusion and recommendations. preferences of the shipping lines based on the
behaviour of the shipping lines and actual choices
made by them. This refers to the key factors observed
to be influencing the port choice of the shipping lines
II. Methodology based on their actions and actual choices made by
them. The revealed preferences of the shipping lines
The analysis is based on the results of the survey are derived based on the questions posed to them in the
conducted by Tongzon (2002) among selected ship-
survey, asking them (1) to state their hub port of call
ping lines comprising a sample size of 31. The study
and (2) rate the chosen factors at their hub port of call.
focuses on South East Asia. But, since the major
‘Hub port of call’ is referred to as the most preferred
concentration of port competition within this region
and most frequently used port by the shipping lines.
is between Singapore and Malaysia, the focus is
This would indirectly reveal the key factors which have
further narrowed down to determine the port choice
influenced the port choice of the shipping lines when
of the shipping lines based in Singapore and
they selected their own hub ports of call. This is
Malaysia. Hence the sample size comprises of the
examined with the application of Binary Logistic
shipping lines in Singapore and Penang, Malaysia.
Regression (BLR) and Model selection.
Downloaded by [Oregon State University] at 23:30 08 January 2015

There are two main parts to the study. The first


part examines the stated preferences of the shipping
lines for the factors identified to be important by the
shipping lines in their decision to choose a port. The
stated preferences of the shipping lines are derived III. Empirical Evidence: Stated Preference
based on the questions posed to them in the survey
about how important they thought were the chosen Based on observation and the overview of the
factors in their decisions to choose a particular port. Boxplot given subsequently, we can see that ‘effi-
‘Chosen factors’ refers to the factors considered in the ciency’ is stated to be of prime importance. We can
study as determinants of port choice, i.e. efficiency, see from the Descriptive statistics (Table 1), that the
location, adequacy of infrastructure, port charges, median of efficiency is 6.00, which indicates that the
connectivity, cargo size and wide range of port median concentration for efficiency is more densely
services. They were asked to rate the factors with concentrated in the upper quartile region and is
7 being the highest rating and 1 being the lowest higher relative to the other factors. Also the whiskers
rating. These ratings are evaluated with the help of on the boxplot (which directs the skewness) indicate
the Boxplot analysis to determine the most important that the ratings given to efficiency are skewed to the
factors stated by the shipping lines. left, i.e. they are skewed toward the low ratings which

Table 1. Descriptive statistics for the ratings of stated preference of shipping lines
Variable Factors N Mean Median SD
Ratings Cargo size 31 3.000 2.000 2.280
Connectivity 31 4.581 5.000 1.766
Efficiency 31 5.710 6.000 1.532
Infrastructure 31 4.194 4.000 1.922
Location 31 4.194 4.000 2.072
Port charges 31 4.387 5.000 1.764
Port services 31 3.097 3.000 2.039
Variable Factors SE mean Minimum Maximum Q1 Q3
Ratings Cargo size 0.410 1.000 7.000 1.000 5.000
Connectivity 0.317 1.000 7.000 3.000 6.000
Efficiency 0.275 2.000 7.000 5.000 7.000
Infrastructure 0.345 1.000 7.000 2.000 6.000
Location 0.372 0.000 7.000 3.000 6.000
Port charges 0.317 1.000 7.000 3.000 6.000
Port services 0.366 0.000 7.000 1.000 5.000
Note: This table is the MINITAB output for the Boxplots shown in figure 1. This indicates that their ratings are less variable
and more densely packed around their means relative to the others (except efficiency). Hence we can conclude that port
charges and connectivity are the second most important factors to the shipping lines in their choice of a port.
480 J. L. Tongzon and L. Sawant
indicates that the ratings are sparse and more spread Boxplots of ratings by factors
out between the median and the lower quartile. In
addition, the inter-quartile range6 for efficiency is
7
very small: 7  5 ¼ 2. This shows that the ratings are
6
condensed around the median and there is not much
variability in the general consensus about the 5

Ratings
importance of efficiency. This can also be verified 4
from the SD of efficiency which is 1.532 and it is the 3
smallest compared to that of the other factors. 2
Looking at the Boxplots of port charges and
1
connectivity, they both look similar since both have
the same median values, same upper and lower 0
quartile values, same interquartile ranges and same Factors

Cargo size

Connectivity

Efficiency

Infrastructure

Location

Port charges

Wide range
of port
services
SD values. Also they both are skewed to the left and
their whiskers are of the same length as they both
have the same minimum and maximum values.
Hence, we can conclude that they both are identical
Downloaded by [Oregon State University] at 23:30 08 January 2015

Fig. 1. Boxplot of ‘Stated preference’ of the shipping lines.


