- Email: [email protected]
The determinants of vessel capacity utilization: The case of Brazilian iron ore exports
Transportation Research Part A xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Transportation Research Part A journal homepage: www.elsevier.com/locate/tra
The determinants of vessel capacity utilization: The case of Brazilian iron ore exports Roar Adland a,b,⇑, Haiying Jia b, Siri P. Strandenes a a b
Norwegian School of Economics (NHH), Helleveien 30, 5045 Bergen, Norway Center for Applied Research (SNF), Helleveien 30, 5045 Bergen, Norway
a r t i c l e
i n f o
Article history: Received 30 December 2015 Received in revised form 11 November 2016 Accepted 30 November 2016 Available online xxxx Keywords: Bulk shipping Capacity utilization Trade flows
a b s t r a c t The objective of this paper is to determine the drivers behind the utilization of a vessel’s cargo-carrying capacity on individual voyages. Based on maritime economic theory we propose that a vessel’s capacity utilization - defined as the ratio of cargo size divided by DWT - should be positively correlated with freight rates, as poor market conditions will force vessels to compete for lower-than-optimal stem sizes. Furthermore, we propose that capacity utilization is dependent on the distance sailed, the fuel price and the value of the cargo. Using a unique data set sourced from port agent lineup reports and covering nearly 10,000 individual shipments of iron ore from Brazil between 2009 and 2014 we estimate a multiple regression model consisting of macroeconomic, route-specific and vessel-specific determinants. Our empirical results suggest that vessel-specific determinants (DWT) dominate the impact of general market conditions, with smaller vessels typically having lower capacity utilization. The impact of freight market conditions conforms to our a priori expectations. Our findings and modeling approach contributes to maritime environmental policymaking by enabling more accurate bottom-up estimation of emissions. The research is also crucial for improved modeling of real vessel earnings and tonne-mile demand based on observations of global ship movements from AIS data. Ó 2016 Elsevier Ltd. All rights reserved.
1. Introduction Vessel capacity utilization refers to the share of a vessel’s total carrying capacity occupied by paying cargo (Alizadeh and Talley, 2011a). Capacity utilization is a general concept that applies to all segments of shipping. However, the way we measure cargo carrying capacity is defined by the type of ship and may refer to, for instance, tonnes (bulk carriers), cubic meters (gas carriers), TEU (container vessels) or lane meters (RoRo vessels). Typically, the intake of light and voluminous cargoes such as containers is constrained by volume, and so capacity utilization refers to the fraction of total volume available for loading, while the intake of heavy cargoes such as coal and iron ore is constrained by weight relative to a vessel’s allowed Deadweight Tonnage (DWT). A ship’s DWT is typically quoted as being the maximum carrying capacity applicable under the International Load Line Regulations when floating at the Summer Load Line draught. To complicate matters, the actual density of certain cargoes (e.g. light oil products or light grains) will decide if the cargo intake on a particular trip is constrained by the available volume or weight, even for tankers and bulk carriers. In the case of container vessels, actual capacity utilization may also be constrained by vessel stability calculations (i.e. the stacking sequence and resulting centre of gravity) ⇑ Corresponding author at: Norwegian School of Economics (NHH), Helleveien 30, 5045 Bergen, Norway. E-mail address: [email protected] (R. Adland). http://dx.doi.org/10.1016/j.tra.2016.11.023 0965-8564/Ó 2016 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
2
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
(Zerby and Conlon, 1978). To avoid some of these challenges, the theoretical exposition in our paper relate to the transport of high-density cargoes on bulk carriers, where capacity utilization can be thought of as a continuous function and the cargo intake is always weight constrained. In this case, the definition of capacity utilization is simply cargo size divided by DWT, a ratio which is sometimes also referred to as the load factor (Wijnolst and Wergeland, 1996, p. 312). We emphasize that a vessel cannot be loaded to 100% DWT utilization, as the deadweight measure also includes the weight of fuel, freshwater, supplies and crew that the vessel is allowed to carry for a particular draught. Capacity utilization is a key input in both macroeconomic models of the freight market and microeconomic models of vessel/firm profitability. At the macro level, overall fleet productivity is determined by capacity utilization, sailing speed and ballast ratio (see e.g. Wijnolst and Wergeland, 1996). The higher the capacity utilization, the more transportation work a given fleet will produce per time unit. Conversely, low average capacity utilization implies there is slack or spare capacity available on the supply side. At the micro level – be it an individual voyage or an individual company – capacity utilization will have a large impact on the profitability and unit transport cost. This is because most costs can be taken as fixed once a voyage has been accepted or, in the case of liner shipping, a string of vessels have been employed with a defined route and service frequency. To illustrate the importance of capacity utilization for the economics of bulk carriers, Fig. 1 shows the relationship between capacity utilization and timecharter-equivalent (TCE) vessel earnings and transportation cost, respectively. For the purpose of illustration we here use the Capesize voyage from Rotterdam via Tubarao to Qingdao, with fuel and freight market prices as of 26th June 2015 and Baltic Exchange (2015) route assumptions. The unit transport cost refers to the $/tonne breakeven rate when chartering the vessel at the prevailing tripcharter ($/day) rate. For the sake of simplicity we have here ignored the second-order effect caused by the dependency of fuel consumption on cargo size, i.e. the impact of a higher draught on fuel consumption. From Fig. 1 it is obvious that sailing with very low capacity utilization is detrimental to profitability. Indeed, there will be a point of utilization below which accepting the cargo becomes uneconomic – in this example below 45%. From a policy point of view, capacity utilization is a key input in the bottom-up estimation of air pollution from ships. The reason is twofold: (i) it directly affects the fuel consumption and air emissions as the main determinant of the draught (displacement) of a vessel, and (ii) it affects the level of emissions per tonne mile, both as a measure of the amount of cargo onboard and a determinant of the fuel consumption. In order to get a true picture of energy use and related emissions per tonne-mile, particularly in the comparison between land and sea-based transportation modes, it is crucial to use the true cargo carried rather than DWT capacity (Hjelle, 2011). In this context, low capacity utilization would jeopardize the comparative advantage of maritime transport alternatives, and it is therefore a crucial variable to monitor and understand for policy makers. As a related point, poor capacity utilization would question the economic and environmental sustainability of the mega-ships currently in operation and under construction in the container liner industry. In general, lower capacity utilization will imply higher emissions per unit of transportation work, all else equal. As a related policy issue, capacity utilization is also closely tied to the cost of maritime transportation in international trade. For instance, when raw materials are carried on vessels hired on a timecharter ($/day) basis, the real unit cost per tonne of cargo is inversely related to utilization. The use of part loading (low utilization) will, all else equal, therefore lead to higher transport costs and lower international maritime trade (Korinek and Sourdin, 2010; Valentine et al., 2013). More generally, higher capacity utilization of all vessels will increase the available supply of transportation by the existing fleet and reduce transports costs (freight rates) by exploiting economies of scale in general (see, e.g. Hummels, 2007, for a related discussion). Despite the importance of capacity utilization at the micro level in regards to environmental policy, the true cost of international trade and economic sustainability, there are no published empirical studies on its determinants and dynamics. This is likely due to the difficulty in obtaining reliable voyage-specific data on cargo sizes. Our paper is a first attempt at filling this gap in the literature. We contribute to the literature by providing the first theoretical exposition and empirical analysis of the micro- and macroeconomic determinants of capacity utilization in bulk shipping markets. Additionally we investigate this important issue by using an entirely new set of data on cargo sizes from port agents. The remainder of the paper is organised as follows: Section 2 reviews the relevant literature on capacity utilization, Section 3 presents the theoretical basis for our model and variable selection, Section 4 presents our data and descriptive statistics, Section 5 presents our empirical results and Section 6 concludes with suggestions for future research.
