Sie sind auf Seite 1von 11

Transport Policy 18 (2011) 665–675

Contents lists available at ScienceDirect

Transport Policy
journal homepage: www.elsevier.com/locate/tranpol

Vessel and voyage determinants of tanker freight rates and contract times
Amir H. Alizadeh a, Wayne K. Talley b,n
a
Cass Business School, City University, London EC1Y 8TZ, United Kingdom
b
College of Business and Public Administration, Old Dominion University, Norfolk, VA 23529, United States

a r t i c l e i n f o abstract

Available online 5 February 2011 The behavior of shipping freight (charter) rates and the timing of shipping contracts affect the
Keywords: transportation costs of charterers and the operating cash flows of shipowners. Although the literature
Freight rate has established macroeconomic determinants of shipping freight rates, there has been no systematic
Tanker investigation of microeconomic determinants of freight rates and the delivery time of chartered ships
Shipping routes (the laycan period) in the tanker market. Therefore, the aim of this paper is to investigate the
Laycan importance of vessel and contract specific factors in the determination of tanker freight rates and
laycan periods in shipping contracts. Individual tanker shipping contracts from January 2006 to March
2009 are used to estimate freight rates and laycan periods using a system of simultaneous equations.
The estimation results suggest that the duration of the laycan period is an important determinant of the
shipping freight rate and vice versa. Other determinants of freight rates include the vessel’s hull type,
fixture deadweight utilization ratio, vessel age, and voyage routes. Determinants of the laycan period
include the former determinants as well as the Baltic Dirty Tanker Index and its volatility.
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction (e.g., the size and age of the vessel and the geographical route
in which the vessel is employed) are also important in determin-
In 2008 the cargo carrying capacity of the world tanker fleet of ing shipping freight rates. In addition, the terms and conditions of
402 million tonnes was 38% of the total world shipping fleet and the charter contract such as loading date in relation to contract
the number of tanker ships exceeded 4800.1 During the same year date and cargo size in relation to vessel capacity are important
the tanker fleet transported 2795 million metric tonnes (mmt) of determinants of freight rates. Yet, to our knowledge, there has
liquid bulk commodities around the world, out of which been no systematic attempt in the literature to investigate
2043 mmt was crude oil and 752 mmt was petroleum products. microeconomic determinants of tanker freight rates and the
The tanker shipping market is by far the largest sector of the delivery time of tanker charter contracts. This paper aims to
world shipping industry in terms of trading volume and weight. investigate the effect of microeconomic variables including vessel
Therefore, it is not surprising that a large number of studies have and voyage specific factors in determination of tanker freight
been devoted to analyzing tanker freight (charter) rates, charter- rates. In addition, determinants of the timing of the freight
ing decisions and policies, transportation strategies, and fleet contract in relation to the actual loading date, a period known
deployments and operations of the tanker shipping industry. as the laycan period, as well as the interrelationship between the
Freight rates in the tanker shipping industry fluctuate con- freight rate and laycan period are investigated.
siderably in short run (Glen and Martin, 1998; Kavussanos, 2003). There are two important dates in any vessel charter contract.
Such fluctuations affect the formation of shipping policies, trans- The first date is the day on which negotiations between the
actions, and contracts as well as cash flows and costs of ship- shipowner and the charterer are completed, i.e., the day on which
owners and charterers (see Brown et al., 1987; Laulajainen, 2008). the contract is agreed and the charter-party is signed. This date is
While these fluctuations are argued to be due to the global known as the fixture date or the hire date. The second date is the
economic activity as well as the state of the tanker shipping date on which the ship must present herself at the loading
market, variables related to vessel and route characteristics port ready for loading the cargo, i.e., the layday. In practice, once
the loading or lifting day for the cargo has been determined, the
trader (charterer) will enter the market to find and charter
n
Corresponding author. the most suitable tanker for transportation of the cargo from
E-mail addresses: a.alizadeh@city.ac.uk (A.H. Alizadeh),
wktalley@odu.edu (W.K. Talley).
the loading port to the destination port. Depending on the nature
1
Statistics are from Clarkson Research Services Limited for tankers greater of the trade, the route, and the type of cargo, the loading date may
than 10,000 deadweight tonne capacity. be anytime from a few days to a couple of months ahead of the

0967-070X/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tranpol.2011.01.001
666 A.H. Alizadeh, W.K. Talley / Transport Policy 18 (2011) 665–675

time when the trading contract and transportation decisions are and Randers and Göluke (2007) also use macroeconomic variables
made. Hence, assuming vessels are available in the market at all in a system dynamic setting to model and forecast tanker
time and at a constant flow, the trader has the option to enter into freight rates.
the freight market and hire a vessel anytime until the very last Other studies have examined the time series properties of
minute, as long as it is practical, before loading of the cargo. shipping freight rates such as their dependence on past values;
Therefore, it is the trader’s – or charterer’s – decision for all further, they use univariate or multivariate time series models to
practical purposes as to when to enter the market and charter (or capture the dynamics of freight rates and their volatilities. These
hire) a ship. For instance, if the conditions are not favorable and models are then used to forecast shipping freight rates and
there is enough time prior to the cargo loading date, the trader volatilities (Veenstra and Franses, 1997; Kavussanos and
may wait and not inform the shipbroker about the cargo and the Alizadeh, 2002; Kavussanos, 2003; Adland and Cullinane, 2005,
need for a ship. The charterer’s decision of when to charter a ship 2006; Lyridis et al., 2004; Batchelor et al., 2007). These studies
is dependent on such market conditions as current and expected utilize aggregate data and macroeconomic variables in an attempt
freight rates, the volatility of freight rates, and the cost to be to capture the dynamics and fluctuations in shipping freight rates
incurred of not being able to find a ship to charter if the decision and accordingly utilize the models for forecasting purposes;
to hire is delayed. however, their performance in predicting shipping freight rates
The purpose of this paper is fourfold: (1) investigate vessel and has not been promising. The underperformance of structural and
voyage determinants (e.g., vessel age, hull type, route, and dead- time series models using macroeconomic variables for forecasting
weight utilization of the fixture) of individual tanker freight rates; shipping freight rates may be attributed to the use of aggregate
(2) investigate vessel and voyage determinants of the delivery and macroeconomic data. While aggregate macroforecasts of
time of chartered tanker ships (the laycan period); (3) examine freight rates could be useful for medium to long term investment
differences in tanker freight rates and laycan periods across purposes, shipowners and charterers require microforecasts for
different routes; and (4) investigate the relationship between making operational decisions, cash flow analyses, and budgeting
individual tanker freight rates and laycan periods of tanker (e.g., microforecasts of shipping freight rates for specific routes).
charter contracts. From the point of view of charterers the timing The literature on microeconomic determinants of shipping
of entering into the market for fixing a vessel may affect the freight rates is limited to the studies by Tamvakis (1995)
demand in the market and consequently move freight rates. and Tamvakis and Thanopoulou (2000). In particular, Tamvakis
However, waiting for more favorable freight rates may be risky (1995) hypothesized that the construction and employment of
as freight rates tend to move very sharply in a very short period of double-hull tankers as mandated by the US Oil Pollution Act 1990
time. From the shipowners’ point of view, information on the role (OPA) would affect the freight rates for double-hull versus single-
of vessel and voyage specific factors in determination of freight hull tankers. That is to say, the increasing pressure on oil
rates can be used in investment, operations, and deployments importers and charterers to hire double-hull vessels would result
strategies. in a relative increase in the demand for double- versus single-hull
The remainder of the paper is as follows. Section 2 presents a tankers, consequently resulting in higher rates for double-hull
brief review of previous studies on tanker shipping and freight tankers. However, the empirical results were mixed; no signifi-
determination. Section 3 presents the methodology (i.e., econo- cant premia was found in the hiring of double- versus single-hull
metric equation specifications) for investigating vessel and voy- tankers. In the study by Tamvakis and Thanopoulou (2000), the
age determinants of individual freight rates and laycan periods for existence of a two-tier dry-bulk ship charter market, reflecting
tanker shipping contracts. The data are described in Section 4, the age of vessel, was investigated. However, the empirical results
while Section 5 presents the empirical results. Finally, conclusions based on the time period 1989–1996 (that covered different
are presented in the last section. stages of the shipping cycle) found no significant difference
between freight rates paid for newer versus older vessels.
In a theoretical study, Strandenes (1999) argues that in order
2. Review of literature for a two-tier liquid-bulk ship charter market to exist, the
demand for quality tankers has to increase by more than 30%
The tanker freight market is characterized by the interaction of from its levels at the end of 1990s. She also points out that stricter
supply and demand for tanker shipping services (Tinbergen, enforcement of OPA compliance in Europe of substandard tankers
1934; Koopmans, 1939; Zannetos, 1966; Beenstock and may trigger a two-tier tanker market in which quality tankers
Vergottis, 1993). The demand for tanker services is a derived trade at a premium. However, such market conditions may last
demand, in the sense that it is derived from the international for only 3–5 years as substandard tankers will be forced out of the
trade in oil and oil products, which in turn depends on world market. Mokia and Dinwoodie (2002) investigate the spatial
economic activity and imports and consumption of energy com- organization, segmentation, contractual terms, and pricing
modities (see Stopford, 2009). The supply of tanker shipping aspects of tanker operations and lay-times.2 They observe sig-
services, on the other hand, depends, for example, on the size of nificant differences in lay-times across tanker shipping routes and
the tanker fleet, the tonnage that is available for trading, tanker ship sizes and suggest that the differences in lay-times should
shipbuilding activities, bunker fuel prices, the scrapping rate of appear in tanker charter contracts. However, they do not inves-
the fleet, and the productivity of the tanker fleet at any point tigate differences in laycan periods and charter rates of tanker
in time. Studies by Hawdon (1978), Strandenes (1984), and contracts across routes.
Beenstock and Vergottis (1989, 1993), among others, empirically Although numerous studies have investigated the formation
investigate the determination of shipping freight rates, the price and behavior of shipping freight rates, specific vessel and voyage
for sea transportation, through the interaction between supply determinants of these rates in the tanker freight market have not
and demand for sea transportation. They find that factors such as been investigated in the literature. This paper undertakes such an
world economic activity, growth in industrial production, sea-
borne trade in commodities, oil prices, availability of vessel
tonnage, new vessel buildings on order, and tanker shipbuilding 2
In ship chartering terminology, ‘‘laytime’’, usually quoted in hours, is defined
deliveries and the scrapping rates determine freight rates for sea as the time allowed for a ship to load and discharge its cargo and is usually stated
transportation. More recent studies by Dikos et al. (2006) in ship charter contracts.
A.H. Alizadeh, W.K. Talley / Transport Policy 18 (2011) 665–675 667