and equally important to the shipping lines in making 7 ^ highest rating; 1 ^ lowest rating
their decisions about choosing a port. In comparison
to the remaining factors (except efficiency), we see infrastructure is the highest and its SD is the lowest
that port charges and connectivity have a higher relative to the other two factors which indicates it has
median concentration, lower SD and a lower inter- a higher median concentration and lower variability
quartile range, except that their interquartile range is in its ratings. Also infrastructure is symmetric at the
on par with that of location. median value, while cargo size and wide range of port
We now consider the remaining four factors: services are right skewed which implies that the
infrastructure, location, cargo size and wide range of ratings given to the latter two factors are more
port services. Among these, we can clearly see that the towards the lower side. Hence we can safely conclude
median concentration is higher for infrastructure and that infrastructure is the fourth most important
location relative to cargo size and wide range of port factor to the shipping lines in their choice of a port.
services. Hence, we first compare the boxplots for Finally, comparing the boxplots of cargo size and
infrastructure and location. It can be seen from Fig. 1 wide range of port services, we can see that wide
and Table 1 that they have the same median value. But range of port services has a higher median concen-
a smaller interquartile range of location (6  3 ¼ 3) tration and lower SD relative to cargo size, indicating
than that of infrastructure (6  2 ¼ 4) indicates the smaller variability in the ratings of the former. Their
ratings for location are more concentrated around the interquartile range is the same and both are right
mean than the ratings for infrastructure. The SD of skewed. But cargo size is more skewed to the right
location is larger than that of infrastructure. However, than wide range of port services which implies that
we must take note of the fact that the SD of location the ratings given to cargo size are more inclined
has been influenced by the extreme values.7 The range8 towards the lower side than those for wide range of
for location is 7  0 ¼ 7 while the range for infra- port services. Hence, we can conclude that wide range
structure is 7  1 ¼ 6. Based on the formula for SD, i.e. of port services is the fifth most important factor and
pP
[(x  x)/n  1], we can see that the rating of 0 cargo size is the sixth important factor as stated by
has led the SD of ‘location’ to be slightly higher. the shipping lines in their decision to choose a
Hence, we can safely conclude that location would be particular port.
the third most important factor for the shipping lines
in their choice process.
Considering the remaining three boxplots of
infrastructure, cargo size and wide range of port IV. Empirical Evidence: Revealed Preference
services, we see that all three have the same
interquartile range, i.e. all three reflect the same The survey sample includes shipping lines based in
spread in the data. However, the median value for Singapore and Malaysia and their ‘hub port of call’
6
Interquartile range ¼ Upper quartile  Lower quartile.
7
Location has the lower extreme value of 0 which is the rating given by one of the shipping lines in the sample.
8
Range ¼ Maximum value  Minimum value.
Port choice in a competitive environment: from the shipping lines’ perspective 481
has been either PSA or a Malaysian port. Hence, the Y^ ¼ 1:442 þ 0:3683 Adequate Infrastructure
dependent variable would be binary in nature,  0:4859 Port charges
i.e. whether shipping lines choose PSA or do not
choose PSA as their most preferred port and a test þ 0:4156 Wide range of port services
using BLR is conducted. The response variable at 15% level of significance:11
is defined as:
 The result of the final model shows that port charges,
1, if shipping line chooses PSA wide range of port services and adequate infrastruc-
Yi ¼
0, if shipping line does not choose PSA, ture are revealed to be the key factors influencing the
port choice of the shipping lines. Here we see a lack of
where i represents the shipping line. consistency with the results of the ‘stated preference’
The Multiple Logistic Regression is fitted using the of the shipping lines in Section III. This inconsistency
following independent variables: between the stated and the revealed preferences
X1 ¼ Adequate infrastructure, X2 ¼ Location, would be justified in the next section.
X3 ¼ Efficiency, X4 ¼ Port charges,
X5 ¼ Cargo size, X6 ¼ Connectivity and Justification for revealed preference results
Downloaded by [Oregon State University] at 23:30 08 January 2015

X7 ¼ Wide range of port services To identify the underlying reasons for the incon-
sistency between the ‘stated preference’ and the
‘revealed preference’ would require the justification
Results of the binary logistic regression to be divided into two parts. In Section IV, the
We can see from the Logistic regression table, the justification for the factors included in the final model
fitted logit is given by: for ‘revealed preference’ of the shipping lines would
Y^ i ¼ 0:128 þ 0:4532 Adequate infrastructure be outlined. Followed by this, in Section IV an
elaborate justification would be given with the help of
 0:0892Location  0:1535Efficiency
a statistical and perceptual approach for the factors
 0:6762Port charges þ 0:1335Cargo size which have been deleted out of the final model for the
 0:3344Connectivity ‘revealed preference’.
þ 0:5551Wide range of port services:9
Justification for the factors included in the ‘revealed
Analysis of binary logistic regression preference’ of the shipping lines. First, as discussed
previously, it should be highlighted that the world is
To see whether the factors collectively contribute in
moving towards globalization, mergers, alliances and
explaining the logit,10 we test the hypothesis that the
severe inter-port competition. Hence, the shipping
coefficients of the independent variables (1, 2, 3,
lines are also diversifying their operations to cater to
4, 5, 6, 7) are all zero. Based on the results, we
the door-to-door philosophy from mere transport
conclude that at the 5% level of significance the
providers to integrated logistics organizations.
variables in the study are significant and thus
Moreover, the perceived improvements in port
collectively contribute to the model.
performance and the varied choice of ports of call
We further conduct Model selection and the final
have resulted in the lines concentrating on the
model chosen is:
maritime route that best serve their interest of faster
Y^ ¼ 0:031  0:4812 Port charges and cheaper transit rather than a particular
hinterland.(Slack, 1993). At such a juncture with the
þ 0:4326 Wide range of port services at 5%
fear of overlapping hinterlands and two ports in the
level ofsignificance and same region being at the threshold level of efficiency

9
In BLR, instead of predicting Yi we fit the model to predict the logits, log [/(1  )] where  represents the probability that
the shipping line chooses PSA i.e. Pr [Yi ¼ 1/Xi] ¼  and (1  ) represents the probability that the shipping line does not
choose PSA i.e. Pr [Yi ¼ 0/Xi] ¼ 1  . Thus our fitted regression equation is of the form, Y^ i ¼ =ð1  Þ ¼
e0 þ1 X1 þ2 X2 þ3 X3 þ4 X4 þ5 X5 þ6 X6 þ7 X7 . Taking the natural logarithm of both sides we get, log ½=ð1  Þ ¼ 0 þ 1 X1 þ
2 X2 þ 3 X3 þ 4 X4 þ 5 X5 þ 6 X6 þ 7 X7 .
10
Logit refers to the port choice of the shipping line. It is the response variable of the BLR equation in this study.
11
Menard (2001) has stated that the usual 5% criterion for the significance level is too low and could result in the exclusion of
some important variable from the model. Hence, he has tested and proven that we could use a criterion of 15–20% level of
significance.
482 J. L. Tongzon and L. Sawant
Table 2. Binary logistic regression for the ratings of revealed preference of the shipping lines
Link function: logit
Variable Value Count

Response information
Port selection 1 14 (Event)
0 17
Total 31
Odds 95% CI

Predictor Coefficient SE coeff. Z P Ratio Lower Upper

Logistic regression table


Constant 0.128 2.148 0.06 0.953
Infrastructure 0.4532 0.3060 1.48 0.139 1.57 0.86 2.87
Location 0.0892 0.2645 0.34 0.736 0.91 0.54 1.54
Efficiency 0.1535 0.2697 0.57 0.569 0.86 0.51 1.46
Port charges 0.6762 0.3346 2.02 0.043 0.51 0.26 0.98
Cargo size 0.1335 0.2519 0.53 0.596 1.14 0.70 1.87
Downloaded by [Oregon State University] at 23:30 08 January 2015

Connectivity 0.3344 0.3078 1.09 0.277 0.72 0.39 1.31


R. port services 0.5551 0.3037 1.83 0.068 1.74 0.96 3.16
Log-likelihood ¼ 12.965
Test that all slopes are zero: G ¼ 16.754, DF ¼ 7, p-value ¼ 0.019
Note: This results were obtained for the Full Model (FM).