2. Literature review While it is clear from the introduction that capacity utilization is a key driver of economic and environmental performance in the maritime supply chain, we do not yet have a good understanding of its determinants. Although capacity utilization is an input in most standard supply/demand models and emissions accounting models, it is often assumed to be of a certain functional form based on industry ‘‘rules of thumb”. Wijnolst and Wergeland (1996, p. 314) states that ‘‘the load factor can theoretically become as high as 0.975, will in practice hardly ever be higher than 0.95, but may come as low as 0.65 if demand is very low and part loading and multi porting are common phenomena”. Smith et al. (2013) suggest that when the freight market is in a state of overcapacity, operators are forced to compete harder for cargoes, resulting in a willingness to accept a greater number of part-load cargoes. Both sources thus point to a positive relationship between freight market conditions and capacity utilization. It is important in this conPlease cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
3
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
$/tonne
$/day 30000
30
25000
25
20000
20
15000
15
10000
10
5000 0
TCE (l.h.s.)
5
Breakeven (r.h.s.) 0 45% 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 100%
Capacity ulizaon Fig. 1. Profitability and breakeven rates as a function of vessel capacity utilization.
text to specify the time horizon on which the capacity utilization for an individual voyage is set. In microeconomic analysis of freight rate formation for individual voyages (see e.g. Alizadeh and Talley, 2011a,b; Adland et al., 2016) we are concerned with the immediate equilibrium between the number of vessels and cargoes in a particular loading area. Cargoes must be shipped within a certain time window (laycan), and only vessels that are commercially available and physically able to meet the laycan can offer transportation. Depending on whether there are more ships than cargoes or vice versa, cargoes are then matched with ships (subject to capacity constraints) in either a perfectly competitive micro-market or an auction-like process, respectively (Stopford, 2009). If we ignore speed changes and the possibility of inter-temporal substitution (Zannetos, 1966), it follows that available supply (DWT capacity) is fixed in the short run (in the order of days). Consequently, an increase in short-run demand and, thus, a higher freight rate, will lead to higher contemporaneous capacity utilization. In the longer run (weeks), high freight rates in one region of the world would attract more vessels which could push freight rates and utilization down. On an even longer time horizon (years), high average freight rates would encourage the construction of more vessels, which would also push freight rates and capacity utilization down. We also note that in oligopolistic shipping sectors, where owners can influence cargo volumes by price differentiation, we may observe a negative shortrun relationship between freight rates and volumes transported by a specific carrier1 (see, Zerby and Conlon, 1978, for a discussion of capacity utilization in the case of price differentiation). In early work, Kalindaga (1990) estimates the impact of part-cargo loading based on a sample of tanker fixtures for each of the main loading areas. Similarly, using fixture data for crude oil tankers between January 2006 and March 2009, Alizadeh and Talley (2011a,b) find that reported utilization is higher for the larger vessels (average 90.8% for VLCCs compared to 77% for Aframax tankers). This negative relationship between vessel size and utilization is expected. Given the economies of scale in ship size (Stopford, 2009), which also applies to fuel consumption and crew size, the share of DWT reserved for ‘non-cargo’ use is necessarily smaller for larger ships. It follows that the maximum utilization is negatively and non-linearly related to vessel size (DWT). We note that only a fraction of all voyages are observable in public fixture data, as some will be performed by in-house tonnage and others are purposely kept ‘‘off market” by the parties involved. Veenstra and Dalen (2008) provide a thorough review of the related problem of constructing freight rate indices based on incomplete public fixture data. In general, the type of charterparty will influence whether the charterer is incentivised to optimize cargo intake (i.e. maximizing capacity utilization). Under a time charter, the charterer is in commercial control of the vessel and pays a fixed daily hire irrespective of the cargo size (Stopford, 2009). Accordingly, the charterer will achieve the minimum unit transport cost by maximising capacity utilization. Conversely, under a voyage charter, the charterer pays a fixed $/tonne irrespective of the cargo size. For the shipowner the situation is the reverse, with no economic gain from an increased cargo intake under a time (trip) charter, and a proportional improvement in earnings as a function of the cargo size under the voyage charter. One solution to approximating capacity utilization in the absence of cargo size data is to use vessel draughts for individual voyages reported by the Automated Identification System (AIS). Jia et al. (2015) compare different models for the conversion between vessel draughts and cargo sizes in the drybulk market and conclude that they can give good approximations of the onboard cargo size. However, they note that this approach relies heavily on the geographical coverage of the data and the accuracy of the manual draught reports by the ship’s crew. Smith et al. (2013) use AIS-reported vessel draughts as a basis for estimating capacity utilization across shipping segments and find it to be approximately 80% or less in the case of crude oil tankers, but does not investigate its determinants. The literature on activity-based energy use and vessel emissions similarly rely heavily on rules of thumb and the use of average, static, capacity utilization. Endresen et al. (2007) applies a flat 80% DWT utilization for the reconstruction of historical emission levels over four decades, but note that this variable represents the greatest uncertainty in the estimates of fleet productivity. Behrens et al. (2003) report cargo utilization of 91% for bulk carriers and 87% for tankers above 50,000 DWT for the year 2001. 1
We thank an anonymous referee for pointing out the nuances with regards to time horizon and market structure.
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
4
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
In summary, the literature on capacity utilization lacks theoretical structure and has not yet attempted to model the determinants or dynamics of capacity utilization at the micro (individual voyage) level. Our paper fills this gap in the literature. 3. Theoretical background and methodology 3.1. Deriving optimal capacity utilization As a starting point for our capacity utilization model, it is useful to start with the classical profit maximization problem for a shipowner (see, for instance, Ronen, 1982; Strandenes, 1999). However, rather than assuming that the daily fuel consumption is a function purely of vessel speed we adopt the more general approach that fuel consumption is determined by both speed and capacity utilization. If we ignore port days and port costs, the operating profit function can be expressed as:
p ¼ s DWT u
D 24v
p f ðv ; uÞ
ð1Þ
where s is the spot rate ($/tonne); DWT is the ship’s deadweight tonnage (tonnes); u is the utilization rate; D is the trip distance (nautical miles); v is the average voyage sailing speed (knots); p is the bunker price ($/tonne); f(v, u) is daily fuel consumption at speed v and utilization rate of u (tonnes/day). The general point to make here is that there exists a trade off between the marginal revenue of one extra tonne of cargo (the spot freight rate) and the marginal cost of an extra tonne of cargo (i.e. the incremental fuel consumption due to higher displacement). The optimal utilization rate u⁄ is, thus, the ratio where these two effects are identical. The daily fuel consumption function f(v, u) can be expressed as a function of displacement and speed (MAN Diesel and Turbo, 2013; Gorski et al., 2013):
f ðv ; uÞ ¼
v vd
3 2=3 r Fd
ð2Þ
rd
where vd is the design (maximum) sailing speed (knots); r is the voyage displacement (tonnes); rd is the full load displacement (tonnes); Fd is the daily fuel consumption at maximum speed and displacement (tonnes/day). The displacement of a vessel is the sum of its lightweight (i.e. the total weight of steel, engine, equipment etc.) and its deadweight. As a first order approximation, the displacement ratio rr can be replaced by the utilization rate (u) provided d
that the vessel’s lightweight (LDT) is much lower than its DWT.2 In this case, Eq. (2) can be rewritten as
p ¼ s DWT u 24v =D p
v vd
3 F d f ðuÞ2=3
ð3Þ
The utilization rate that maximizes the profit p for given values of the spot freight rate and bunkers price can be found by taking the partial derivative of p with respect to u and solving the equation by setting it to zero, i.e. @@up ¼ 0 subject to umin 6 u 6 umax . The solution for the optimal utilization ratio is
u ¼
!3 2ðvv Þ3 F d p D d
3s DWT 24v
ð4Þ
Interestingly, the optimal utilization ratio is related to the ratio of fuel prices and spot rates, as is the optimal speed of a vessel (Ronen, 1982). We emphasize that it is not the solution to Eq. (4) per se that is interesting in our context, as the optimal utilization will generally be the corner solution u⁄ = umax (i.e. full load). It is only in extreme cases of very low spot rates and very high bunker prices that the theoretical optimal load may be less than full load.3 However, Eq. (4) is useful as it gives us a first indication of the variables that play a role. 2