investigation for shipping freight rates as well as for the timing of However, it might be the case that freight rates and laycan
charter contracts (i.e., laycan periods) in this market. The results periods are interrelated and should be modeled as a system of
of the investigations will be useful to shipowners, charterers, and simultaneous equations. If so, the individual estimation of the
investors in their decision-making processes with respect to above two equations using the Ordinary Least Squares (OLS)
tanker ships. method will yield biased and inconsistent parameter estimates.
If a simultaneous interrelationship exists between the laycan
period and the freight rate of a shipping contract, then
3. Methodology Eqs. (1) and (2) must be re-specified to reflect this simultaneous
interrelationship—i.e., by including LCi,t as an explanatory vari-
In order to investigate the vessel and voyage specific determi- able in the fri,t equation (1) and including fri,t as an explanatory
nants of the freight rates for tanker shipping contracts, the variable in the LCi,t equation (2). Thus, Eqs. (1) and (2) rewritten as
following model can be estimated: a simultaneous equations system may be specified as
fri,t ¼ a0 þ a1 BDTIt þ a2 HTi þ a3 UTi þ a4 AGEi þ a5 AGE2i fri,t ¼ a0 þ a1 BDTIt1 þ a2 LCi,t þ a3 HTi þ a4 UTi þ a5 AGEi þ a6 AGE2i
X K
þ a6 VOLt þ dj RTj,i,t þvi,t ; vi,t  iidð0, s2 Þ ð1Þ X
K
þ a7 VOLt1 þ d1,j RTj,i,t þ vi,t LCi,t ¼ b0 þ b1 fri,t þ b2 LCi1,t
j¼1
j¼1
þ b3 HTi þ b4 UTi þ b5 AGEi þ b6 VOLt1
!
where fri,t is the log of the freight rate of the ith fixture (contract) XK vi,t
at time t, HTi is a dummy variable representing the vessel’s hull þ d2,j RTj,i,t þ ei,t xi ¼  IIDð0, RÞ ð3Þ
j¼1
ei,t
type in the ith fixture (i.e., a single hull versus a double hull), UTi
is the deadweight utilization ratio of the fixture expressed as the
ratio of cargo tonnes to deadweight (dwt) of the vessel,3 and AGEi where xi is the vector of residuals, which follows a bivariate
is the vessel’s age in contract i. The square of the age variable, distribution with zero mean and variance–covariance matrix R.
AGE2i , also appears in Eq. (1) since the operational performance, When specifying a system of simultaneous equations, there are
technological efficiency, and quality standard of ships decline as two main issues that should be considered. The first issue is the
they get older, and consequently their hire rate becomes a problem of identification, which arises when parameters of the
nonlinear function of age. The binary dummy variable, RTi,j,t, structural equation cannot be deduced from the parameters of
identifying route j for contract i at time t, is included in Eq. (1) the reduced form model. Therefore, for a system of simultaneous
to differentiate the fixtures according to the route in which the equations to be identified, both the order and the rank conditions
vessel is employed (see Section 4 for detailed explanation of should be satisfied.4 The system of equations presented in equation
routes). Thus, the coefficient of dummy variable, dj, signifies the system (3) satisfies both the order and the rank conditions.5 The
difference in average freight rate for route j in comparison to second issue is the estimation method. Because of the correlation
average freight rate in all unidentified fixture routes (‘‘other between variables and error terms in each equation as well as cross
routes’’), depicted by constant a0. Furthermore, to control for correlation among error terms, the OLS estimation method is not an
macro-determinants of tanker freight rates in the model, we appropriate estimation technique for estimating parameters of
include the log of the Baltic Dirt Tanker Index (BDTIt), a bench- simultaneous equations system. An appropriate estimation techni-
mark indicator for the condition of the tanker freight market, and que is the Three Stage Least Squares (3SLS) estimation technique,
the BDTI (VOLt), measuring the volatility of the freight market. since it yields unbiased, consistent, and efficient estimates of the
Determinants of the timing of tanker shipping contracts by parameters of the system in the presence of heteroscedasticity and
charterers are investigated by regressing the fixture laycan period contemporaneous correlation in the errors across equations.
on the hypothesized explanatory variables, including the route,
tanker market condition, and various vessel characteristics such
as hull type, utilization ratio of the vessel, and vessel age. Hence, 4. Description of data
the laycan regression model is specified as
LCi,t ¼ b0 þ b1 HTi þ b2 UTi þ b3 AGEi þ b4 VOLt The tanker fleet is broadly differentiated into five size classes:
XK Very Large Crude Carriers (VLCC) (more than 160,000 dwt, nor-
þ lj RTj,i,t þ ei,t ; ei,t  iidð0, s2 Þ ð2Þ mally 250,000–320,000 dwt), Suezmax (130,000–160,000 dwt),
j¼1 Aframax (70,000–120,000 dwt), Panamax (50,000–70,000 dwt),
and Handysize (20,000–45,000 dwt).6 The two larger size tankers,
where variables HTi, UTi, AGEi, and VOLt, are hypothesized to namely VLCC and Suezmax, are only involved in transportation of
determine charterers’ decisions on the timing of hiring tankers,
i.e. the laycan period. The route dummy variable, RTj,i,t, is also 4
There are two conditions that must be satisfied for the system of simulta-
included to examine possible differences in laycan periods of neous equations to be identified. These are the order and the rank conditions of
contracts across different tanker routes. Thus the coefficient of the identifications. The order condition, which is a necessary but not sufficient
condition for identification of simultaneous system of equations, requires the
route dummy variable, lj, signifies the difference in the laycan
number of excluded and predetermined variables in each equation to be at least
duration for route j in comparison to average laycan period for all less than the number of endogenous variables minus one. The two equations in
unidentified routes (‘‘other routes’’) depicted by constant b0. the system defined in (3) satisfy this condition; that is, the freight equation is
exactly identified and the laycan equation is over-identified. The over-identifica-
tion is not a major problem because we use the 3SLS estimation method, which
3
The utilization variable UT is constructed as the ratio of cargo size to yields efficient, unbiased, and consistent estimates.
5
deadweight (dwt) of the vessel in a fixture. This is because the high correlation The variable LCi  1,t is included in the laycan equation to achieve both
between the cargo size and dwt can lead to the problem of multicolinearity when identification of the system and reduce serial correlation in residuals.
6
estimating the model, whereas inclusion of the ratio of cargo to deadweight in the In December 2008, the tanker fleet was comprised of 517 VLCC (153.6 mil-
model eliminates this problem. A negative sign for the coefficient of this variable is lion dwt), 364 Suezmax (55.4 million dwt), 771 Aframax (79.9 million dwt), 363
expected, because as dwt increases the freight rate decreases and for a vessel with Panamax (25.6 million dwt), and 2858 Handysize (89.3 million dwt) tankers. The
a given dwt, the smaller the size of the cargo, the higher should be the freight rate proportion of single-hull (single side or bottom) tankers in each sector was VLCC
to satisfy the owner to accept the contract. 23.6%, Suezmax 12.9%, Aframax 11.9%, Panamax 15.2%, and Hnadysize 15.2%.
668 A.H. Alizadeh, W.K. Talley / Transport Policy 18 (2011) 665–675