(e.g. PSA vs. PTP), an integrated package of lower time, the efforts of PTP to rival PSA cannot be
charges, new value added services and adequate ignored with the construction of a bunker port and a
infrastructure at the port would be vital for a port petrochemical complex near PTP12 (The Shipping
to gain comparative advantage over the other. Times, 26 July 2002). Also, the terrorist attacks in
Hence, most ports are now exploiting the oppor- New York and Washington have led to immediate
tunities to provide efficient warehousing and dis- actions to improve the global supply chain security at
tribution activities such as inventory control, data ports like Singapore which handle more than 20% of
management, packing and processing (Slack, 1993). the world’s total transhipment business.
In addition, several ports are also facilitating the Besides, since the port charges need to be
concept of ‘one-stop-shopping’ by providing services continuously reduced, the only way to maintain a
like bunkering, repairs and shipyard, customs clear- higher margin of profits is by complementing the low
ance, inland transport, security, banking facilities and charges with more value added services. Also the fact
forwarding agents. With the provision of these that the world’s busiest transhipment port and the
services, the ports can benefit not only from the sale second biggest container terminal operator, PSA, is
of these services but also from the additional cargo seen to be conducting a major overhaul of its strategy
that flows through the port. in its bid to stay cost competitive clearly indicates
A telephone interview with a Malaysian port agent that today port charges might be the one of the
revealed that, although the Malaysian ports and PSA foremost factor influencing port choice.
are at the threshold level of efficiency, PSA still Likewise, the physical infrastructure at the port
attracts a substantial amount of transhipment traffic such as cranes, berths, yards, tugs and storage system
due to its cutting edge in the provision of wide range have been the traditional forms of physical infra-
of port services. For example, while PSA remains the structure available at almost all ports. However, the
top bunkering port in the world, recent developments present era seems to be diversifying its definition of
at the port with the introduction of an Accreditation infrastructure by including services such as (1) inland
Scheme for Bunker Suppliers, intensifying bunker transportation system to serve the hinterland, city
quality checks and adopting the ‘Improved Bunker surrounding the port and even the neighbouring
Sampling Method’ have further enhanced its produc- countries (2) A ‘free port’ status which can be defined
tivity (MPA press release, 30 June 2003). At the same as a zone in the port where equipment, materials and
12
This project has aimed to capture a slice of Singapore’s bunkering business worth over S$5 billion a year. (The Shipping
Times, 26 July 2002).
Port choice in a competitive environment: from the shipping lines’ perspective 483
Table 3. Descriptive statistics for the ratings given to factors at PSA and Malaysian ports

Factors Efficiency Location Cargo size Connectivity

Port PSA Malaysia ports PSA Malaysia ports PSA Malaysia ports PSA Malaysia ports
Median 4.0 5.0 5.5 5.0 4.5 4.0 3.0 4.0
SD 2.0 2.317 2.065 2.149 2.326 2.339 2.381 1.654
Q1 2.0 2 2.75 2.0 1 1 1.75 2.5
Q3 5.25 6.5 6.0 6.0 6 6 6.25 5.5

Table 4. Tabulated percentage for ratings given to factors at PSA and Malaysian ports
Efficiency Location Cargo size Connectivity
0 1 0 1 0 1 0 1
1–3 7 5 1–3 6 5 1–3 8 6 1–3 5 8
41.18% 35.71% 35.29% 35.71% 47.06% 42.86% 29.41% 57.14%
4–7 10 9 4–7 11 9 4–7 9 8 4–7 12 6
58.82% 64.29% 64.29% 64.71% 52.94% 57.14% 70.59% 42.86%
Downloaded by [Oregon State University] at 23:30 08 January 2015

All 17 14 All 17 14 All 17 14 All 17 14


100% 100% 100% 100% 100% 100% 100% 100%

goods are introduced free of duties and taxes, the descriptive statistics (Table 3) calculated for the
goods being then stored (in a commercial free port) or ratings given by the shipping lines to the services
processed (in an industrial free port) and re-exported. provided at these ports. The results of the descriptive
This is evident from the recently announced $461.6m statistics would be further verified with the help of
growth plan of Port Klang to adopt the Dubai model tabulated percentage (Table 4) for the same ratings.
and create a ‘free zone’ distribution centre with Here the ratings are divided into two categories, the
commercial and logistic services to boost its status as first category consisting of ratings ranging from 1 to 3
a transhipment hub (The New paper, 19 February which would signify ratings on the lower side and
2003) and 3) the ‘electronic data interchange’ (EDI) the second category consisting of ratings ranging
system which speeds up communications between the from 4 to 7 which would signify the ratings on the
ports and the shipping lines producing a higher higher side.
service level and gains in efficiency through lower
operational cost. Some shipping experts have stated Efficiency
that the denial of PSA to meet the demands of From Table 3 we can see that the median concentra-
Maersk Sealand for dedicated berths and lower tion for the Malaysian ports is higher than that for
charges led to the move of Maersk Sealand to PTP. PSA. However the interquartile range14 and the SD
for PSA are smaller than that for the Malaysian
Justification for the factors deleted from the ‘revealed ports. This indicates that while the median concen-
preference’ of shipping lines. One evident reason for tration for Malaysian ports is higher, the ratings for
the deletion of the other factors could be that, given PSA are more concentrated around the mean and
the increasing advancements in port facilities, today have a lower variance. Based on this, we can conclude
most ports have achieved the threshold level of that the efficiency levels of PSA and the Malaysian
performance required by the shipping lines. Hence, ports are almost compatible. Also the tabulated
the deleted factors would not be identified as percentage (Table 4) of the ratings for efficiency
prominent factors by the shipping lines. To test this, shows PSA received 64.29% and the Malaysian ports
a comparison would be made between the factors at 58.82%. The 6% of higher ratings for the Malaysian
PSA and the Malaysian ports13 with the help of the ports has been influenced by the high ratings given to
13
Since the ‘hub port of call’ of the shipping lines in the sample has been either PSA or one of the Malaysian ports such as
PTP or Port Klang, a comparison is made between these ports. As this study focuses on determining the factors influencing
the port choice of the shipping lines rather than doing a comparison between two ports, the ratings for PTP and Port Klang
have been included under one identity of ‘Malaysian ports’.
14
Interquartile range ¼ upper quartile  lower quartile, i.e. Q3 – Q1 in Table 3.
484 J. L. Tongzon and L. Sawant
the efficiency at PTP. But we must note that the range, i.e. the ratings for location of PSA are less
workload and the operational cost at PTP are much variable and more concentrated around the mean
less compared to that at PSA. This is evident relative to the ratings for location of Malaysian ports.
from the fact that PTP handled around 2.8 million Based on this, we can say that the location for PSA is
20-ft-equivalent units (TEUs) in 2002 compared to slightly better than that for the Malaysian ports.
PSA’s 16.8 million TEUs local throughput (The However, the tabulated percentage (Table 4) shows
Straits Times, 19 February 2003). Also given the that the ratings given to the location of PSA
small clientele base at PTP, it has not been put constitutes 64.71% while that for Malaysian ports is
through a rigorous test as yet. Thus, based on the 64.29%. The ratings seem on par because the values
aforementioned statistical evidence, we can say the of the median, SD and interquartile ranges between
efficiency level for the ports of the two countries is them differ very slightly. Hence, we can say that the
almost on par. location of the Malaysian ports is almost on par with
Moreover, there have been growing concerns about the location of PSA.
the need for ports to improve their efficiency to Several researchers have stressed upon the inherent
prevent being pushed out of the increasingly compe- potential for a port with a sheltered and deep-water
titive market situation. As a result, several ports have harbour and extensive areas of level land to have an
invested heavily to upgrade their facilities by employ- added advantage in attracting significant amount of
Downloaded by [Oregon State University] at 23:30 08 January 2015