MAN Diesel and Turbo (2013) suggests an average DWT/LDT ratio of 6 for tankers and bulk carriers. Strictly speaking there exists a lower bound to the spot freight rate which is contingent on the bunkers price (Adland, 2012) and so some combinations of s and b are not plausible. 3
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
5
3.2. Building an empirical model With reference to Eq. (4) we can divide variables into three groups: macroeconomic (spot freight rate s, fuel prices p), voyage-specific (speed v and distance D) and vessel-specific variables (DWT). We expand on their expected influence on capacity utilization below. We also add two new variables - a Southern hemisphere winter dummy and the iron ore spot price - to control for the potential effects of seasonal weather restrictions and the impact of commodity markets. 3.2.1. Macroeconomic variables Spot freight rate. Following Wijnolst and Wergeland (1996) and Smith et al. (2013) we expect that freight market conditions, as proxied here by the $/tonne spot freight rate for a voyage charter, are positively correlated with capacity utilization. During a weak freight market, with a corresponding oversupply of vessels relative to the volume of available cargoes, it will be rational for operators to compete for low utilization (part cargoes) as long as the resulting TCE is positive and there is no alternative employment. Conversely, in strong freight markets where transport capacity is scarce, charterers have a strong incentive to maximize the cargo intake for every voyage they undertake, subject to technical constraints. Bunker price. High fuel prices should encourage reduced sailing speeds and daily fuel consumption. Due to the non-linear relationship between sailing speed and fuel consumption in Eq. (2), this will translate into a lower required volume of fuel onboard for a given voyage.4 Provided that the cargo size is maximized, the lower intake of fuel for the voyage will increase utilization, all else equal. Accordingly, we expect a negative relationship between fuel prices and DWT utilization. Iron ore price. China imports 50% of Brazilian iron ore (UN Comtrade, 2016). Lower global prices makes imported iron ore more competitive compared to high-cost domestic producers and this substitution effect will increase seaborne demand, all else equal. We therefore expect a negative relationship between the iron ore price and capacity utilization. 3.2.2. Voyage-specific variables Distance. The relationship between distance and utilization is not clear cut. In principle, longer distances require higher bunker intake at the start of a voyage and, correspondingly, lower capacity utilization as long as the cargo intake is maximized. However, long voyages also have a higher opportunity cost of part loads (i.e. lower earnings and higher unit transport cost for longer) which should encourage higher utilization by economically rational agents. Moreover, long voyages in the drybulk market will typically serve the modern deep-water ports in North-East Asia which have fewer draught restrictions than regional ports closer to the main export areas (Brazil in our empirical case), which again points to higher utilization. Winter dummy. Voyages that pass through ‘winter zones’ as per the International Load Lines regulations will have reduced maximum draughts, and therefore lower utilization all else equal. In our empirical case, as 80% of Brazilian iron ore is passing the Cape of Good Hope eastbound to the Middle East and Asia (UN Comtrade, 2016), we would anticipate lower draughts in the Southern Hemisphere winter. Our hypothesis thus becomes that voyages commencing in Q4/Q1 have lower average utilization. Speed. We do not include speed in our study due to the unavailability of speed data for individual voyages for our entire sample. Adland and Jia (2016) study voyage speeds for bulk carriers in the 2011–2012 period and conclude that speeds are relatively constant throughout this period, with no statistical significant impact of freight market conditions or fuel prices for the laden leg. We are therefore fairly confident that omitting the voyage speed variable does not negatively affect our results. 3.2.3. Ship-specific variables DWT. We expect a negative and non-linear relationship between vessel size and capacity utilization as the non-cargo share of DWT (fuel, fresh water and fixed weights) is smaller for large ships due to economies of scale in ship operation. However, aside from this technicality, commercial practice and physical port constraints may also affect utilization as a function of vessel size. For instance, draught restrictions in the load or discharge port may limit the cargo size for very large vessels and reduce capacity utilization. Similarly, the availability of storage areas or the commercial needs of the buyer may limit cargo sizes. Finally, there may be hold strength limitations, particularly for heavy cargo such as iron ore. We note that in some cases, the shipowner may be compensated for such ‘short loading’ by way of a higher freight rate. We here test different specifications with regards to vessel size: DWT, DWT2 and dummies for DWT size groups. 3.3. Regression model To test the importance of the above determinants of the utilization rate empirically, we estimate various specifications of the following general model using a pooled OLS regression:
4 By way of example, a standard 175,000DWT Capesize sailing a 10,000 n.m. laden voyage will consume a total of 1555 tonnes of fuel at an average speed of 15 knots and only 691 tonnes at 10 knots. This releases an additional 864 tonnes of cargo capacity, equivalent to 0.5%-points utilization.
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
6
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
U j ¼ a0 þ a1 sj þ a2 pj þ a3 qj þ a4 Dj þ a5 Wdj þ
X
xi DWT i;j þ ej
ð5Þ
i
where Uj is the observed utilization rate for voyage j; sj is the spot freight rate; pj is the bunker price; qj is the iron ore spot price; Dj is the trip distance; Wdj is the winter dummy variable; and DWTi,j is the set of i variables defining the relationship between size and utilization. ej is a random perturbation such that E(ej) = 0 and Var(ej) = r2. Pooled OLS is appropriate in this context because there are, on average, few observations along the individual vessel dimension in our panel data and so we care less about time-invariant effects. As each freight rate, fuel price and iron ore price is matched to individual voyages which need not be consecutive in the time dimension we also do not worry about the serial correlation that normally would be present in time series of these macro variables. We apply robust regression estimation using Huber-White sandwich estimators to control for heteroscedasticity, lack of normality, and outlier observations that exhibit large residuals, leverage or influence (Huber, 1967; White, 1980). We also check for multicollinearity using the Variance Inflation Factor (VIF) diagnostic. An important point is whether the reduced-form model in Eq. (5) could suffer from an endogeneity problem. Fortunately, it is straightforward to establish that our independent variables are not caused by DWT utilization in the short run. The bunker price is a derivative of the global crude oil market, of which marine fuel consumption is a tiny share. As DWT utilization will vary within a relatively small range and the speed and fleet size can be taken as constant in the short run, it follows that the impact of draught variations on the global demand for oil is negligible. The demand for commodities is generally taken to be inelastic with regards to freight market conditions (Adland and Strandenes, 2007). Provided that ocean transportation is not a bottleneck that affects the supply of the commodity, the commodity price is therefore not caused by freight market conditions in general or DWT utilization in particular. Voyage distances are given by the location of sources and consumers of the raw material in question. In our empirical case, neither the location or production of iron ore mines or the location or production of steel mills can reasonably be assumed to depend on vessel capacity utilization. The short-run distribution of vessel sizes (DWT) is given by the geographical positions and commercial availability of vessels and therefore constant in the very short run. Additionally, iron ore cargoes are sold almost exclusively on a Cost Insurance Freight (CIF) basis, which means it is the seller of the cargo (in our empirical case, Brazilian miner Vale) that must arrange ocean transportation. The cargo shipment programme is a mix of long-term volume contracts and spot cargoes, but all cargoes will be sold in the iron ore market some time before the freight is arranged. It follows that the cargo size, or cargo size range, is defined before a new cargo is offered in the freight market. The momentary demand for transportation is therefore given by the number of cargoes and their destination, with the spot freight rate a result of the matching process with the available fleet. Consequently, there is no causality from DWT utilization to the spot freight rate in the very short run. 4. Data 4.1. Cargo size data We collected monthly reports of vessels leaving the six main iron ore export terminals in Brazil from port agents LBH Group over the period November 1, 2008 to August 1, 2014. Table 1 summarizes the physical characteristics of the six terminals. The port agent reports contain the following data fields:
Report date Vessel name DWT Arrival time, ETA (Expected Time of Arrival) Berthing time, ETB (Expected Time of Berthing) Departure time, ETD (Expected Time of Departure) Destination Cargo quantity.