crude oil. Aframax vessels are also involved in crude oil trans- with 378 fixtures. The other routes constitute between 1.43% and
portation, but occasionally are hired for transportation of oil 7.69% of the total fixtures in the VLCC sector.
products. Smaller size tankers such as the Panamax and Handy- The number of routes defined for the Suezmax market is 15
size vessels are mainly involved in the transportation of clean and (see Table 1) and covers 54.01% of the liquid trading routes; the
dirty oil products, although they are occasionally employed in remainder of the fixtures are spread across many routes with less
short-haul crude oil transportation.7 This paper focuses on the trading activities. The most active Suezmax routes are the West
three larger size tankers, since they are involved in transportation Africa–US Gulf with 345 fixtures (8.64% of total fixtures), Black
of one commodity (crude oil), serving a relatively limited number Sea–UK and continent Europe with 340 fixtures (8.51% of total
of trading routes, thereby simplifying the empirical analysis of fixtures), and West Africa–US Atlantic Coast with 310 fixtures
this paper. (7.76% of total fixtures). The routes with less trading activities are
The data for this study were collected from Clarkson’s the Persian Gulf–China (42 fixtures), Ceyhan–UK and Continent
Research Services Ltd. website, Shipping Intelligence Network Europe (44 fixtures), and Ceyhan–US Gulf with 55 fixtures. Other
(SIN),8 and comprise information on tanker spot fixtures for the routes with less trading activities constitute between 1.48% and
period January 2006 to March 2009. The fixtures for VLCC, 5.53% of the total trading activities in the Suezmax sector.
Suezmax, and Aframax tankers are considered for the analysis. In the Aframax trades, 6682 fixtures are utilized in this study.
The data include such information as vessel characteristics, The 16 high-liquidity routes found in Table 1 represent 55.97% of
voyage characteristics, shipowner, and charterer. After filtering the total trading activities in this market. The most liquid routes
the data for missing values, omitted information, and other in the Aframax market are the Caribbean–US Gulf with 617
unusable observations, a total of 3435, 3994, and 6682 fixture fixtures (9.23% of total fixtrues), followed by Black Sea–
observations remain for VLCC, Suezmax, and Aframax tankers, Mediterranean with 589 fixtures (8.81% of total fixtures), and
respectively.9 Specifically, the information on each fixture con- Primorsk–UK and Continent Europe with 562 fixtures (8.41% of
sists of the vessel’s name, size, age, type, dwt, and the owner of total fixtures). Other routes comprise between 0.55% and 6.26% of
the vessel as well as the loading, discharging ports and regions, the total Aframax trades.
the freight rate for the fixture, the fixture’s date, the layday, the The descriptive statistics of the laycan period of fixtures for
cargo type and size, and the charterer. The Baltic Dirty Tanker different size tankers are also found in Table 1. It is interesting to
Index (BDTI) values were also obtained from SIN over the same note that the average (or mean) laycan period is directly related
period. to vessel size. For instance, average laycan period for VLCC
Spot (or voyage) tanker freight rates that have been expressed, fixtures is 20.7 days, while the average laycan period for Suezmax
negotiated ,and agreed upon are reflected in the Worldscale and Aframax tanker fixtures are 19.4 and 12.6 days, respectively.
index. The Worldscale index is essentially a measure of the The variation (standard deviation) of laycan period for different
breakeven rate of a standard tanker on a round trip between a tanker sizes seems to be more or less similar with 6.4 days for
loading port and the discharging port under certain assumptions VLCC fixtures, 6.6 days for Suezmax, and 5.9 days for Aframax
regarding the vessel’s specifications, port charges, fuel prices, and fixtures. One reason for the existence of a direct relationship
other factors. This index has evolved and has been used by the between laycan period and tanker size may be the duration of the
tanker industry since World War II. The breakeven rate for the voyage. The average duration of the voyage for larger tankers is
standard tanker is calculated on a US dollar per metric ton ($/mt) longer than that for smaller tankers because larger tankers due to
basis by the Worldscale association for every possible tanker their economies of scale are employed to serve trades on longer
route around the world and published every January. The calcu- routes, whereas smaller tankers (Aframax) are employed in
lated breakeven freight rate in every tanker route is known as the relatively shorter routes. This means that owners and charters
‘‘flat rate’’, which is considered as Worldscale 100 for that route.10 of larger tankers tend to plan their voyage requirements and
Following the publication of these rates, charterers, tanker own- chartering activities relatively earlier compared to owners and
ers, tanker operators, and brokers negotiate and draft shipping charterers of smaller tankers, which results in a longer average
contracts on the basis of Worldscale. Perhaps the main benefit of laycan period for larger tankers compared to smaller ones.
the Worldscale index in the tanker freight market is that if fuel The average laycan period, however, appears to vary across
and other costs are assumed to be the same across different different tanker routes. For instance, the average laycan period for
routes, then one can easily compare the profitability of voyages by the VLCC fixtures varies between 17.2 days for route 10 (Persian
simply comparing the market or negotiated rates across tanker Gulf–India) and 26.3 days for route 13 (West Africa–US Gulf). The
routes. standard deviations of VLCC laycan periods across different routes
Table 1 presents the routes used in this study for each vessel seem to be more stable with values between 4.8 and 6.7 days. For
type. For the VLCC trade, 14 major routes are found in Table 1, Suezmax fixtures, the average laycan period varies from 11.3 days
representing more than 75% of the fixtures; the remaining in route 1 (Caribbean–US Gulf) and 25.2 days in route 10 (West
fixtures are for many minor routes with less trading activities. Africa–US Atlantic Coast), with standard deviations ranging
Eleven of these fourteen routes are out of the Persian Gulf, two are between 4.3 and 6.7 days across different routes. The data
out of West Africa, and one out of the North African port of Arzew in Table 1 also suggest that tanker routes out of West Africa
(Algeria). The most active route in the VLCC market is the Persian seem to have longer laycan periods compared to tanker routes in
Gulf–Korea with 426 fixtures, followed by West Africa–US Gulf the Caribbean. In the Aframax sector, the average laycan period is
between 6.5 days in route 2 (East Coast Mexico–US Gulf) and 19.8
7
days in route 11 (Persian Gulf–India), while standard deviations
The employment of Handysize tankers in oil product transportation is
range from 3.0 to 9.1 days. It can be noted that while the average
mainly due to the small parcel size of petroleum products, which rarely exceed
60,000 tons. laycan period seems to be lower in the Aframax sector compared
8
www.clarksons.net. to the VLCC and Suezmax sectors, there seems to be higher
9
The data used here is tanker voyage contracts by observations every day. variation in the laycan period of Aframax tanker fixtures.11
Therefore, although the structure of the data can be considered both cross-
sectional and time-series, the difference here is that the number of observations
every day may vary depending on the number of actual fixtures reported.
10 11
See www.worldscale.co.uk for detailed definitions and explanations on how The relationship between the duration of the laycan and vessel routes will
the Worldscale flat rate is calculated. also be addressed in the empirical results section of this paper.
A.H. Alizadeh, W.K. Talley / Transport Policy 18 (2011) 665–675 669

Table 1
Descriptive statistics for routes and laycan periods for different size tankers.

Route Number of fixtures Laycan period (days)

No. Percentage Mean Median Min Max SD

VLCC
1 Persian Gulf–US Gulf 203 5.91 21.1 21.0 6.0 35.0 6.1
2 Persian Gulf–UK/Continent 49 1.43 22.1 21.0 9.0 35.0 6.5
3 Persian Gulf–Singapore 124 3.61 19.2 19.0 0.0 27.0 5.1
4 Persian Gulf–Japan 213 6.20 20.8 21.0 7.0 35.0 4.8
5 Persian Gulf–China 264 7.69 19.0 19.0 0.0 41.0 5.1
6 Persian Gulf–Taiwan 199 5.79 17.7 18.0 0.0 46.0 5.2
7 Persian Gulf–Korea 426 12.40 20.1 20.0 1.0 35.0 5.5
8 Persian Gulf–Thailand 230 6.70 18.5 19.0 2.0 51.0 5.5
9 Persian Gulf–East 138 4.02 20.3 21.0 2.0 33.0 5.2
10 Persian Gulf–India 146 4.25 17.2 18.0 1.0 35.0 6.6
11 Persian Gulf–Ain Sukhna 99 2.88 17.5 17.0 1.0 49.0 6.0
12 Arzew/Algeria–US Gulf 59 1.72 22.7 23.0 3.0 35.0 6.2
13 West Africa–US Gulf 378 11.00 26.3 27.0 0.0 40.0 5.8
14 West Africa–China 61 1.78 25.4 25.0 11.0 52.0 6.3
15 Other routes 846 24.63 20.7 21.0 0.0 49.0 6.7
Total fixtures 3435 20.7 21.0 0.0 52.0 6.4