ing superior quality cranes, building more berths, traffic. Although this claim cannot be rejected, it is
creating more storage capacity and providing better important to consider that lately several ports are
wage incentives. As Notteboom et al. (2000) have artificially building these facilities. While Port Klang
stated, ‘the industry is dominated by the belief that and PSA are deep-water ports, PTP has invested
the best workable strategy to defeat competitors is heavily in building a deep-water port and deeper
building new and highly efficient terminals’. Such drafts to accommodate the large vessels. Besides, the
great amount of pressure has driven the relatively greater part of the waterway of the Straits of
new and smaller ports to re-think and re-define their Malacca17 runs through the territorial waters of
strategies in order to attain compatibility in the Indonesia and Malaysia with the much shorter
operational environment within which the commer- Singapore straits joining it at the southern end.
cial ports operate. This is evident from the most Also the ‘Traffic Separation Scheme’ (TSS), adopted
recent developments of PTP15 and the South China by the International Maritime Organization (IMO) to
ports which are relatively newcomers in the port avoid casualties at the complex pattern of the Straits
industry and yet have proven to be a major threat to comprising of heavy traffic, stretches from Port
the established providers of port services, i.e. PSA Klang through PTP and PSA. Hence, all of these
and the Hong Kong port, respectively. In addition, three major ports bordering the Malacca straits
the relocation of the operations of two major provide an adequate stopover for the cargo opera-
shipping lines, Maersk Sealand and Evergreen from tions of containerships. Based on the aforementioned
PSA to PTP has been a remarkable evidence of the arguments, we can confirm the compatibility of the
growing competency of PTP which came into location of these ports and hence it thus justifies
existence only in 1999. The expertise acquired by the deletion of location from the final model of the
Port Klang has also been alarming.16 From this, we ‘revealed preference’ of the shipping lines.
can infer that the major ports in South East Asian,
Connectivity
i.e. PSA and Malaysian ports like Port Klang and
The deletion of connectivity from the final model of
PTP, are almost on par in terms of the minimum level
the ‘revealed preference’ shows that although in the
of efficiency required by the shipping lines.
‘stated preference’ of the shipping lines (Section III) it
Location was rated at exactly the same level as port charges, the
Table 3 reflects that PSA has a higher median ‘revealed preference’ indicates that port charges is
concentration, smaller SD and a smaller interquartile given a higher weightage compared to connectivity.

15
PTP has continuously invested in building larger berths and employing superior quality cranes.
16
Port Klang has two major attractions to the shipping lines: cargo from the industrial hinterland of the Klang Valley and low
rates. It has also been expanding over the last year with the conversion of conventional berths to container handling facilities.
(Lloyd’s List, 28 February 2003)
17
The Malacca strait is approximately 500 miles long and is the longest straits in the world used for international navigation.
It forms the main seaway connecting the Indian Ocean with the China Sea and provides the shortest route for vessels trading
between the Middle and East Asian countries.
Port choice in a competitive environment: from the shipping lines’ perspective 485
As mentioned earlier, connectivity is not merely a Table 3 show that the cargo size at PSA and at the
function of the maritime routes. With globalization it Malaysian ports have the same interquartile range
is becoming more so a function of the port’s and SD. The median concentration for cargo size at
connection through all modes of faster and cheaper PSA is slightly higher than that of the Malaysian
transport to states within the country as well as ports. Also Table 4 reveals the cargo size for PSA at
neighbouring countries. Based on an interview with 57.14% and that of the Malaysian ports 52.14%.
Captain Sawant,18 we confirm that the main reason With this, we can deduce that the handling of
for the deletion of connectivity is that it has not been of cargo size is almost compatible in the Malaysian
primary concern to the shipping lines in their ports and PSA.
operations (Interview, 15 January 2003). This is Although the general consensus has been that
because the operations of the shipping lines have larger vessels generate economies of scale; in the
prominently consisted of delivery of goods from one present era with the economic downturn and
port to the other port. It is mainly the responsibility of volatility in the level of trade and freight rates, it
the freight forwarders to pick up the goods from the has been difficult for the larger vessels to generate
port and deliver them to the point of consumption, i.e. sufficient cargo to fill the large ships and to deal with
to the various consignees. Based on this, we can say the increasing operational cost.19 Most ocean carriers
that good connectivity would be of greater concern to including Cosco Container Lines had envisaged
Downloaded by [Oregon State University] at 23:30 08 January 2015