The sample consists of 9862 individual port calls by 3954 vessels ranging from 24,000 DWT to 405,000 DWT with an average size of 152,000 DWT. Table 2 shows the descriptive statistics for cargo size by port. Sailing distances for the laden leg are obtained from a standard marine distance calculator (axsmarine.com) under the assumption that vessels avoid pirate zones where applicable. Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
7
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx Table 1 Brazil iron ore terminal details. Source: Compiled by authors from vale.com, samarco.com. Terminal name
Pier
Shipper
No. Berths
Latitude
Longtitude
Max DWT
Max LOA
Max draught
Load rate (MTPH)
Ponta da Maderia
Pier I Pier II Pier III
Vale
2
2.57
44.38
420,000 155,000 220,000
345/280
25
16,000 8000 8000
Ponta Ubu
Westside Pier Eastside Pier
Samarco
2
20.78
40.58
250,000 150,000
308 240
15.5 16.8
10,000 7000
CSN Terminal
Pier 401
Ferteco
2
22.87
43.77
n/a
17.3
5000
Tubarao
Pier 1 Pier 2
Vale Vale
2 2
20.28 20.28
40.25 40.25
200,000 365,000
285/320 350
17 20
7000 16,000
Vale Vale
1 1
22.87 23.00
43.77 44.02
230,000 300,000
Itaguai GIT
18 15.2
Table 2 Descriptive statistics for cargo size by port (tonnes). Source: LBH Group, compiled by authors.
CSN GIT Itaguai PDM Tubarao Ubu
Obs
Mean
Std. Dev.
Min
Max
965 1114 834 2745 3165 992
156,482 186,669 149,624 197,710 169,826 119,638
19,772 55,285 34,450 73,401 81,096 54,189
39,285 29,220 27,500 20,052 8000 29,998
198,236 285,902 274,380 395,384 395,373 204,350
4.2. Macroeconomic data As China is by far the largest importer of Brazilian iron ore, we use the Baltic Capesize Index C3 for the route Tubarão to Qingdao as reference for the freight market. As per Baltic Exchange (2015) standard route assumptions, vessels are delivered for the trip in the Antwerp-Rotterdam-Amsterdam (ARA) area and then sail in ballast to Tubarão for loading. Assuming that vessels are fixed upon departing the last discharge port, we match each laden voyage with the Baltic spot freight rate 21 days prior to the loading port ETA5 in order to reflect freight market conditions at the time of the fixture. We use the 380cst Rotterdam heavy fuel price at the same time as a reference for fuel costs. Both the spot freight rate and fuel price series were obtained from Clarkson Research (2016). We base our proxy for the value of the cargo on the 62%Fe-content spot iron ore index as constructed by The Steel Index (TSI) and obtained from Bloomberg. The TSI 62% spot iron ore price index is defined as the price per tonne of iron ore delivered in Tianjin, China, on a Cost, Insurance and Freight (CIF) basis. In order to avoid accounting for the freight component twice in our model, we calculate the Free On Board (FOB) price by subtracting the above freight rate from the CIF price and using the FOB price in our estimations. Similarly to the freight rate, we match each cargo with the iron ore price 21 days prior to the time of arrival at the load port anchorage in Brazil in order to reflect market conditions at the time of sale.6 Fig. 2 shows the development in our macro variables during the sample period. 4.3. Capacity utilization Fig. 3 shows the kernel density of the utilization, using bandwidth 0.725% and the Gaussian kernel function. We see that the majority of voyages are performed with fairly high capacity utilization (>85%), with a peak around 96%. Indeed, the average capacity utilization across the sample is 92%. The observations with very low capacity utilization may be either a result of vessels loading at multiple ports in Brazil, making it appear as if the cargo intake at each port is unusually low, or errors in the data entry process at the port agent. For instance, the cargo size or DWT may be wrongly entered, or a vessel name may not be unique to a certain size vessel. Fig. 4 shows a scatter plot of vessel capacity utilization vs. DWT. We see a clustering of vessel sizes into roughly four main DWT groups: Panamax and smaller (<100,000 DWT), Capesizes (<250,000 DWT), iron ore carriers and Valemaxes (>400,000 DWT). We note the presence of some ‘‘mini-Capes” around 115,000 DWT size but choose to include these with their bigger Capesize cousins. Table 3 summarizes the utilization statistics by size group. Both Fig. 4 and Table 3 confirm that smaller vessels (Panamax and smaller) have substantially lower utilization on average. 5 This includes 17 days sailing in ballast from Rotterdam plus an assumed average of 4 days between fixture and delivery as per Baltic Exchange (2015) definitions. 6 The sale of the cargo will probably have taken place even earlier than the fixture of the ship though this would be unavailable private information.
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
8
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
$/tonne
$/tonne 800
200 180
700
160 600 140 500
120
400
100 80
300
60 200 40 100
20 0 21/10/08 21/10/09 21/10/10 21/10/11 21/10/12 21/10/13 21/10/14 iron ore (LH)
freight (LH)
0
bunker price (RH)
10 0
5
Density
15
20
Fig. 2. Freight rate, bunker price and iron ore price development. Source: Clarkson Research (2016), Bloomberg.
.2
.4
.6
.8
1
utilization Fig. 3. Kernel density of vessel utilization.
Fig. 4. Scatter plot of utilization vs. DWT.
We note that there is some non-linearity in the mean-utilization-DWT relationship, with particularly the sub-Capsize segment but also the very large vessels showing lower utilization than the standard Capesize segment (around 180,000 DWT).
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
9
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx Table 3 Descriptive statistics for capacity utilization by size group and port. DWT range
Obs
Mean
Std. Dev.
Min
Max
0–99,999 100,000–249,999 250,000–399,999 400,000+
1589 6405 1652 169
0.801 0.944 0.936 0.934
0.175 0.069 0.074 0.079
0.285 0.153 0.027 0.650
1.00 1.00 1.00 0.99
Port name
Obs
Mean
Std. Dev.
Min
Max
CSN GIT Itaguai PDM Tubarao Ubu
965 1114 834 2745 3165 992
0.916 0.954 0.924 0.945 0.904 0.860
0.081 0.058 0.075 0.080 0.131 0.149
0.230 0.388 0.153 0.140 0.027 0.250
1.00 1.00 1.00 1.00 1.00 1.00
Table 4 Regression results. Dependent variable: utilization % (1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
Adj. R-squared VIF
0.104
0.042
0.232
0.236 1.39
0.242 1.49
0.238 1.380
0.241 1.530
0.256 1.54
0.261 1.44
0.264 1.41
Constant
0.84 (201.78) [0.00]
0.89 (399.22) [0.00]
0.80 (182.94) [0.00]
0.77 (132.30) [0.00]
0.83 (112.51) [0.00]
0.80 (108.30) [0.00]
0.85 (154.41) [0.00]
0.77 (175.89) [0.00]
0.77 (104.92) [0.00]
0.79 (109.86) [0.00]
DWT (tonne) (’000000)
0.46 (23.69) [0.00]
0.14 (32.17) [0.00]
0.14 (32.47) [0.00]
0.14 (32.43) [0.00]
0.14 (32.26) [0.00]
0.14 (32.31) [0.00]
0.12 (23.61) [0.00]
0.12 (23.81) [0.00]
0.12 (24.01) [0.00]
DWT_3 (tonne) 250,000–399,999t
0.14 (28.56) [0.00]
0.14 (28.91) [0.00]
0.14 (29.17) [0.00]
0.14 (28.89) [0.00]
0.14 (29.05) [0.00]
0.11 (19.80) [0.00]
0.11 (20.16) [0.00]
0.11 (20.41) [0.00]
DWT_4 (tonne) 400,000+t
0.13 (17.91) [0.00]
0.14 (18.40) [0.00]
0.14 (18.79) [0.00]
0.14 (18.27) [0.00]
0.14 (18.57) [0.00]
0.11 (13.78) [0.00]
0.11 (14.16) [0.00]
0.12 (14.68) [0.00]
1.08 (7.51) [0.00]
0.69 (4.54) [0.00]
0.75 (4.88) [0.00]
0.73 (4.92) [0.00]
0.68 (4.85) [0.00]
DWT^2 (tonne) (’E+12)
0.74 (18.44) [0.00]
DWT_group_dummy (<100,000t benchmark) DWT_2 (tonne) 100,000–249,999t
Freight rate (C3 route $/tonne) (’000) Bunker price (380cst Rotterdam $/tonne) (’000)
[0.08] (7.49) [0.00]
Iron ore price (62% FOB $/tonne) (’000)
0.03 (0.66) [0.51]
[0.08] (7.67) [0.00] 0.17 (6.08) [0.00]
[0.07] (9.50) [0.00]
0.07 (1.90) [0.06]
0.14 (4.96) [0.00]
Distance (nautical mile) (’000)
5.97E03 (13.46) [0.00]
5.14E04 (13.03) [0.00]
5.71E03 (12.81) [0.00]
Winter_dummy
3.51E04 (0.18) [0.85]
5.14E04 (0.27) [0.79]
1.20E03 (0.63) [0.53]
Note: Figures in () are t-stats; figures in [ ] are the p- values of H0: zero coefficients.