Suezmax
1 Caribbean–US Gulf 75 1.88 11.3 11.0 3.0 21.0 4.5
2 East Coast Mexico–Spain 59 1.48 14.9 16.0 0.0 28.0 5.5
3 Black Sea–UK/Continent 340 8.51 17.6 17.5 2.0 44.0 4.6
4 Sidi Kerrir–Mediterranean 131 3.28 13.0 13.0 2.0 26.0 4.5
5 Arzew/Algeria–US Gulf 71 1.78 16.7 18.0 1.0 28.0 6.1
6 Ceyhan–UK/Continent 44 1.10 16.0 16.5 5.0 30.0 5.1
7 Ceyhan–US Gulf 55 1.38 16.7 17.0 6.0 27.0 4.6
8 West Africa–Brazil 102 2.55 20.0 20.0 0.0 51.0 5.8
9 West Africa–UK/Continent 215 5.38 23.7 24.0 5.0 44.0 4.5
10 West Africa–US Atlantic Coast 310 7.76 25.2 25.0 2.0 45.0 4.3
11 West Africa–US Gulf 345 8.64 23.0 24.0 1.0 33.0 4.9
12 West Africa–ST Criox 84 2.10 22.8 24.0 1.0 33.0 5.1
13 Persian Gulf–India 221 5.53 19.2 20.0 4.0 33.0 6.0
14 Persian Gulf–China 42 1.05 19.0 19.0 4.0 31.0 6.2
15 Marsa Bahhayer–Far East 63 1.58 19.3 20.0 5.0 34.0 6.5
16 Other routes 1837 45.99 18.5 19.0 0.0 41.0 6.7
Total fixtures 3994 19.4 20.0 0.0 51.0 6.6

Aframax
1 Caribbean–US Gulf 617 9.23 8.6 9.0 0.0 39.0 3.8
2 East Coast Mexico–US Gulf 418 6.26 6.5 7.0 0.0 33.0 3.0
3 Bahamas–US Gulf 62 0.93 7.7 9.0 1.0 13.0 3.1
4 Primorsk–UK/Continent 562 8.41 13.0 13.0 3.0 44.0 3.8
5 North sea–UK/Continent 170 2.54 9.1 9.0 1.0 17.0 3.0
6 Black Sea–Mediterranean 589 8.81 14.7 15.0 1.0 28.0 3.9
7 Ceyhan–Mediterranean 172 2.57 11.3 11.0 2.0 43.0 4.7
8 Sidi Kerir–Mediterranean 327 4.89 11.1 11.0 0.0 46.0 4.0
9 Libya–Mediterranean 195 2.92 10.1 10.0 1.0 17.0 3.2
10 Algeria–Mediterranean 60 0.90 10.2 10.0 1.0 18.0 3.7
11 Persian Gulf–India 85 1.27 19.8 19.0 4.0 48.0 9.1
12 Malaysia–India 92 1.38 16.7 17.0 0.0 27.0 5.8
13 Indonesia–Australia 113 1.69 15.9 15.0 3.0 29.0 5.6
14 Indonesia–Korea/Japan 126 1.89 15.1 15.0 2.0 28.0 4.6
15 Marsa Bahhayer–Far East 115 1.72 19.5 19.0 1.0 39.0 6.2
16 Vung Tau–Australia 37 0.55 15.9 17.0 4.0 26.0 5.0
17 Other routes 2942 44.03 13.8 13.0 0.0 40.0 6.3
Total fixtures 6682 12.6 12.0 0.0 48.0 5.9

Sample period: 1 January 2006–31 March 2009, consists of 3435 VLCC fixtures, 3994 of Suezmax fixtures, and 6682 Aframax fixtures.

The descriptive statistics in Table 2 reveal that worldscale The descriptive statistics in Table 2 for tanker age show that
freight (charter) rates are higher for smaller than larger tankers. the average age of the VLCC fleet is 10.31 years versus 7.48 and
For instance, the average (or mean) freight rate for a VLCC in the 7.49 years for Suezmax and Aframax tankers, respectively. It is
sample is 90.91WS points, while the average freight rate for also interesting to note that the maximum ages of Suezmax and
Suezmax and Aframax tankers are 139.25WS and 164.04WS Aframax tankers are 32 and 38 years, respectively, which are
points, respectively. This is expected, since vessels exhibit econo- higher than the maximum age of VLCC tankers of 25 years.
mies of ship size at sea, i.e., the $/tonne of transport cost declines Utilization ratio statistics are also found in Table 2. The
as vessel size increases. The freight rates for all tanker sizes show utilization ratio is the ratio of the amount of cargo in each fixture
significant skewness and kurtosis, with the Bera and Jarque to the dwt (or carrying) capacity of the vessel. The average ratio of
(1980) test for normality rejecting the null hypothesis that 90.8% for VLCC tankers is higher than 84% and 77% for Suezmax
freight rates are normally distributed at conventional significance and Aframax tankers, respectively. The standard deviations of the
levels. utilization ratios are inversely related to the sizes of vessels, e.g.,
670 A.H. Alizadeh, W.K. Talley / Transport Policy 18 (2011) 665–675

Table 2
Descriptive statistics of explanatory variables.

Age Dwt Cargo Utilization Laycan Rate BDTI


(years) (000 mt) (000 mt) ratio days (WS) index

VLCC
Mean 10.31 289.84 262.41 0.908 20.66 90.91 1255.34
Median 11.00 298.68 260.00 0.907 21.00 80.00 1214.00
Maximum 25.00 442.47 400.00 1.000 52.00 330.00 2347.00
Minimum 0.00 239.78 80.00 0.265 0.00 23.00 524.00
Std. dev. 6.16 20.155 13.795 0.053 6.38 43.19 331.63
Skewness  0.061  0.376  3.372  2.380  0.017 1.587 0.681
Kurtosis 1.76 4.08 44.16 26.01 3.72 5.92 3.84
JB test 221.4 248.5 248935.1 79013.5 74.8 2663.9 367.4
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000

Suezmax
Mean 7.48 153.05 129.12 0.84 19.36 139.25 1264.74
Median 6.00 151.46 130.00 0.86 20.00 130.00 1218.00
Maximum 32.00 167.55 150.00 1.00 51.00 450.00 2347.00
Minimum 0.00 121.00 50.00 0.34 0.00 30.00 511.00
Std. dev. 5.92 7.59 11.19 0.07 6.55 52.82 350.39
Skewness 0.67  0.80  3.68  2.49  0.21 1.14 0.63
Kurtosis 2.46 4.83 17.04 11.95 2.92 5.03 3.54
JB test 349.3 976.2 41801.7 17468.1 30.3 1555.7 314.0
Probability 0.000 0.000 0.000 0.000 0.000 0.000 0.000

Aframax
Mean 7.49 104.12 80.26 0.77 12.64 164.04 1266.63
Median 6.00 105.25 80.00 0.76 12.00 153.00 1222.50
Maximum 38.00 117.10 110.00 1.00 48.00 430.00 2347.00
Minimum 0.00 81.28 30.00 0.31 0.00 49.00 540.00
Std. dev. 6.05 7.07 8.59 0.09 5.92 58.91 349.38
Skewness 1.00  0.42 0.82 0.34 0.86 0.92 0.54
Kurtosis 3.75 2.96 4.33 3.09 4.56 3.97 3.44
JB test 1273.1 199.2 1236.3 128.9 1498.8 1205.3 384.7
Probability 0.000 0.000 0.000 0.000 0.000 0.000 0.000

Sample period: 1 January 2006–31 March 2009, consists of 3435 VLCC fixtures, 3994 of Suezmax fixtures, and 6682 Aframax fixtures.
JB is the Bera and Jarque (1980) test for normality, which follows w2ð2Þ . The 5% critical value for this test is 5.991.