the freight forwarders than the shipping lines. the new container vessel deliveries to reduce by the
From Tables 3 and 4, we can infer that connectivity second half of 2002. Also Hans Payer, member of
at the Malaysian ports is better than that at PSA. The the executive board of Germanisher Lloyd,20 said, ‘It
ratings for the Malaysian ports have a higher median is doubtful that the economies of scale in large ships
concentration and smaller SD and interquartile range can be carried on indefinitely’ (Containerisation
than those for PSA. The results are such because International, June 2002)
Malaysia is more densely connected by roads and Thus, the above analysis justifies the redundancy of
railway within the country as well as to the the factors deleted from the ‘revealed preference’ of
neighbouring countries. This has benefited the freight the shipping lines. This once again confirms the
forwarders at the Malaysian ports. For example, PTP validity of the ‘revealed preference’ of the shipping
is in an advantageous position to offer a fast road lines in the present era of regional competition and
network, a growing selection of short sea feeder globalization. With this, we have also successfully
operations and a rail connection to central Malaysia investigated the inconsistency derived between the
and southern Thailand. The PTP officials have stated stated and the revealed preference of the shipping
that, ‘the port has been able to stem the flow of lines in the previous section. The analyses in the two
southern Malaysian produced goods across the sub sections thus confirm the inconsistency between
causeway due to its price competitiveness and the ‘stated preference’ and the ‘revealed preference’ of
increased connectivity’. (Containerisation the shipping lines for factors influencing their port
International, June 2002). Port Klang also enjoys choice. The analyses also clearly reflect the under-
good connectivity through its dense network of road lying reasons for the inconsistency between the two
and railways. The importance of connectivity to preferences.
freight forwarders is also seen from the comments of
the manager of Kenwood Logistics which set up its
South East Asian distribution hub in PTP. He says,
‘PTP offered us the ability to consolidate, value add V. Conclusion
and redistribute all of our cargo within this region’.
Hence it’s the freight forwarders that attach greater This study has provided a valuable insight in
importance to connectivity rather than shipping lines. evaluating the determinants of port choice from the
Cargo size perspective of the shipping lines in the present era of
Cargo size has been defined as the capacity of the regional competition. The findings of this study
port to handle the amount of cargo being carried by a have clearly shown that the stated preference of
vessel during a particular voyage. The results in the shipping lines markedly differs from their
18
Captain Sawant is a Master Mariner by profession who has been in the shipping industry for over 25 years. The interview
was held to discuss the importance of connectivity to the shipping lines in their operations. The interview confirmed the
changing definition of connectivity.
19
Furthermore, carrying empty containers would create an imbalance in box movement and hence the lines settle for lower
freight rates to recover a portion of their cost.
20
It is the largest classification society for containerships.
486 J. L. Tongzon and L. Sawant
revealed preference. Hence, the results have shown no hinterlands and two ports within the same region
consistency between the two. being at the threshold level of efficiency; lower
The study involved two parts. The examination of charges, provision of new value added services and
the stated preferences has shown that efficiency was adequate infrastructure in an integrated package
of prime importance followed by port charges and would be vital for a port to gain a comparative
connectivity being on par. Location was the third advantage over the other.
most important, infrastructure fourth, wide range of The findings have provided theoretical and empiri-
port services was the fifth most important factor and cal support to justify that the revealed preference of
cargo size the sixth most important factor in their the shipping lines should be incorporated in policy
choice of a port. The test for the revealed preference formulations. Further, the influence of port charges
has identified port charges and wide range of port relative to other factors in influencing port choice as
services to be the significant factors for shipping lines inferred in the revealed preference is a clear indication
at 5% level of significance and infrastructure at 15% that, although several studies have shown that quality
level of significance. The remaining factors came out always takes precedence over price, there is only a
to be insignificant. maximum price that the shipping lines would be
There are several reasons for the revealed prefer- willing to pay. Hence, only as long as the price is
ences of shipping lines to differ from their stated below this maximum limit then the quality of port
Downloaded by [Oregon State University] at 23:30 08 January 2015

preferences. Most importantly, with increasing com- services would take precedence over price. While we
petition shipping lines are restructuring their opera- consider the relevance of port charges we must,
tions to more on the provision of new value added however, take note that there is still a minimum level
services to the shippers and also on outsourcing for of service and efficiency on which the shipping lines
logistic operations. This has led them to concentrate will not compromise. Hence in the present era it is the
more on the maritime route that best serves their total package that counts.
interest of faster and cheaper transit, rather than a The above position can be translated to the present
particular hinterland. Commenting on the uncer- context in the announcement made by PSA to slash
tainty of transhipment hubs, Maersk Sealand Asia’s 14% of its workforce in its bid to cut operation cost
regional chief executive, Mr. Bjarne Hansen, cau- and be cost competitive with its neighbouring ports,
tioned, ‘One should always bear in mind that if Port Klang and PTP (The Straits Times, 19 February
shipping lines find a better, more efficient and less 2003). Thus, in spite of achieving a higher load factor
expensive place along their route, they’ll shift. by automating the back-end of operations and
Nothing is given; you have to remain competitive at offering high level of efficiency, PSA seems to be
all times’ (The Straits Times, 22 February 2003). The facing tough competition due to the rapid develop-
brunt of this is felt by the ports which are seen to ments and future plans of the neighbouring ports to
undertake huge investments to meet the increasing increase capacity and maintain handling charges
needs of the shipping lines. As a result, most ports are about 50% lower than that at PSA.21 Moreover, the
seen to have achieved the threshold level of efficiency. dominant contribution of the wide range of port
Further, the value of a strategic location is disappear- services in the study justifies the need for the ports to
ing as most ports today have invested heavily in provide for them to keep a strong hold on their
artificially building a deep-water port and deeper footloose customers. This involves facilitating the
drafts to accommodate larger vessels. Cargo size is concept of ‘one-stop-shopping’ by providing services
another factor that has lost its significance as the like bunkering, repairs, shipyard, customs clearance,
current economic downturn has made it difficult for inland transport, security, banking facilities, inven-
the shipping lines to generate sufficient cargo to fill tory and warehouse management and distribution
the larger vessels. This has raised fears for the and supply chain management.22 Finally, provision
shipping lines to face diseconomies of scale. Finally, of adequate infrastructure such as efficient inland
connectivity has also been insignificant because this is transportation, EDI and dedicated berths would add
not really of significance to the shipping lines, but value to the services at the port in addition to the
rather more of significance to the freight forwarders. available physical infrastructure which is easily
Thus, at such a juncture with the fear of overlapping available at any port.
21
PTP has capacity to handle four million TEUs, with plans to raise that to six million TEUs by 2006 (The Straits Times,
19 February 2003). Port Klang has announced its $461.6m growth plans to adopt the Dubai model and create a ‘free zone’
distribution centre supported by commercial and logistic services to boost its status as a transshipment hub.
22
Although Singapore remains the top bunkering port, the Malaysian plan to build a bunker port and a petrochemical
complex at PTP poses a threat to Singapore in the near future.
Port choice in a competitive environment: from the shipping lines’ perspective 487
Thus in addition to technical and labour efficiency, Fung, K.-F. (2001) Competition between the ports of Hong
special consideration should be given by the port Kong and Singapore: a structural vector error correc-
tion model to forecast the demand for container
operator to adopt ‘differentiation’ in their approach handling services, Maritime Policy and Management,
to offer services. ‘Differentiation’ refers to providing 28, 3–22.
specific port services in market niches distinct from Gray, R. (1982) Behavioural approaches to freight trans-
those provided by other ports, offering greater value port modal choice, Transport Reviews, 2, 161–84.
Hosmer, D. and Lemeshow, S. (2000) Applied Logistic
to the port user (Notteboom et al., 2001, p. 83). In Regression, John Wiley & Sons, New York; Singapore.
retrospect, this article has highlighted the prevailing Jerman, R., Anderson, R. and Constantin, James (1978)
difference between the stated preference and revealed Shipper Vs Carrier perceptions of carrier selection
preference of the shipping lines. These findings are variables, International Journal of Physical
important because efficient allocation of resources in Distribution, 9, 29–38.
Lloyd’s List, 28 February (2003).
the present era of regional competition creates a need Marr, N. (1980) Do managers really know what service
to decipher between the stated preference and their customers require?, International journal of
revealed preference of the shipping lines for imple- physical distribution, 10, 433–44.
menting effective and competitive port policies. Maritime P. Authority press release, 30 January 2003,
Singapore.
Although the econometric estimation was based on McGinnis, M. (1980) Shipper attitude towards Freight
a small sample, the results have provided some useful
Downloaded by [Oregon State University] at 23:30 08 January 2015