5. Empirical results Table 4 shows the results for the various specifications of our multiple regression model. We note that the high statistical significance and estimated signs of all parameters are stable across specifications, which is encouraging for the validity of our
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
10
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
hypotheses and correct specification of our empirical model. Moreover, the VIF diagnostic test suggests the absence of multicollinearity in all cases. Of the three different relationships between vessel size (DWT) and capacity utilization (specifications 1–3), the size group dummies perform the best. Our results are similar to those found for tankers in Smith et al. (2013) in the sense that smaller vessels operate with lower average capacity utilization. However, the relationship is non-linear, with vessel sizes above 100,000 DWT showing only marginal changes. For the remainder of the specifications we use only the DWT group dummies as they have the highest explanatory power. In specification 4 we add the impact of the freight market. As expected, the spot freight rate loads positively and is highly statistically significant. From an economic point of view, a change in the spot freight rate from $30/tonne observed at recent peaks down to the $10/tonne at the cycle lows implies a reduction in the vessel capacity utilization of 2.2%-points, all else being equal. Given that the ship-specific variation in capacity utilization is relatively low (ref. the peaked distribution in Fig. 3), this is economically meaningful. When we add the fuel price and/or commodity price as explanatory variables (specifications 5 and 6), this effect is reduced somewhat. The bunker fuel price is also a highly significant determinant, and the iron ore price less so, though both conform to our a priori expectations. However, we note that the fuel price and commodity price appears to be picking up the same ‘commodity-boom’ effect, as inclusion of the former in the regression (specifications 5 and 7) makes the latter insignificant at standard levels of significance. The economic intuition for including the fuel price as a determinant, aside from it showing up in Eq. (4), is arguably less appealing than our other macro variables, particularly with low variation in voyage speeds. The distance variable is also highly significant and loads positively, which suggests that the alternative cost effect and/or port infrastructure constraints dominate the effect on fuel intake. In economic terms, the increase in utilization between a short and long voyage, say Rotterdam (4965 nautical miles) vs Qingdao (11,017 nautical miles), is approximately 3.6%points. Seasonality, at least how we have defined it here, does not seem to make a difference to vessel capacity utilization. 6. Concluding remarks We have shown that several macro- and microeconomic determinants, notably freight market conditions, fuel prices, vessel size, sailing distance and vessel size, influence the vessel capacity utilization of drybulk carriers. From a maritime economic perspective this is an important finding, as it means that real vessel earnings receive a ‘double blow’ during poor freight markets from both low spot rates and reduced capacity utilization. The latter effect results both from increased competition for sub-optimal cargo sizes and, likely, more efficient sailing patterns in times of overcapacity. In terms of maritime environmental policy, our results suggest that the emissions per tonne-mile of cargo transported are greater during low freight markets, due to the lower estimated utilization. Consequently, the typical assumption of static capacity utilization in the literature will lead to biases in estimates that are correlated with the freight market cycle. Our study is a first step to improve the accuracy and quality of national and global emission accounting for policy making along this important dimension. We acknowledge that there remains a large unexplained component in the observed vessel capacity utilization data. While some of this is likely due to measurement errors in the original dataset, future revisions of this work should try to further improve the explanatory power of the models herein. In particular, vessel age and hold structure should be considered. Older tonnage will expectedly have lower capacity utilization – both because it is less attractive on the spot market and therefore may have to bid for sub-optimal cargo sizes but also because of reduced structural strength. This is particularly a concern for heavy cargo such as iron ore, where the fatigue caused by numerous loading and discharge cycles over the life of the vessel can eventually cause a collapse of hull sections (Paik and Melchers, 2008). Similarly, specialised iron ore carriers and other bulk carrier vessels denoted as ‘‘strengthened for ore” will expectedly have a higher capacity utilization than vessels not built to this higher specification. This is in part to avoid failures in the ship structure, particularly during the loading process. However, ore carriers also have a different hold shape than standard bulk carriers, ensuring a higher centre of gravity and better sea keeping performance. It is also worth modeling the relationships between maximum vessel draughts and any draught restrictions in the loading or discharge port in detail, as this is known to sometimes be a limiting factor. The challenge here is that the final port of discharge will not always be known. Acknowledgements This research was partly funded by the Research Council of Norway’s SMARTRANS programme under the project CARGOMAP – mapping vessel behaviour and cargo flows (project #239104). We thank two anonymous referees for their valuable comments as well as participants at the International Forum for Shipping, Ports and Airports (IFSPA 2015) at the Hong Kong Polytechnic University for input on an earlier version of this paper. References Adland, R., 2012. Microeconomic modelling of the supply function in bulk shipping revisited. In: International Association of Maritime Economists (IAME) Annual Conference, Taipei. September 5–8. Adland, R., Jia, H., 2016. Dynamic speed choice in bulk shipping. Marit. Econ. Logist. http://dx.doi.org/10.1057/s41278-016-0002-3. Adland, R., Strandenes, S.P., 2007. A discrete-time stochastic partial equilibrium model of the spot freight market. J. Transp. Econ. Policy 41 (2), 189–218. Adland, R., Cariou, P., Wolff, F.-C., 2016. The influence of charterers and owners on bulk shipping freight rates. Transp. Res. Part E 86, 69–82.
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
11
Alizadeh, A.H., Talley, W.K., 2011a. Vessel and voyage determinants of tanker freight rates and contract times. Transp. Policy 18 (5), 665–675. Alizadeh, A.H., Talley, W.K., 2011b. Microeconomic determinants of dry bulk shipping freight rates and contract times. Transportation 38 (3), 561–579. Baltic Exchange, 2015. Manual for Panelists,. Behrens, H.L., Endresen, Ø., Mjelde, A., Garmann, C., 2003. Environmental Accounting System for Norwegian Shipping – EASNoS Phase 1, DNV Report 2002– 1645. Det Norske Veritas, Høvik, Norway. Clarkson Research, 2016. Shipping Intelligence Network, . Endresen, Ø., Sørgård, E., Behrens, H.L., Brett, P.O., Isaksen, I.S.A., 2007. A historical reconstruction of ships’ fuel consumption and emissions. J. Geophys. Res. 112, 1–17. Gorski, W., Abramowics-Gerigk, T., Burciu, Z., 2013. The influence of ship operational parameters on fuel consumption. Sci. J. 36 (108), 49–54. Hjelle, H.M., 2011. The double load factor problem of Ro-Ro shipping. Marit. Policy Manage. 38 (3), 235–249. Huber, P.J., 1967. The behavior of maximum likelihood estimates under nonstandard conditions. Proceedings of the Firth Berkeley Symposium on Mathematical Statistics and Probability, 221–233. Hummels, D., 2007. Transport costs and international trade in the second era of globalization. J. Econ. Perspect. 21 (3), 131–154. Jia, H., Prakash, V., Smith, T., 2015. Estimating Vessel Utilization in the Drybulk Freight Market: The Reliability of Draught Reports in AIS Data Feeds. In: ECONSHIP Conference, Chios. 23–26 June. Kalindaga, Y.C., 1990. Estimation of capacity utilization in world shipping. Marit. Policy Manage. 17 (1), 41–46. Korinek, J., Sourdin, P., 2010. Clarifying trade costs: maritime transport and its effect on agricultural trade. Appl. Econ. Perspect. Policy 32 (3), 417–435. MAN Diesel & Turbo, 2013. Basic Principles of Ship Propulsion. Paik, J.K., Melchers, R.E., 2008. Condition Assessment of Aged Structures. Woodhead Publishing, Cambridge. Ronen, D., 1982. The effect of oil price on the optimal speed of ships. J. Operat. Res. Soc. 33 (11), 1035–1040. Smith, T., O’Keefe, E., Aldous, L., Agnolucci, P., 2013. Assessment of Shipping’s Efficiency Using Satellite AIS Data. UCL Energy Institute, London, UK. Stopford, M., 2009. Maritime Economics. Routledge, London. Strandenes, S.P., 1999. Is there a potential for a two-tier tanker market? Marit. Policy Manage. 26 (3), 249–264. UN Comtrade, 2016. UN comtrade database,
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
Contents lists available at ScienceDirect
Transportation Research Part A journal homepage: www.elsevier.com/locate/tra
The determinants of vessel capacity utilization: The case of Brazilian iron ore exports Roar Adland a,b,⇑, Haiying Jia b, Siri P. Strandenes a a b
Norwegian School of Economics (NHH), Helleveien 30, 5045 Bergen, Norway Center for Applied Research (SNF), Helleveien 30, 5045 Bergen, Norway
a r t i c l e
i n f o
Article history: Received 30 December 2015 Received in revised form 11 November 2016 Accepted 30 November 2016 Available online xxxx Keywords: Bulk shipping Capacity utilization Trade flows
a b s t r a c t The objective of this paper is to determine the drivers behind the utilization of a vessel’s cargo-carrying capacity on individual voyages. Based on maritime economic theory we propose that a vessel’s capacity utilization - defined as the ratio of cargo size divided by DWT - should be positively correlated with freight rates, as poor market conditions will force vessels to compete for lower-than-optimal stem sizes. Furthermore, we propose that capacity utilization is dependent on the distance sailed, the fuel price and the value of the cargo. Using a unique data set sourced from port agent lineup reports and covering nearly 10,000 individual shipments of iron ore from Brazil between 2009 and 2014 we estimate a multiple regression model consisting of macroeconomic, route-specific and vessel-specific determinants. Our empirical results suggest that vessel-specific determinants (DWT) dominate the impact of general market conditions, with smaller vessels typically having lower capacity utilization. The impact of freight market conditions conforms to our a priori expectations. Our findings and modeling approach contributes to maritime environmental policymaking by enabling more accurate bottom-up estimation of emissions. The research is also crucial for improved modeling of real vessel earnings and tonne-mile demand based on observations of global ship movements from AIS data. Ó 2016 Elsevier Ltd. All rights reserved.