the standard deviations in Table 2 increase from 5.3% for VLCC interrelated. However, before estimating the system of simulta-
tankers to 7.0% and 9.0% for Suezmax and Aframax tankers, neous equations, it has to be examined whether a simultaneous
respectively. In fact, the estimated standard deviations suggest relationship exists between tanker freight rates and laycan periods.
that the size of cargoes transported by larger vessels is more in This is investigated by using the Hausman (1978) test for simulta-
line with size of these vessels, while cargo sizes of smaller vessels neity. The Hausman test is performed using the two-step method
are less in line with their size (given the greater variability in their of Davidson and Mackinnon (1993). The test is performed by first
utilization ratios). estimating the laycan equation (2) as the auxiliary regression and
The average size of the cargo transported by different size then using the estimated residuals from this estimated equation as
tankers in Table 2 reveals that VLCC tankers transported 262,410 observations for the explanatory variable (e^ i,t ) in the freight rate
metric tonnes (mt) on average, while Suezmax and Aframax equation (1). In addition, the Hausman test is performed by first
tankers transported 129,120 and 80,260 mt on average, respec- estimating freight equation (1) as the auxiliary regression and then
tively. The variation (or standard deviation) in cargo size for using the estimated residuals (v^ i,t ) from this equation as observa-
different size tankers declines in moving from larger size tankers tions for the explanatory variable in the laycan equation (2). The
to smaller ones—i.e., declines from 13.8 for VLCC tankers to 11.2 results of the Hausman tests are found in Table 3. It can be seen that
and 8.6 for Suezmax and Afamax tankers, respectively. However, the coefficient of fitted-residuals variable (e^ i,t ) in the freight
when the standard deviations are adjusted with respect to their equation and the coefficient of fitted-residuals variable (v^ i,t ) in the
means, i.e., divided by their means to compute the Coefficient of laycan equation are both statistically significant at the 1% level for
Variation, CV, a different conclusion follows with respective to VLCC and Suezmax tankers, supporting the notion of a contem-
relative variability—i.e., the relative variability in cargo size poraneous relationship between laycan periods and freight rates.
increases as tankers become smaller in size. For instance, the However, the null hypothesis of no simultaneity between freight
ratio of cargo size variation to average cargo size for VLCC is 5.3% rate and laycan period could not be rejected in the Aframax market.
(13.795 mt/262.410 mt), while the same ratio for Suezmax and For consistency in the presentation of estimation results, the
Aframax tankers is 8.7% (11.190 mt/129.120 mt) and 10.7% estimate of the simultaneous equation system (3) for the Aframax
(8.590 mt/80.260 mt), respectively. market is presented as well. Estimates of the simultaneous-equa-
tion system (3), using the 3SLS estimation method, are reported in
Tables 4–6 for VLCC, Suezmax, and Aframax tankers, respectively.12
5. Empirical results
12
As suggested by an anonymous referee, the system of simultaneous Eq. (3)
The empirical investigation is undertaken by estimating the was also estimated using Full Information Maximum Likelihood (FIML) and
parameters of the system of simultaneous equation (3), which that Generalized Method of Moments (GMM). The results of both estimation methods
assumes the laycan period and freight rates of tanker fixtures are are consistent with the ones from 3SLS. However, the benefit of 3SLS is that the
A.H. Alizadeh, W.K. Talley / Transport Policy 18 (2011) 665–675 671

Table 3
Hausman test for simultaneity of freight rates and laycan periods.

P
K
Main regression LCi,t ¼ b0 þ b1 HTi þ b2 UTi þ b3 AGEi þ b4 VOLt þ lj RTj,i,tr þ ei,t ; ei,t  iidð0, se 2 Þ
j¼1
P
K
Test regression fri,t ¼ a0 þ a1 BDTIt þ a2 HTi þ a3 UTi þ a4 AGEi þ a5 AGE2i þ a6 VOLt þ dj RTj,i,t þ fe^ i,t þ vi,t ; vi,t  iidð0, sv 2 Þ
j¼1
VLCC Suezmax Aframax

Coeff. p-val Coeff. p-val Coeff. p-val


Significance of f 0.007 0.000 0.001 0.000 0.0004 0.413

P
K
Main regression fri,t ¼ a0 þ a1 BDTIt þ a2 HTi þ a3 UTi þ a4 AGEi þ a5 AGE2i þ a6 VOLt þ dj RTj,i,t þ vi,t ; vi,t  iidð0, sv 2 Þ
j¼1
P
K
Test regression LCi,t ¼ b0 þ b1 HTi þ b2 UTi þ b3 AGEi þ b4 VOLt þ lj RTj,i ,t þ fv^ i,t þ ei,t ; ei,t  iidð0, se 2 Þ
j¼1
VLCC Suezmax Aframax

Coeff. p-val Coeff. p-val Coeff. p-val


Significance of f 3.510 0.000 1.004 0.047 0.280 0.373

Sample period: 1 January 2006–31 March 2009, consist of 3435 VLCC fixtures, 3994 of Suezmax fixtures, and 6682 Aframax fixtures.
The standard errors of the test regression are corrected for heteroscedasticity and autocorrelation using the Newey and West (1987) method.

5.1. VLCC tankers utilization ratio coefficient of  0.876 indicates that a 1% increase
in the utilization ratio will result in a decline in the freight rate by
The estimated coefficients of simultaneous equation system (3) 0.876%. The signs of the coefficients for age and age-squared
for the VLCC contracts are reported in Table 4. The results reveal suggest a quadratic relationship between age and VLCC
that in the freight equation 10 of the 14 route dummy variables freight rates.
are statistically significant at the 5% level. The estimated coeffi- Estimated coefficients of the laycan equation in equation
cients for 4 out of these 10 routes are negative, indicating lower system (3) for VLCC tankers found in Table 4 reveal that
freight rates on average compared to other routes. These four significant differences exist in the laycan period of the VLCC
routes are—Persian Gulf–the US Gulf (route 1) and Persian Gulf– charter contracts as indicated by the values of the coefficients of
UK and Continent Europe (route 2), where larger than standard the dummy route variables. Specifically, the laycan periods for
VLCCs and ULCCs operate, as well as the West Africa–US Gulf routes 3, 5, 6, 8, 10, and 11 are shorter than those for the other
(route 13) and West Africa–China (route 14). The first three routes routes as indicated by the negative and significant coefficients of
are westbound routes, while the last is a back-haul route—VLCCs their route dummy variables. On the other hand, the significant
are returning to the Persian Gulf, where vessel owners would be and positive coefficients of dummy variables for routes 12, 13,
expected to lower their rates on this route in order to fix a and 14 indicate that the laycan period for charter contracts in
contract that optimizes their overall voyage revenue on the way these routes is longer than in other routes by 2.561, 5.647, and
back to the East from either the US Gulf or Europe (which is 5.160 days, respectively. These routes extend from North Africa
expected to be lower compared to other routes). The routes with and West Africa to the US Gulf and from West Africa to China. The
positive and significant coefficients of dummy variables are all empirical evidence suggests that the laycan periods of VLCC
east bound routes from the Persian Gulf. The estimated coefficient routes out of these regions are generally longer. The longer laycan
of the route dummy variable is not significantly different from periods could be due to trade and terminal requirements as well
zero for route 4 (Persian Gulf–Japan), route 7 (Persian Gulf– as availability of tankers in these regions. In the VLCC market, the
Korea), and route 12 (Algeria–US Gulf). normal waiting region is considered to be the Persian Gulf, where
The significant and positive coefficient of the log of BDTI level vessels are directed for their next cargo. However, any fixture out
was expected, since this variable reflects the tanker market of the general trading route for VLCCs should be done with ample
condition as well as the effect of macroeconomic determinants time. Furthermore, route 14, in particular, is considered to be a
of tanker freight rates. In addition, the magnitude of its estimated back-haul route for vessels trying to utilize the ballast return leg
coefficient (1.165) indicates that for every 1% change in BDTI, the from the US Gulf to the Persian Gulf. As a result, some chartering
VLLC freight rates change by 1.165%, revealing high sensitivity in managers try to find employment in this route in advance to
this sector to changes in the market. The positive coefficient of maximize the round trip voyages from the Persian Gulf to the US
lagged BDTI volatility (0.227) indicates that as market volatility and Europe.
increases, the level of freight rate for hiring a tanker increases as The positive coefficient of the freight rate (2.552) in the laycan
well. This is in line with asset pricing theories, which suggest a equation suggests that increases in freight rates lead to increases
direct relationship between risk and return. The negative esti- in laycan periods. Higher freight rates may be an indication of a
mated coefficients of hull type, deadweight utilization ratio, and lower supply or higher demand (or both) and unavailability of
age suggest that these variables have a negative effect on tanker ships. If so, charterers are expected to enter the market earlier and
freight (charter) rates. For instance, the coefficient estimates thus hire tankers earlier, thereby avoiding disruptions that would
indicate that freight (charter) rates for single-hull VLCCs tankers occur in not being able to hire tankers at desired times. The
are 13.8% lower than for double-hull VLCCs. The deadweight negative coefficient of BDTI volatility (  3.679) suggests that
increases in this volatility reduce the laycan periods of charter
contracts. Specifically, the laycan period declines by 3.7 days if
(footnote continued)
the BDTI volatility increases by one standard deviation. The
standard errors are corrected for heteroscedastiocity thought the estimation positive coefficient of the deadweight utilization ratio (10.885)
procedure. suggests that VLCC laycan periods are longer for full-load than for
672 A.H. Alizadeh, W.K. Talley / Transport Policy 18 (2011) 665–675

Table 4
3SLS estimation results for VLCC freight rate and laycan period.