Transportation choice: a factor analytic study,


insights into port choice determinants from the International Journal of Physical Distribution, 10, 25–34.
perspective of the shipping lines. Menard, S. (2001) Applied Logistic Regression Analysis,
To date several researchers have attempted to Sage Publications, Thousand Oaks, CA.
Meyrick, S. and D’este, G. (1989) More than the bottom
develop statistical models to determine the relevant line – how users select a shipping service. Proceedings
factors affecting the port industry. However, it would of the 14th Australian Transport Research Forum,
be beneficial if some amount of research was devoted Perth, Australia, 65–81.
to understanding the behaviour of the shipping lines Notteboom, T. and Winkelmans, W. (2001) Structural
changes in logistics: how will port authorities face the
and their preferences in the present era of regional challenge?, Maritime Policy and Management, 28,
competition. It would be interesting to take this study 71–89.
further to investigate for a consistency between the Notteboom, T., Coeck, C. and van den Broeck, J. (2000)
stated preference and the revealed preference on an Measuring and explaining the relative efficiency of
international scale. A comparison could be made to container terminals by means of Bayesian Stochastic
Frontier models, International journal of Maritime
see if the findings derived for the South East Asian Economics, 2, 83–106.
region are similar to those found for another region. Schaller, B. (1999) Enhancing Transit’s Competitiveness: A
Studies could also be conducted to verify if the Survey Methodology, http://www.schallerconsult.com/
dominance of port charges would continue to be pub/enhancef.htm
Slack, B. (1993) Pawns in the Game: ports in a global
persistent in the near future. transportation system, Growth and change, 24, 579–88.
Tabernacle, J. (1995) A study of the changes in perfor-
mance of quayside container cranes, Maritime Policy
and Management, 22, 115–24.
Tongzon, J. (1994) Determinants of port performance and
References efficiency, Transportation Research-A, 29A, 245–52.
Tongzon, J. (2002) A survey of shipping lines at selected
Baird, A. (1996) Containerisation and the decline of the ports of Southeast Asia (unpublished), Department of
upstream urban port in Europe, Maritime policy and Economics, National University of Singapore.
Management, 23, 145–56. The Shipping Times, ‘NOL hopes to finish talks on PSA
Brooks, M. R. (1985) An alternative theoretical approach terminal deal soon’, 23 July 2002.
to the evaluation of liner shipping (Part 2, Choice The Shipping Times, ‘Pelepas will handle 10m containers by
Criteria), Maritime Policy and Management, 12, 2005: Dr M’, 26 July 2002.
145–55. The Shipping Times, ‘Connectivity counts’, 3 August 2002.
Chatterjee, S., Hadi, A. and Price, B. (2000) Regression The New Paper, ‘ Port Klang’s $461.6m growth plan’, 19
Analysis by Example, Wiley, New York. February 2003.
Containerisation International (News), June (2002). The Straits Times, ‘Job cuts at PSA, Don’t blame us:
Fleming, D. and Hayuth, Y. (1994) Spacial characteristics Tanjung Pelepas port’, 19 February 2003.
of transportation hubs: centrality and intermediacy, The Straits Times, ‘PSA needs to slim down so it won’t go
Journal of Transport Geography, 2, 3–18. under water’, 22 February 2003.
488 J. L. Tongzon and L. Sawant
Appendix
arbitrary and has no particular meaning. 4) Lastly,
A1-Overview of the Box Plot we can detect unusually large or small data values
called outliers. These will be displayed in the
With the help of the Box Plot we get a clear display of MINITAB output with an ‘’ sign.
the ‘stated preference’ of the shipping lines. The
Boxplot examines the following areas: (1) the location
A2-Validity of the factors in the study
of the median of the ratings. The median which is the
middle value in the ordered list indicates the 50th To see whether the factors collectively contribute in
percentile of the data, i.e. 50% of the data lies below explaining the logit,26 we test the hypothesis that the
the median and the other 50% lies above the median. coefficients of the independent variables (1, 2, 3,
Based on the location of the median in the range from 4, 5, 6, 7) are all zero. Hence, we would test the
7 to 1, we can observe the median concentration of Full Model (FM) against the Reduced model (RM).
the ratings. For example, a higher median value
^ ¼ b0 þ b1 Adequate infrastructure
FM: Y
would have a higher median concentration indicating
that 50% of the ratings given to the factor are in þ b2 Location þ b3 Efficiency
the upper range.23 (2) The direction of skewness of þ b4 Port charges þ b5 Cargo size
the data which is indicated by the length of the
þ b6 Connectivity
Downloaded by [Oregon State University] at 23:30 08 January 2015