1. Introduction Vessel capacity utilization refers to the share of a vessel’s total carrying capacity occupied by paying cargo (Alizadeh and Talley, 2011a). Capacity utilization is a general concept that applies to all segments of shipping. However, the way we measure cargo carrying capacity is defined by the type of ship and may refer to, for instance, tonnes (bulk carriers), cubic meters (gas carriers), TEU (container vessels) or lane meters (RoRo vessels). Typically, the intake of light and voluminous cargoes such as containers is constrained by volume, and so capacity utilization refers to the fraction of total volume available for loading, while the intake of heavy cargoes such as coal and iron ore is constrained by weight relative to a vessel’s allowed Deadweight Tonnage (DWT). A ship’s DWT is typically quoted as being the maximum carrying capacity applicable under the International Load Line Regulations when floating at the Summer Load Line draught. To complicate matters, the actual density of certain cargoes (e.g. light oil products or light grains) will decide if the cargo intake on a particular trip is constrained by the available volume or weight, even for tankers and bulk carriers. In the case of container vessels, actual capacity utilization may also be constrained by vessel stability calculations (i.e. the stacking sequence and resulting centre of gravity) ⇑ Corresponding author at: Norwegian School of Economics (NHH), Helleveien 30, 5045 Bergen, Norway. E-mail address: [email protected] (R. Adland). http://dx.doi.org/10.1016/j.tra.2016.11.023 0965-8564/Ó 2016 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
2
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
(Zerby and Conlon, 1978). To avoid some of these challenges, the theoretical exposition in our paper relate to the transport of high-density cargoes on bulk carriers, where capacity utilization can be thought of as a continuous function and the cargo intake is always weight constrained. In this case, the definition of capacity utilization is simply cargo size divided by DWT, a ratio which is sometimes also referred to as the load factor (Wijnolst and Wergeland, 1996, p. 312). We emphasize that a vessel cannot be loaded to 100% DWT utilization, as the deadweight measure also includes the weight of fuel, freshwater, supplies and crew that the vessel is allowed to carry for a particular draught. Capacity utilization is a key input in both macroeconomic models of the freight market and microeconomic models of vessel/firm profitability. At the macro level, overall fleet productivity is determined by capacity utilization, sailing speed and ballast ratio (see e.g. Wijnolst and Wergeland, 1996). The higher the capacity utilization, the more transportation work a given fleet will produce per time unit. Conversely, low average capacity utilization implies there is slack or spare capacity available on the supply side. At the micro level – be it an individual voyage or an individual company – capacity utilization will have a large impact on the profitability and unit transport cost. This is because most costs can be taken as fixed once a voyage has been accepted or, in the case of liner shipping, a string of vessels have been employed with a defined route and service frequency. To illustrate the importance of capacity utilization for the economics of bulk carriers, Fig. 1 shows the relationship between capacity utilization and timecharter-equivalent (TCE) vessel earnings and transportation cost, respectively. For the purpose of illustration we here use the Capesize voyage from Rotterdam via Tubarao to Qingdao, with fuel and freight market prices as of 26th June 2015 and Baltic Exchange (2015) route assumptions. The unit transport cost refers to the $/tonne breakeven rate when chartering the vessel at the prevailing tripcharter ($/day) rate. For the sake of simplicity we have here ignored the second-order effect caused by the dependency of fuel consumption on cargo size, i.e. the impact of a higher draught on fuel consumption. From Fig. 1 it is obvious that sailing with very low capacity utilization is detrimental to profitability. Indeed, there will be a point of utilization below which accepting the cargo becomes uneconomic – in this example below 45%. From a policy point of view, capacity utilization is a key input in the bottom-up estimation of air pollution from ships. The reason is twofold: (i) it directly affects the fuel consumption and air emissions as the main determinant of the draught (displacement) of a vessel, and (ii) it affects the level of emissions per tonne mile, both as a measure of the amount of cargo onboard and a determinant of the fuel consumption. In order to get a true picture of energy use and related emissions per tonne-mile, particularly in the comparison between land and sea-based transportation modes, it is crucial to use the true cargo carried rather than DWT capacity (Hjelle, 2011). In this context, low capacity utilization would jeopardize the comparative advantage of maritime transport alternatives, and it is therefore a crucial variable to monitor and understand for policy makers. As a related point, poor capacity utilization would question the economic and environmental sustainability of the mega-ships currently in operation and under construction in the container liner industry. In general, lower capacity utilization will imply higher emissions per unit of transportation work, all else equal. As a related policy issue, capacity utilization is also closely tied to the cost of maritime transportation in international trade. For instance, when raw materials are carried on vessels hired on a timecharter ($/day) basis, the real unit cost per tonne of cargo is inversely related to utilization. The use of part loading (low utilization) will, all else equal, therefore lead to higher transport costs and lower international maritime trade (Korinek and Sourdin, 2010; Valentine et al., 2013). More generally, higher capacity utilization of all vessels will increase the available supply of transportation by the existing fleet and reduce transports costs (freight rates) by exploiting economies of scale in general (see, e.g. Hummels, 2007, for a related discussion). Despite the importance of capacity utilization at the micro level in regards to environmental policy, the true cost of international trade and economic sustainability, there are no published empirical studies on its determinants and dynamics. This is likely due to the difficulty in obtaining reliable voyage-specific data on cargo sizes. Our paper is a first attempt at filling this gap in the literature. We contribute to the literature by providing the first theoretical exposition and empirical analysis of the micro- and macroeconomic determinants of capacity utilization in bulk shipping markets. Additionally we investigate this important issue by using an entirely new set of data on cargo sizes from port agents. The remainder of the paper is organised as follows: Section 2 reviews the relevant literature on capacity utilization, Section 3 presents the theoretical basis for our model and variable selection, Section 4 presents our data and descriptive statistics, Section 5 presents our empirical results and Section 6 concludes with suggestions for future research.