X
K
fri,t ¼ a0 þ a1 BDTIt1 þ a2 LCi,t þ a3 HTi þ a4 UTi þ a5 AGEi þ a6 AGE2i þ a7 VOLt1 þ d1,j RTj,i,t þ vi,t !
j¼1 vi,t
xi ¼ ei,t  IIDð0, RÞ
X
K
LCi,t ¼ b0 þ b1 fri,t þ b2 LCi1,t þ b3 HTi þ b4 UTi þ b5 AGEi þ b6 VOLt1 þ d2,j RTj,i,t þ ei,t
j¼1

fri Equation LCi Equation

Variables Coeff. p-val Variables Coeff. p-val

a0 Constant  3.803 0.000 b0 Constant  1.666 0.489


a1 Log BDTI, BDTIt  1 1.165 0.000 b1 Log freight rate, fri 2.552 0.000
a2 Laycan, LCi 0.032 0.000 b2 Layan, LCi  1 0.155 0.000
a3 Hull type, HTi  0.138 0.000 b3 Hull type, HTi 0.892 0.017
a4 Utilization, UTi  0.876 0.000 b4 Utilization, UTi 10.885 0.000
a5 Age, AGEi 0.009 0.000 b5 Age, AGEi  0.094 0.001
a6 AGE2i  0.0003 0.012
a7 BDTI volatility, VOLt  1 0.227 0.000 b6 BDTI volatility VOLt  1  3.679 0.000

d1,1 Persian Gulf–US Gulf  0.240 0.000 d2,1 Persian Gulf–US Gulf 0.821 0.069
d1,2 Persian Gulf–UK/Continent  0.121 0.003 d2,2 Persian Gulf–UK/Continent 1.290 0.118
d1,3 Persian Gulf–Singapore 0.061 0.024 d2,3 Persian Gulf–Singapore  1.511 0.005
d1,4 Persian Gulf–Japan 0.002 0.943 d2,4 Persian Gulf–Japan  0.271 0.529
d1,5 Persian Gulf–China 0.050 0.014 d2,5 Persian Gulf–China  1.752 0.000
d1,6 Persian Gulf –Taiwan 0.103 0.000 d2,6 Persian Gulf–Taiwan  2.886 0.000
d1,7 Persian Gulf–Korea 0.016 0.355 d2,7 Persian Gulf –Korea  0.553 0.103
d1,8 Persian Gulf–Thailand 0.091 0.000 d2,8 Persian Gulf–Thailand  1.826 0.000
d1,9 Persian Gulf–East 0.034 0.179 d2,9 Persian Gulf–East  0.827 0.105
d1,10 Persian Gulf–India 0.111 0.000 d2,10 Persian Gulf–India  3.395 0.000
d1,11 Persian Gulf–Ain Sukhna 0.090 0.006 d2,11 Persian Gulf–Ain Sukhna  3.361 0.000
d1,12 Arzew/Algeria–US Gulf  0.062 0.105 d2,12 Arzew/Algeria–US Gulf 2.561 0.001
d1,13 West Africa–US Gulf  0.171 0.000 d2,13 West Africa–US Gulf 5.647 0.000
d1,14 West Africa–China  0.190 0.000 d2,14 West Africa–China 5.160 0.000

R-bar squared 0.589 0.240


JB 69.17 [0.000] 43.48 [0.000]
WH 10.76 [0.000] 623.4 [0.000]
DW 1.129 2.018
SC 15332.0 [0.000]

Sample consist of 3994 individual VLCC fixtures over the period January 2006 to March 2009.
The system is estimated using 3SLS estimation method. Standard errors are sequentially updated and corrected for heteroscedasticity and serial correlation.
JB is the Jarque and Bera (1980) test for normality of estimated residuals of the system.
WH is the White (1980) test for heteroscedasticity.
DW is the Durbin–Watson test for 1st order serial correlation.
SC test is the system residual portmanteau test for 10th order Autocorrelations in residuals.

part cargoes—likely explained by the fact that full-load cargoes Algeria–US Gulf (route 5), West Africa–UK/Continent (route 9),
generally move on main trade routes and part cargoes mainly West Africa–US Gulf (routes 11), and Persian Gulf–India (route
move on back-haul routes. That is to say, charterers tend to fix 13). The East Coast Mexico–Spain route is considered a back-haul
tankers for part cargoes mainly in the direction that is opposite to route and is thus expected to trade at a rate discount. The results
that of main trade flows; further, for part cargoes, they tend to also reveal that the coefficients of trade route dummy variables 1,
hire vessels that are available at the loading area for immediate 3, 4, 8, and 15 are positive, indicating that the freight rates trade
hire. The negative coefficient of the age variable ( 0.094) at a rate premium that ranges between 4.3% and 23.5% higher
suggests that the laycan period for newer VLCCs is longer than than average (i.e. than the unidentified routes). The estimated
that for older VLCCs; i.e., modern vessels are likely to be hired coefficients of log BDTI and volatility of BDTI are in line with those
earlier than older ones, everything else being equal. Finally, the of the VLCC estimate, i.e., increases in both BDTI level and
positive and significant coefficient of hull type in the laycan volatility lead to increases in Suezmax freight rates. Furthermore,
equation suggests that on average the laycan period for single a positive relationship exists between the freight rate and the
hull tankers is almost one day (0.892) longer than that of double- laycan period, which suggests that there is a premium to hire a
hull tankers. vessel early, while existence of a negative relationship between
freight rates and the hull type of Suezmax tankers, as expected,
5.2. Suezmax tankers suggests that single hull vessels are traded at a discount to
double-hull tankers. The significant coefficient of the laycan
In Table 5 the estimate of the freight rate equation for period indicates that charterers tend to pay a premium when
Suezmax tankers reveals significant differences in hire rates in hiring vessels early. The estimated coefficient of deadweight
different routes—11 of the 15 route dummy variables are statis- utilization variable also reveals that there is a negative relation-
tically significant at the 5% level. Five of the 11 significant route ship between Suezmax tanker freight rates and the deadweight
dummy variables have negative coefficients and thus the trade is utilization ratio, which indicates that freight rates are relatively
at a rate discount, i.e., the East Coast Mexico–Spain (route 2), lower for fixtures with higher deadweight capacity utilization
A.H. Alizadeh, W.K. Talley / Transport Policy 18 (2011) 665–675 673

Table 5
3SLS estimation results for Suezmax freight rate and laycan period.

X
K
fri,t ¼ a0 þ a1 BDTIt1 þ a2 LCi,t þ a3 HTi þ a4 UTi þ a5 AGEi þ a6 AGE2i þ a7 VOLt1 þ d1,j RTj,i,t þ vi,t !
j¼1 vi,t
xi ¼ ei,t  IIDð0, RÞ
X
K
LCi,t ¼ b0 þ b1 fri,t þ b2 LCi1,t þ b3 HTi þ b4 UTi þ b5 AGEi þ b6 VOLt1 þ d2,j RTj,i,t þ ei,t
j¼1

fri Equation LCi Equation

Variables Coeff. p-val Variables Coeff. p-val

a0 Constant  2.226 0.000 b0 Constant  2.714 0.155


a1 Log BDTI, BDTIt  1 1.047 0.000 b1 Log freight rate, fri 2.050 0.000
a2 Laycan, LCi 0.018 0.019 b2 Layan, LCi  1 0.063 0.000
a3 Hull type, HTi  0.087 0.000 b3 Hull type, HTi 1.074 0.006
a4 Utilization, UTi  0.910 0.000 b4 Utilization, UTi 13.474 0.000
a5 Age, AGEi 0.007 0.000 b5 Age, AGEi  0.108 0.000
a6 AGE2i  0.0004 0.000
a7 BDTI volatility, VOLt  1 0.175 0.000 b6 BDTI volatility, VOLt  1  1.186 0.022

d1,1 Caribbean–US Gulf 0.235 0.000 d2,1 Caribbean–US Gulf  7.442 0.000
d1,2 East Coast Mexico–Spain  0.398 0.000 d2,2 East Coast Mexico–Spain  3.651 0.000
d1,3 Black Sea–UK/Continent 0.131 0.000 d2,3 Black Sea–UK/Continent  1.834 0.000
d1,4 Sidi Kerrir–Mediterranean 0.176 0.000 d2,4 Sidi Kerrir–Mediterranean  5.624 0.000
d1,5 Arzew/Algeria–US Gulf  0.063 0.033 d2,5 Arzew/Algeria–US Gulf  1.961 0.005
d1,6 Ceyhan–UK/Continent 0.098 0.012 d2,6 Ceyhan–UK/Continent  3.100 0.000
d1,7 Ceyhan–US Gulf  0.018 0.592 d2,7 Ceyhan–US Gulf  2.263 0.004
d1,8 West Africa–Brazil 0.043 0.067 d2,8 West Africa–Brazil 1.349 0.021
d1,9 West Africa–UK/Continent  0.093 0.022 d2,9 West Africa–UK/Continent 4.738 0.000
d1,10 West Africa –US Atlantic Coast  0.097 0.059 d2,10 West Africa–US Atlantic Coast 6.196 0.000
d1,11 West Africa–US Gulf  0.080 0.024 d2,11 West Africa–US Gulf 4.106 0.000
d1,12 West Africa–St Criox  0.056 0.153 d2,12 West Africa–St Criox 4.002 0.000
d1,13 Persian Gulf–India  0.050 0.007 d2,13 Persian Gulf–India 1.092 0.020
d1,14 Persian Gulf–China  0.029 0.373 d2,14 Persian Gulf–China 0.145 0.872
d1,15 Marsa Bahhayer–Far East 0.155 0.000 d2,15 Marsa Bahhayer–Far East  0.053 0.943