‘whisker’.24 If the data are left skewed it would imply


that very few shipping lines have given lower ratings þ b7 Wide range of port services
for the factor and vice versa if the data are right vs:
skewed. (3) The spread of the ratings judged by the ^ ¼ b0
RM: Y
‘interquartile range’. Interquartile range ¼ Upper
quartile25  Lower quartile. A small value of the The hypothesis is tested based on the ‘G’ statistic
interquartile range indicates a small amount of which has a chi-square distribution with 7 degrees
variability in the ratings for the factor which implies of freedom. From Table 2, G ¼ 16.754. A small value
the ratings are more densely packed around the of chi-square implies that the variables are not
mean. Since the interquartile range represents the significant while a large value of chi-square indicates
width of the middle half of the data it is robust the significance of the variables in the model
against outliers. The height of the box is usually (Chatterjee et al., 2000). From the chi-square

Table A1. Logistic regression table for fitting the model without wide range of port services

Odds 95% CI

Predictor Coef. SE coef. Z P Ratio Lower Upper


Logistic regression Table
Constant 0.918 2.127 0.43 0.666
Adequate 0.5077 0.2773 1.83 0.067 1.66 0.96 2.86
Location 0.1757 0.2421 0.73 0.468 0.84 0.52 1.35
Efficien 0.1297 0.2310 0.56 0.574 0.88 0.56 1.38
Port cha 0.6757 0.3044 2.22 0.026 0.51 0.28 0.92
Cargo si 0.2359 0.2563 0.92 0.357 1.27 0.77 2.09
Connection 0.0638 0.2333 0.27 0.785 0.94 0.59 1.48
Log-likelihood ¼ 14.991
Test that all slopes are zero: G ¼ 12.703, DF ¼ 6, p-value ¼ 0.048

23
The sample median is a better estimation than the sample mean as the median is not sensitive to outliers.
24
‘Whisker’ refers to the tails projecting from the left and right of the boxplot. In Fig. 1, it is to the top and bottom of the
Boxplots.
25
‘Quartiles’ are data values that divide an ordered data set into four quarters: the lower quartile, lower middle quartile, upper
middle quartile and the upper quartile. Upper quartile region represents 75th percentile of the data and lower quartile
represents 25th percentile of the data.
26
Logit refers to the port choice of the shipping line which is the response variable of the BLR equation in this study.
Port choice in a competitive environment: from the shipping lines’ perspective 489
distribution table, the 5% critical value of the chi- for the FM and the RM. G ¼ 2  [12.965 
square distribution with 7 degrees of freedom is (14.991)] ¼ 4.052. From the chi square distribution
14.07. As 16.754 > 14.07 we can conclude at the 5% table, the 5% critical value of the chi-square
level of significance that the variables in the study distribution with 1 degree of freedom27 is 3.84.
collectively contribute to the model. Also the p-value Based on the criteria for ‘G’ statistic28 we can
of the test: P [2 (7) > 16.754] ¼ 0.019 is considerable conclude that since 4.052 > 3.84, the variable wide
smaller than 0.05. This indicates that the variables are range of port services is significant.
significant. Thus we reject the RM that all coefficients
Case 2: Test to determine the significance of
are zero.
‘connectivity’
A3-Model Selection
^ ¼ b0 þ b1 Adequate infrastructure
RM: Y
Test at 5% level of significance: þ b2 Location þ b3 Efficiency
Case 1: Test to determine the significance of ‘wide þ b4 Port charges þ b5 Cargo size
range of port services’
þ b7 Wide range of port services
^ ¼ b0 þ b1 Adequate infrastructure
RM: Y vs:
Downloaded by [Oregon State University] at 23:30 08 January 2015

þ b2 Location þ b3 Efficiency ^ ¼ b0 þ b1 Adequate infrastructure


FM: Y
þ b4 Port charges þ b5 Cargo size
þ b2 Location þ b3 Efficiency
þ b6 Connectivity
þ b4 Port charges þ b5 Cargo size
vs:
þ b6 Connectivity
^ ¼ b0 þ b1 Adequate infrastructure
FM: Y
þ b7 Wide range of port services
þ b2 Location þ b3 Efficiency
To check the significance of connectivity we test the
þ b4 Port charges þ b5 Cargo size
RM against the FM and find the ‘G’ statistic as
þ b6 Connectivity mentioned in the earlier case. G ¼ 2  [12.965 
þ b7 Wide range of port services (13.621)] ¼ 1.312. The 5% critical value of the chi-
square distribution with 1 degree of freedom is 3.84.
To check the significance of ‘wide range of port Since 1.312 < 3.84, we conclude that the connectivity
services’ we find the ‘G’ statistic which is obtained is not significant in this model and hence can be
from twice the difference between the Log-Likelihood deleted from the model.

Table A2. Logistic regression table for fitting the model without connectivity
Odds 95% CI

Predictor Coef. SE coef. Z P Ratio Lower Upper


Logistic regression Table
Constant 0.500 2.120 0.24 0.814
Adequate 0.5130 0.2876 1.78 0.074 1.67 0.95 2.93
Location 0.1661 0.2488 0.67 0.504 0.85 0.52 1.38
Efficien 0.2164 0.2561 0.84 0.398 0.81 0.49 1.33
Port cha 0.6285 0.3068 2.05 0.040 0.53 0.29 0.97
Cargo si 0.1455 0.2408 0.60 0.546 1.16 0.72 1.85
Wide ran 0.4016 0.2515 1.60 0.110 1.49 0.91 2.45
Log-likelihood ¼ 13.621
Test that all slopes are zero: G ¼ 15.443, DF ¼ 6, p-value ¼ 0.017

27
12.965 is the Log-likelihood for the FM with DF ^ 7. For verification, see Table 2. The DF for the RM ¼ 6 (Table A1).
Hence DF to be used to find the chi-square distribution at the 5% level of significance is, DF ¼ DFFM  DFRM, i.e. DF of
FM  DF of RM. Hence DF to be used in the test is 7  6 ¼ 1.
28
A small value of chi-square would imply that the variables do not contribute significantly to the model while a large value of
chi-square would prove the significance of the variables in the model.
490 J. L. Tongzon and L. Sawant
Table A3. Logistic regression table for fitting the model without cargo size

Odds 95% CI

Predictor Coef. SE coef. Z P Ratio Lower Upper


Constant 0.105 1.978 0.05 0.958
Adequate 0.4866 0.2841 1.71 0.087 1.63 0.93 2.84
Location 0.1521 0.2442 0.62 0.533 0.86 0.53 1.39
Efficien 0.2430 0.2484 0.98 0.328 0.78 0.48 1.28
Port cha 0.5588 0.2651 2.11 0.035 0.57 0.34 0.96
Wide ran 0.4251 0.2466 1.72 0.085 1.53 0.94 2.48
Log-likelihood ¼ 13.812
Test that all slopes are zero: G ¼ 15.060, DF ¼ 5, p-value ¼ 0.010