2. Literature review While it is clear from the introduction that capacity utilization is a key driver of economic and environmental performance in the maritime supply chain, we do not yet have a good understanding of its determinants. Although capacity utilization is an input in most standard supply/demand models and emissions accounting models, it is often assumed to be of a certain functional form based on industry ‘‘rules of thumb”. Wijnolst and Wergeland (1996, p. 314) states that ‘‘the load factor can theoretically become as high as 0.975, will in practice hardly ever be higher than 0.95, but may come as low as 0.65 if demand is very low and part loading and multi porting are common phenomena”. Smith et al. (2013) suggest that when the freight market is in a state of overcapacity, operators are forced to compete harder for cargoes, resulting in a willingness to accept a greater number of part-load cargoes. Both sources thus point to a positive relationship between freight market conditions and capacity utilization. It is important in this conPlease cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
3
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
$/tonne
$/day 30000
30
25000
25
20000
20
15000
15
10000
10
5000 0
TCE (l.h.s.)
5
Breakeven (r.h.s.) 0 45% 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 100%
Capacity ulizaon Fig. 1. Profitability and breakeven rates as a function of vessel capacity utilization.
text to specify the time horizon on which the capacity utilization for an individual voyage is set. In microeconomic analysis of freight rate formation for individual voyages (see e.g. Alizadeh and Talley, 2011a,b; Adland et al., 2016) we are concerned with the immediate equilibrium between the number of vessels and cargoes in a particular loading area. Cargoes must be shipped within a certain time window (laycan), and only vessels that are commercially available and physically able to meet the laycan can offer transportation. Depending on whether there are more ships than cargoes or vice versa, cargoes are then matched with ships (subject to capacity constraints) in either a perfectly competitive micro-market or an auction-like process, respectively (Stopford, 2009). If we ignore speed changes and the possibility of inter-temporal substitution (Zannetos, 1966), it follows that available supply (DWT capacity) is fixed in the short run (in the order of days). Consequently, an increase in short-run demand and, thus, a higher freight rate, will lead to higher contemporaneous capacity utilization. In the longer run (weeks), high freight rates in one region of the world would attract more vessels which could push freight rates and utilization down. On an even longer time horizon (years), high average freight rates would encourage the construction of more vessels, which would also push freight rates and capacity utilization down. We also note that in oligopolistic shipping sectors, where owners can influence cargo volumes by price differentiation, we may observe a negative shortrun relationship between freight rates and volumes transported by a specific carrier1 (see, Zerby and Conlon, 1978, for a discussion of capacity utilization in the case of price differentiation). In early work, Kalindaga (1990) estimates the impact of part-cargo loading based on a sample of tanker fixtures for each of the main loading areas. Similarly, using fixture data for crude oil tankers between January 2006 and March 2009, Alizadeh and Talley (2011a,b) find that reported utilization is higher for the larger vessels (average 90.8% for VLCCs compared to 77% for Aframax tankers). This negative relationship between vessel size and utilization is expected. Given the economies of scale in ship size (Stopford, 2009), which also applies to fuel consumption and crew size, the share of DWT reserved for ‘non-cargo’ use is necessarily smaller for larger ships. It follows that the maximum utilization is negatively and non-linearly related to vessel size (DWT). We note that only a fraction of all voyages are observable in public fixture data, as some will be performed by in-house tonnage and others are purposely kept ‘‘off market” by the parties involved. Veenstra and Dalen (2008) provide a thorough review of the related problem of constructing freight rate indices based on incomplete public fixture data. In general, the type of charterparty will influence whether the charterer is incentivised to optimize cargo intake (i.e. maximizing capacity utilization). Under a time charter, the charterer is in commercial control of the vessel and pays a fixed daily hire irrespective of the cargo size (Stopford, 2009). Accordingly, the charterer will achieve the minimum unit transport cost by maximising capacity utilization. Conversely, under a voyage charter, the charterer pays a fixed $/tonne irrespective of the cargo size. For the shipowner the situation is the reverse, with no economic gain from an increased cargo intake under a time (trip) charter, and a proportional improvement in earnings as a function of the cargo size under the voyage charter. One solution to approximating capacity utilization in the absence of cargo size data is to use vessel draughts for individual voyages reported by the Automated Identification System (AIS). Jia et al. (2015) compare different models for the conversion between vessel draughts and cargo sizes in the drybulk market and conclude that they can give good approximations of the onboard cargo size. However, they note that this approach relies heavily on the geographical coverage of the data and the accuracy of the manual draught reports by the ship’s crew. Smith et al. (2013) use AIS-reported vessel draughts as a basis for estimating capacity utilization across shipping segments and find it to be approximately 80% or less in the case of crude oil tankers, but does not investigate its determinants. The literature on activity-based energy use and vessel emissions similarly rely heavily on rules of thumb and the use of average, static, capacity utilization. Endresen et al. (2007) applies a flat 80% DWT utilization for the reconstruction of historical emission levels over four decades, but note that this variable represents the greatest uncertainty in the estimates of fleet productivity. Behrens et al. (2003) report cargo utilization of 91% for bulk carriers and 87% for tankers above 50,000 DWT for the year 2001. 1
We thank an anonymous referee for pointing out the nuances with regards to time horizon and market structure.
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
4
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
In summary, the literature on capacity utilization lacks theoretical structure and has not yet attempted to model the determinants or dynamics of capacity utilization at the micro (individual voyage) level. Our paper fills this gap in the literature. 3. Theoretical background and methodology 3.1. Deriving optimal capacity utilization As a starting point for our capacity utilization model, it is useful to start with the classical profit maximization problem for a shipowner (see, for instance, Ronen, 1982; Strandenes, 1999). However, rather than assuming that the daily fuel consumption is a function purely of vessel speed we adopt the more general approach that fuel consumption is determined by both speed and capacity utilization. If we ignore port days and port costs, the operating profit function can be expressed as:
p ¼ s DWT u
D 24v
p f ðv ; uÞ
ð1Þ
where s is the spot rate ($/tonne); DWT is the ship’s deadweight tonnage (tonnes); u is the utilization rate; D is the trip distance (nautical miles); v is the average voyage sailing speed (knots); p is the bunker price ($/tonne); f(v, u) is daily fuel consumption at speed v and utilization rate of u (tonnes/day). The general point to make here is that there exists a trade off between the marginal revenue of one extra tonne of cargo (the spot freight rate) and the marginal cost of an extra tonne of cargo (i.e. the incremental fuel consumption due to higher displacement). The optimal utilization rate u⁄ is, thus, the ratio where these two effects are identical. The daily fuel consumption function f(v, u) can be expressed as a function of displacement and speed (MAN Diesel and Turbo, 2013; Gorski et al., 2013):
f ðv ; uÞ ¼
v vd
3 2=3 r Fd
ð2Þ
rd
where vd is the design (maximum) sailing speed (knots); r is the voyage displacement (tonnes); rd is the full load displacement (tonnes); Fd is the daily fuel consumption at maximum speed and displacement (tonnes/day). The displacement of a vessel is the sum of its lightweight (i.e. the total weight of steel, engine, equipment etc.) and its deadweight. As a first order approximation, the displacement ratio rr can be replaced by the utilization rate (u) provided d
that the vessel’s lightweight (LDT) is much lower than its DWT.2 In this case, Eq. (2) can be rewritten as
p ¼ s DWT u 24v =D p
v vd
3 F d f ðuÞ2=3
ð3Þ
The utilization rate that maximizes the profit p for given values of the spot freight rate and bunkers price can be found by taking the partial derivative of p with respect to u and solving the equation by setting it to zero, i.e. @@up ¼ 0 subject to umin 6 u 6 umax . The solution for the optimal utilization ratio is
u ¼
!3 2ðvv Þ3 F d p D d
3s DWT 24v
ð4Þ
Interestingly, the optimal utilization ratio is related to the ratio of fuel prices and spot rates, as is the optimal speed of a vessel (Ronen, 1982). We emphasize that it is not the solution to Eq. (4) per se that is interesting in our context, as the optimal utilization will generally be the corner solution u⁄ = umax (i.e. full load). It is only in extreme cases of very low spot rates and very high bunker prices that the theoretical optimal load may be less than full load.3 However, Eq. (4) is useful as it gives us a first indication of the variables that play a role. 2
MAN Diesel and Turbo (2013) suggests an average DWT/LDT ratio of 6 for tankers and bulk carriers. Strictly speaking there exists a lower bound to the spot freight rate which is contingent on the bunkers price (Adland, 2012) and so some combinations of s and b are not plausible. 3