R-bar squared 0.755 0.241


JB 257.29 [0.000] 56.33 [0.000]
WH 6.859 [0.000] 9.756 [0.000]
DW 1.518 1.992
SC 3314.0 [0.000]

Sample consists of 3994 individual Suezmax fixtures over the period January 2006–March 2009.
The system is estimated using 3SLS estimation method. Standard errors are sequentially updated and corrected for heteroscedasticity and serial correlation.
JB is the Jarque and Bera (1980) test for normality of estimated residuals of the system.
WH is the White (1980) test for heteroscedasticity.
DW is the Durbin–Watson test for 1st order serial correlation.
SC test is the system residual portmanteau test for 10th order Autocorrelations in residuals.

compared to part-cargo contracts. Finally, the positive and nega- 5.3. Aframax tankers
tive coefficients of age and age-squared, respectively, confirm that
there is a quadratic relationship between the age of the vessel and The 3SLS estimate of the Aframax freight rate equation
freight rate in the Suezmax tanker sector. reported in Table 6 reveals that 10 of the 16 route dummy
The estimated coefficients of route dummy variables in the laycan variables are significant at the 1% level. In particular the results
equation reported in Table 5 reveal that laycan periods for Suezmax suggest that freight rates in routes 1, 2, 3, 6, 7, and 8 are on
fixtures in routes 1–7 are significantly shorter than those for routes average higher than freight rates in routes 5, 11, and 12. The
8–13. The rationale for these findings are that routes 1–7 are all coefficients of the route dummy variables for routes 1, 2, and
westbound routes with the exception of route 2 (East coast Mexico– 3 indicate that on average the Aframax routes from Caribbean,
Spain), while routes 8–12 are routes out of West Africa, and route 13 East Coast Mexico, and Bahamas to the US Gulf trade at a
is the Persian Gulf–India. The estimation results also suggest that premium of 24.7%, 29.6%, and 25.3%, respectively. Further, the
Suezmax laycan periods are positively related to Suezmax freight Aframax Mediterranean routes – Black Sea, Ceyhan, Sidi Kerir, and
rates, type of hull (single versus double), and the utilization factor, but Libya to the Mediterranean – trade at a premium of 5.1%, 9.2%,
negatively related to the BDTI volatility and tanker age. More 5.6%, and 10.0%, respectively, of the average Aframax freight rates.
precisely, single-hull tankers tend to have a longer laycan period On the other hand, the dummy variable coefficients for the routes
than modern double-hull tankers. One explanation could be that Persian Gulf–India (route 11) and Malaysia–India (route 12), are
single hull tankers are mostly traded in routes with longer laycan negative, indicating trading Aframax rate discounts of 32.2% and
periods such as Persian Gulf–India and the Far East, while routes to 17.0%, respectively.
North America and Europe are served only with double-hull tankers. Furthermore, the estimated coefficients of the freight rate
The positive coefficient of the dwt utilization variable suggests that equation indicate that Aframax freight rates are positively related
laycans for full-load take place well in advance of loading, while part to BDTI and its volatility, but negatively related to the hull type
cargoes tend to be contracted at short notices and thus have relatively and the cargo utilization factor. More precisely, on average, the
short laycans. freight rate for single-hull Aframax tankers tends to be 7.9% lower
674 A.H. Alizadeh, W.K. Talley / Transport Policy 18 (2011) 665–675

Table 6
3SLS estimation results for Aframax freight rate and laycan period.

X
K
fri,t ¼ a0 þ a1 BDTIt1 þ a2 LCi,t þ a3 HTi þ a4 UTi þ a5 AGEi þ a6 AGE2i þ a7 VOLt1 þ d1,j RTj,i,t þ vi,t !
j¼1 vi,t
xi ¼ ei,t  IIDð0, RÞ
X
K
LCi,t ¼ b0 þ b1 fri,t þ b2 LCi1,t þ b3 HTi þ b4 UTi þ b5 AGEi þ b6 VOLt1 þ d2,j RTj,i,t þ ei,t
j¼1

fri Equation LCi Equation

Variables Coeff. p-val Variables Coeff. p-val

a0 Constant  1.532 0.000 b0 Constant  0.027 0.986


a1 Log BDTI, BDTIt  1 0.935 0.000 b1 Log Freight rate, fri 1.791 0.000
a2 Laycan, LCi 0.022 0.006 b2 Layan, LCi  1 0.058 0.000
a3 Hull type ,HTi  0.079 0.000 b3 Hull type, HTi 2.096 0.000
a4 Utilization, UTi  0.574 0.000 b4 Utilization, UTi 6.054 0.000
a5 Age, AGEi 0.002 0.113 b5 Age, AGEi  0.058 0.000
a6 AGE2i 0.000 0.206
a7 BDTI volatility, VOLt  1 0.079 0.000 b6 BDTI volatility, VOLt  1  1.307 0.000

d1,1 Caribbean–US Gulf 0.247 0.000 d2,1 Caribbean–US Gulf  4.655 0.000
d1,2 East Coast Mexico–US Gulf 0.296 0.000 d2,2 East Coast Mexico–US Gulf  6.749 0.000
d1,3 Bahamas–US Gulf 0.253 0.000 d2,3 Bahamas–US Gulf  5.684 0.000
d1,4 Primorsk–UK/Continent 0.020 0.268 d2,4 Primorsk–UK/Continent  1.420 0.000
d1,5 North sea–UK/Continent 0.007 0.868 d2,5 North sea–UK/Continent  4.165 0.000
d1,6 Black Sea–Mediterranean 0.051 0.000 d2,6 Black Sea–Mediterranean 1.003 0.000
d1,7 Ceyhan–Mediterranean 0.092 0.000 d2,7 Ceyhan–Mediterranean  2.226 0.000
d1,8 Sidi Kerir–Mediterranean 0.056 0.022 d2,8 Sidi Kerir–Mediterranean  2.587 0.000
d1,9 Libya–Mediterranean 0.100 0.002 d2,9 Libya–Mediterranean  3.502 0.000
d1,10 Algeria–Mediterranean 0.067 0.087 d2,10 Algeria–Mediterranean  3.201 0.000
d1,11 Persian Gulf–India  0.322 0.000 d2,11 Persian Gulf–India 5.041 0.000
d1,12 Malaysia–India  0.170 0.000 d2,12 Malaysia–India 2.909 0.000
d1,13 Indonesia–Australia  0.034 0.211 d2,13 Indonesia–Australia 2.023 0.000
d1,14 Indonesia–Korea/Japan  0.005 0.836 d2,14 Indonesia–Korea/Japan 1.431 0.002
d1,15 Marsa Bahhayer–Far East  0.049 0.316 d2,15 Marsa Bahhayer–Far East 5.287 0.000
d1,16 Vung Tau–Australia  0.029 0.498 d2,16 Vung Tau–Australia 2.266 0.008

R-bar squared 0.577 0.245


JB 72.17 [0.000] 1061 [0.000]
WH 10.68 [0.000] 607.52 [0.000]
DW 1.664 1.997
SC 1714.7 [0.000]

Sample consists of 6682 individual Aframax fixtures over the period January 2006–March 2009.
The system is estimated using 3SLS estimation method. Standard errors are sequentially updated and corrected for heteroscedasticity and serial correlation.
JB is the Jarque and Bera (1980) test for normality of estimated residuals of the system.
WH is the White (1980) test for heteroscedasticity.
DW is the Durbin–Watson test for 1st order serial correlation.
SC test is the system residual portmanteau test for 10th order Autocorrelations in residuals.