Table A4. Logistic regression table for fitting the model without port charges

Odds 95% CI
Downloaded by [Oregon State University] at 23:30 08 January 2015

Predictor Coef. SE coef. Z P Ratio Lower Upper


Constant 3.026 1.866 1.62 0.105
Adequate 0.4304 0.2493 1.73 0.084 1.54 0.94 2.51
Location 0.0682 0.2306 0.30 0.767 0.93 0.59 1.47
Efficien 0.1020 0.2080 0.49 0.624 0.90 0.60 1.36
Wide ran 0.4521 0.2257 2.00 0.045 1.57 1.01 2.45
Log-likelihood ¼ 16.555
Test that all slopes are zero: G ¼ 9.575, DF ¼ 4, p-value ¼ 0.048

Case 3: Test to determine the significance of ‘cargo vs:


size’ ^ ¼ b0 þ b1 Adequate infrastructure
FM: Y
^ ¼ b0 þ b1 Adequate infrastructure
RM: Y þ b2 Location þ b3 Efficiency
þ b2 Location þ b3 Efficiency þ b4 Port charges
þ b4 Port charges þ b7 Wide range of port services
þ b7 Wide range of port services
We test the RM against the FM The ‘G’ statistic is
vs: G ¼ 2  [13.812  (16.555)] ¼ 5.486. The 5% criti-
^ ¼ b0 þ b1 Adequate infrastructure
FM: Y cal value of the chi-square distribution with 1 degree
of freedom is 3.84. Since 5.486 > 3.84 we conclude
þ b2 Location þ b3 Efficiency
that port charges is a significant variable influencing
þ b4 Port charges þ b5 Cargo size the port choice of the shipping lines and hence it
þ b7 Wide range of port services should be retained in the model.
We again test the RM against the FM. The ‘G’ Case 5: Test to determine the significance of
statistic is G ¼ 2  [13.621  (13.812)] ¼ 0.382. The ‘efficiency’
5% critical value of the chi-square distribution with 1 ^ ¼ b0 þ b1 Adequate infrastructure
RM: Y
degree of freedom is 3.84. Since 0.382 < 3.84 we
conclude that cargo size is not significant in this þ b2 Location þ b4 Port charges
model and hence can be deleted from the model. þ b7 Wide range of port services
Case 4: Test to determine the significance of ‘port vs:
charges’ ^ ¼ b0 þ b1 Adequate infrastructure
FM: Y
^ ¼ b0 þ b1 Adequate infrastructure
RM: Y þ b2 Location þ b3 Efficiency
þ b2 Location þ b3 Efficiency þ b4 Port charges
þ b7 Wide range of port services þ b7 Wide range of port services
Port choice in a competitive environment: from the shipping lines’ perspective 491
Table A5. Logistic regression table for fitting the model without efficiency logistic regression table

Odds 95% CI

Predictor Coef. SE coef. Z P Ratio Lower Upper


Constant 0.986 1.858 0.53 0.595
Adequate 0.4109 0.2524 1.63 0.104 1.51 0.92 2.47
Location 0.1284 0.2371 0.54 0.588 0.88 0.55 1.40
Port cha 0.5025 0.2541 1.98 0.048 0.60 0.37 1.00
Wide ran 0.4083 0.2385 1.71 0.087 1.50 0.94 2.40
Log-likelihood ¼ 14.336
Test that all slopes are zero: G ¼ 14.012, DF ¼ 4, p-value ¼ 0.007

Table A6. Logistic regression table for fitting the model without location

Odds 95% CI
Downloaded by [Oregon State University] at 23:30 08 January 2015

Predictor Coef. SE coef. Z P Ratio Lower Upper


Constant 1.442 1.710 0.84 0.399
Adequate 0.3683 0.2367 1.56 0.120 1.45 0.91 2.30
Port cha 0.4859 0.2511 1.93 0.053 0.62 0.38 1.01
Wide ran 0.4156 0.2368 1.75 0.079 1.52 0.95 2.41
Log-likelihood ¼ 14.485
Test that all slopes are zero: G ¼ 13.715, DF ¼ 3, p-value ¼ 0.003

Table A7. Logistic regression table for fitting the model without adequate infrastructure
Odds 95% CI

Predictor Coef. SE coef. Z P Ratio Lower Upper


Constant 0.031 1.332 0.02 0.982
Port cha 0.4812 0.2323 2.07 0.038 0.62 0.39 0.97
Wide ran 0.4326 0.2342 1.85 0.065 1.54 0.97 2.44
Log-likelihood ¼ 15.820
Test that all slopes are zero: G ¼ 11.044, DF ¼ 2, p-value ¼ 0.004

We test the RM against the FM. The ‘G’ statistic is, ^ ¼ b0 þ b1 Adequate infrastructure
FM: Y
G ¼ 2  [13.812  (14.336)] ¼ 1.048. The 5% criti-
cal value of the chi-square distribution with 1 degree þ b2 Location þ b4 Port charges
of freedom is 3.84. Since 1.048 < 3.84, we conclude
that efficiency is not significant and hence can be þ b7 Wide range of port services
deleted from the model. We test the RM against the FM. The ‘G’ statistic is
Case 6: Test to determine the significance of G ¼ 2  [14.336  (14.485)] ¼ 0.298. The 5% criti-
‘location’ cal value of the chi-square distribution with 1 degree
of freedom is 3.84. Since 0.298 < 3.84, we conclude
^ ¼ b0 þ b1 Adequate infrastructure
RM: Y that location is not significant and hence can be
þ b4 Port charges deleted from the model.
þ Cargo size þ b6 Connectivity Case 7: Test to determine the significance of
vs: ‘adequate Infrastructure’
492 J. L. Tongzon and L. Sawant
^ ¼ b0 þ b4 Port charges þ b5 Cargo size
RM: Y To check the significance of adequate infrastructure,
þ b6 Connectivity we test the RM against the FM. The ‘G’ statistic is
G ¼ 2  [14.485  (15.820)] ¼ 2.67. The 5% critical
vs:
value of the chi-square distribution with 1 degree of
^ ¼ b0 þ b1 Adequate infrastructure
FM: Y freedom is 3.84. Since 2.67 < 3.84, we can say that
þ b4 Port charges adequate infrastructure is insignificant at the 5% level
þ b7 Wide range of port services of significance.
Downloaded by [Oregon State University] at 23:30 08 January 2015

You might also like