than that of double-hull tankers, while every 1% increase in decisions on when to enter the market, which may reflect their
utilization ratio can lead to a 0.574% decrease in freight rate. perception of the condition of the market and the fact that their
Additionally, Aframax freight rates increase with increases in the entry may increase instability in the market. Additionally, charterers
laycan period, while age of the vessel seems to have no impact on may wish to wait for the market to stabilize and avoid paying a
Aframax freight rates. premium to fix a vessel when the volatility in the market is high.
The estimate of the Aframax laycan equation, also reported A comparison of estimated coefficients of certain variables
in Table 6, reveals that average laycan periods are shorter for routes across different tanker markets points to some interesting con-
1–10 , with the exception of route 6 (Black Sea–Mediterranean), but clusions. First, the estimated coefficients of log of BDTI in the
longer for routes 11–16. Once again, the estimation results suggest freight rate equations (i.e., 1.165, 1.045, and 0.935 for the VLCC,
that laycan periods are longer for routes in the East (routes 11–16, Suezmax, and Aframax tankers, respectively) indicate that the
east of Suez) than routes in the West (routes 1–10, west of Suez). freight rates for larger vessels are more sensitive to market
The age and hull type coefficients suggest that Aframax laycan changes, i.e., BDTI, compared to smaller vessels. Second, the
periods decrease with the age of the vessel and increase if the vessel estimated coefficients of hull type (i.e.  0.138,  0.087, and
is of single-hull. Further, the coefficient of deadweight utilization  0.079 for VLCC, Suezmax, and Aframax tankers, respectively)
(UT) indicates that laycan periods for full-load contracts are longer in the freight rate equation suggest that the differences in the
than those for part-cargo contracts, while the freight rate level has a freight rates for single hull and double hull tankers are relatively
positive effect and the volatility of BDTI has a negative effect on smaller for smaller tankers compared to larger ones. Finally, the
laycan periods. This is again in line with what is observed in the estimated coefficients of BDTI volatility in the freight rate equa-
market for larger tankers, since a higher freight rate is an indication tions (0.227, 0.175, and 0.079 for VLCC, Suezmax, and Aframax
of a tight supply condition and thus forces charterers to fix their tankers, respectively) reveal that freight rates for larger vessels
transportation requirements well in advance. Higher volatility of the are more sensitive to tanker market volatility compared to freight
freight market, on the other hand, makes the charters delay their rates for smaller vessels. This can be explained by the size of the
A.H. Alizadeh, W.K. Talley / Transport Policy 18 (2011) 665–675 675

fleet for larger tankers compared to the fleet size for smaller freight rates and laycan periods for tanker shipping contracts are
tankers (see footnote 6); the inflexibility of larger tankers in interrelated, i.e., the choice of the laycan period can affect the
carrying different cargoes compared to smaller tankers, and the freight rate and alternatively, the level and volatility of the freight
limited number of routes in which larger tankers can operate. rate can influence the choice of the laycan period in tanker
Thus, freight rates for larger tankers are expected to be more chartering.
sensitive to market uncertainty and changes in the market
relative to smaller tankers.
References

6. Conclusions Adland, R., Cullinane, K., 2005. A time-varying risk premium in the term structure
of bulk shipping freight rates. Journal of Transport Economics and Policy 39,
191–208.
The purpose of this paper has been to investigate vessel and Adland, R., Cullinane, K., 2006. The non-linear dynamics of spot freight rates in
voyage determinants (e.g., vessel age, hull type, deadweight tanker markets. Transportation Research Part E 42 (3), 211–224.
utilization, and route) of individual tanker freight (charter) rates Batchelor, R., Alizadeh, A.H., Visvikis, I., 2007. Forecasting spot and forward prices
in the international freight market. International Journal of Forecasting 23 (1),
and laycan periods as well as to investigate the relationship
101–114.
between these rates and the laycan periods of tanker charter Beenstock, M., Vergottis, A.R., 1989. An econometric model of the world tanker
contracts. Data for the investigation were obtained from Clark- market. Journal of Transport Economics and Policy 23 (2), 263–280.
son’s Research Studies website and comprise information on Beenstock, M., Vergottis, A., 1993. Econometric Modelling of World Shipping.
Chapmann & Hall, London.
tanker spot fixtures for the period January 2006–April 2009 for Bera, A.K., Jarque, C.M., 1980. Efficient tests for normality, heteroskedasticity, and
VLCC, Suezmax, and Aframax tankers. serial independence of regression residuals. Economic Letters 6, 255–259.
The investigation reveals several important findings. First, on Brown, G.G., Graves, G.W., Ronen, D., 1987. Scheduling ocean transportation of
crude oil. Management Science 33 (3), 335–346.
average, freight rates of single-hull tankers are at a discount in Davidson, R., Mackinnon, J., 1993. Estimation and Inference in Econometrics.
comparison to those of double-hull tankers. Second, tanker freight Oxford University Press, Oxford.
rates, as expected, are positively and negatively related to the length Dikos, G., Markos, H.S., Papadatos, M.P., Papakonstantinou, V., 2006. Niver lines: a
system-dynamics approach to tanker freight modeling. Interfaces 36 (4),
of the laycan period and the utilization factor (cargo/dwt), respec-
326–341.
tively. Third, a simultaneous equation relationship exists between Glen, D.R., Martin, B.T., 1998. Conditional modelling of tanker market risk using
tanker freight rates and the length of laycan periods. Fourth, tanker route specific freight rates. Maritime Policy and Management 25 (2), 117–128.
freight rates and laycan periods vary across shipping routes, e.g., Hausman, J.A., 1978. Specification tests in econometrics. Econometrica 46 (6),
p1251–p1271.
tankers employed in back-haul routes trade at discounted freight Hawdon, D., 1978. Tanker freight rates in the short and long run. Applied
rates. The laycan periods of routes out of West Africa and Persian Economics 10, 203–217.
Gulf to India and the Far East are generally longer than routes out of Jarque, C.M., Bera, A.K., 1980. Efficient test for normality, homoscedasticity and
serial dependence of regression residuals. Economics Letters 6, 255–259.
the Mediterranean, the Black Sea, the North Sea, the Caribbean, and Kavussanos, M.G., 2003. Time varying risks among segments of the tanker freight
the Baltic Sea. Fifth, the laycan periods of tanker freight contracts markets. Maritime Economics and Logistics 5 (3), 227–250.
vary directly with freight rates and indirectly with freight rate Kavussanos, M.G., Alizadeh, A., 2002. The expectations hypothesis of the term
structure and risk premia in dry bulk shipping freight markets; an EGARCH-M
volatility. This is expected, since higher freight rates generally reflect
approach. Journal of Transport Economics and Policy 36 (2), 267–304.
a lower availability of tanker tonnage—therefore, vessel charterers Koopmans, T.C., 1939. Tanker Freight Rates and Tankship Building. Haarlem,
anticipating tanker shortages will seek to enter the tanker charter Holland.
market earlier, in order to fix their transportation requirements well Laulajainen, R., 2008. Operative strategy in tanker (dirty) shipping. Maritime
Policy and Management 35 (3), 313–339.
in advance. When the volatility and uncertainty of freight rates Lyridis, D.V., Zacharioudakis, P., Mitrou, P., Mylonas, A., 2004. Forecasting tanker
increase, tanker owners and charterers are expected to delay their market using artificial neural networks. Maritime Economics and Logistics 6
letting or hiring of vessels as long as it is feasible and wait to utilize (2), 93–108 June.
Mokia, Z., Dinwoodie, J., 2002. Spatial aspects of tanker lay-times. Journal of
their option value to delay fixing. Sixth, tanker laycan periods are Transport Geography 10, 39–49.
negatively related to tanker age, i.e., newer tankers have longer Newey, W.K., West, K.D., 1987. A simple positive definite heteroskedasticity and
laycans than older tankers. autocorrelation consistent covariance matrix. Econometrica 55, 703–708.
Randers, J., Göluke, U., 2007. Forecasting turning points in shipping freight rates:
The findings of this study may be used in practice by tanker
lessons from 30 years of practical effort. System Dynamics Review 23 (2/3),
owners and tanker charters in negotiating tanker freight rates and 253–284.
contracts—e.g., tanker owners and charterers can determine and Stopford, M., 2009. Maritime Economics. Routledge, London.
negotiate freight rate premiums and discounts, especially in Strandenes, S.P., 1984. Price Determination in the Time-Charter and Second-Hand
Markets, Working paper no. 06, Centre for Applied Research, Norwegian
regard to specific routes and vessel characteristics. From the School of Economics and Business Administration.
point of view of tanker charterers, the decision to enter the Strandenes, S.P., 1999. Is there potential for a two-tier tanker market? Maritime
market to fix vessels may affect market demand and consequently Policy and Management 26 (3), 249–264.
Tamvakis, M.N., Thanopoulou, H.A., 2000. Does quality pay? The case of the dry
tanker freight rates. However, waiting for more favorable freight bulk market. Transportation Research Part E 36 (4), 297–307.
rates may be risky as freight rates tend to move very sharply in a Tamvakis, M.N., 1995. An investigation into the existence of a two-tier spot freight
very short period of time. Therefore, information on the effects of market for crude oil tankers. Maritime Management and Policy 22 (1), 81–90.
Tinbergen, J., 1934. Scheepsruimte en vrachten, De Nederlandsche Conjunctuur
market volatility, vessel type and size, trading route, and cargo
March, pp. 23–35.
size on tanker freight rates may be of value to charters and their Veenstra, A.W., Franses, P.H., 1997. A co-integration approach to forecasting
chartering decisions. From the owners’ point of view, information freight rates in the dry bulk shipping sector. Transportation Research Part A
on the effects of vessel and voyage specific factors on tanker 31A (6), 447–459.
White, H., 1980. A heteroscedasticity-consistent covariance matrix estimator and a
freight rates can be used in their investment, operations, and direct test for heteroscedasticity. Econometrica 48, 55–68.
vessel deployment strategies. Finally, it has been shown that Zannetos, Z.S., 1966. The Theory of Oil Tankship Rates. MIT Press, Massachusetts.

Das könnte Ihnen auch gefallen