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
5
3.2. Building an empirical model With reference to Eq. (4) we can divide variables into three groups: macroeconomic (spot freight rate s, fuel prices p), voyage-specific (speed v and distance D) and vessel-specific variables (DWT). We expand on their expected influence on capacity utilization below. We also add two new variables - a Southern hemisphere winter dummy and the iron ore spot price - to control for the potential effects of seasonal weather restrictions and the impact of commodity markets. 3.2.1. Macroeconomic variables Spot freight rate. Following Wijnolst and Wergeland (1996) and Smith et al. (2013) we expect that freight market conditions, as proxied here by the $/tonne spot freight rate for a voyage charter, are positively correlated with capacity utilization. During a weak freight market, with a corresponding oversupply of vessels relative to the volume of available cargoes, it will be rational for operators to compete for low utilization (part cargoes) as long as the resulting TCE is positive and there is no alternative employment. Conversely, in strong freight markets where transport capacity is scarce, charterers have a strong incentive to maximize the cargo intake for every voyage they undertake, subject to technical constraints. Bunker price. High fuel prices should encourage reduced sailing speeds and daily fuel consumption. Due to the non-linear relationship between sailing speed and fuel consumption in Eq. (2), this will translate into a lower required volume of fuel onboard for a given voyage.4 Provided that the cargo size is maximized, the lower intake of fuel for the voyage will increase utilization, all else equal. Accordingly, we expect a negative relationship between fuel prices and DWT utilization. Iron ore price. China imports 50% of Brazilian iron ore (UN Comtrade, 2016). Lower global prices makes imported iron ore more competitive compared to high-cost domestic producers and this substitution effect will increase seaborne demand, all else equal. We therefore expect a negative relationship between the iron ore price and capacity utilization. 3.2.2. Voyage-specific variables Distance. The relationship between distance and utilization is not clear cut. In principle, longer distances require higher bunker intake at the start of a voyage and, correspondingly, lower capacity utilization as long as the cargo intake is maximized. However, long voyages also have a higher opportunity cost of part loads (i.e. lower earnings and higher unit transport cost for longer) which should encourage higher utilization by economically rational agents. Moreover, long voyages in the drybulk market will typically serve the modern deep-water ports in North-East Asia which have fewer draught restrictions than regional ports closer to the main export areas (Brazil in our empirical case), which again points to higher utilization. Winter dummy. Voyages that pass through ‘winter zones’ as per the International Load Lines regulations will have reduced maximum draughts, and therefore lower utilization all else equal. In our empirical case, as 80% of Brazilian iron ore is passing the Cape of Good Hope eastbound to the Middle East and Asia (UN Comtrade, 2016), we would anticipate lower draughts in the Southern Hemisphere winter. Our hypothesis thus becomes that voyages commencing in Q4/Q1 have lower average utilization. Speed. We do not include speed in our study due to the unavailability of speed data for individual voyages for our entire sample. Adland and Jia (2016) study voyage speeds for bulk carriers in the 2011–2012 period and conclude that speeds are relatively constant throughout this period, with no statistical significant impact of freight market conditions or fuel prices for the laden leg. We are therefore fairly confident that omitting the voyage speed variable does not negatively affect our results. 3.2.3. Ship-specific variables DWT. We expect a negative and non-linear relationship between vessel size and capacity utilization as the non-cargo share of DWT (fuel, fresh water and fixed weights) is smaller for large ships due to economies of scale in ship operation. However, aside from this technicality, commercial practice and physical port constraints may also affect utilization as a function of vessel size. For instance, draught restrictions in the load or discharge port may limit the cargo size for very large vessels and reduce capacity utilization. Similarly, the availability of storage areas or the commercial needs of the buyer may limit cargo sizes. Finally, there may be hold strength limitations, particularly for heavy cargo such as iron ore. We note that in some cases, the shipowner may be compensated for such ‘short loading’ by way of a higher freight rate. We here test different specifications with regards to vessel size: DWT, DWT2 and dummies for DWT size groups. 3.3. Regression model To test the importance of the above determinants of the utilization rate empirically, we estimate various specifications of the following general model using a pooled OLS regression:
4 By way of example, a standard 175,000DWT Capesize sailing a 10,000 n.m. laden voyage will consume a total of 1555 tonnes of fuel at an average speed of 15 knots and only 691 tonnes at 10 knots. This releases an additional 864 tonnes of cargo capacity, equivalent to 0.5%-points utilization.
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
6
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
U j ¼ a0 þ a1 sj þ a2 pj þ a3 qj þ a4 Dj þ a5 Wdj þ
X
xi DWT i;j þ ej
ð5Þ
i
where Uj is the observed utilization rate for voyage j; sj is the spot freight rate; pj is the bunker price; qj is the iron ore spot price; Dj is the trip distance; Wdj is the winter dummy variable; and DWTi,j is the set of i variables defining the relationship between size and utilization. ej is a random perturbation such that E(ej) = 0 and Var(ej) = r2. Pooled OLS is appropriate in this context because there are, on average, few observations along the individual vessel dimension in our panel data and so we care less about time-invariant effects. As each freight rate, fuel price and iron ore price is matched to individual voyages which need not be consecutive in the time dimension we also do not worry about the serial correlation that normally would be present in time series of these macro variables. We apply robust regression estimation using Huber-White sandwich estimators to control for heteroscedasticity, lack of normality, and outlier observations that exhibit large residuals, leverage or influence (Huber, 1967; White, 1980). We also check for multicollinearity using the Variance Inflation Factor (VIF) diagnostic. An important point is whether the reduced-form model in Eq. (5) could suffer from an endogeneity problem. Fortunately, it is straightforward to establish that our independent variables are not caused by DWT utilization in the short run. The bunker price is a derivative of the global crude oil market, of which marine fuel consumption is a tiny share. As DWT utilization will vary within a relatively small range and the speed and fleet size can be taken as constant in the short run, it follows that the impact of draught variations on the global demand for oil is negligible. The demand for commodities is generally taken to be inelastic with regards to freight market conditions (Adland and Strandenes, 2007). Provided that ocean transportation is not a bottleneck that affects the supply of the commodity, the commodity price is therefore not caused by freight market conditions in general or DWT utilization in particular. Voyage distances are given by the location of sources and consumers of the raw material in question. In our empirical case, neither the location or production of iron ore mines or the location or production of steel mills can reasonably be assumed to depend on vessel capacity utilization. The short-run distribution of vessel sizes (DWT) is given by the geographical positions and commercial availability of vessels and therefore constant in the very short run. Additionally, iron ore cargoes are sold almost exclusively on a Cost Insurance Freight (CIF) basis, which means it is the seller of the cargo (in our empirical case, Brazilian miner Vale) that must arrange ocean transportation. The cargo shipment programme is a mix of long-term volume contracts and spot cargoes, but all cargoes will be sold in the iron ore market some time before the freight is arranged. It follows that the cargo size, or cargo size range, is defined before a new cargo is offered in the freight market. The momentary demand for transportation is therefore given by the number of cargoes and their destination, with the spot freight rate a result of the matching process with the available fleet. Consequently, there is no causality from DWT utilization to the spot freight rate in the very short run. 4. Data 4.1. Cargo size data We collected monthly reports of vessels leaving the six main iron ore export terminals in Brazil from port agents LBH Group over the period November 1, 2008 to August 1, 2014. Table 1 summarizes the physical characteristics of the six terminals. The port agent reports contain the following data fields:
Report date Vessel name DWT Arrival time, ETA (Expected Time of Arrival) Berthing time, ETB (Expected Time of Berthing) Departure time, ETD (Expected Time of Departure) Destination Cargo quantity.
The sample consists of 9862 individual port calls by 3954 vessels ranging from 24,000 DWT to 405,000 DWT with an average size of 152,000 DWT. Table 2 shows the descriptive statistics for cargo size by port. Sailing distances for the laden leg are obtained from a standard marine distance calculator (axsmarine.com) under the assumption that vessels avoid pirate zones where applicable. Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
7
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx Table 1 Brazil iron ore terminal details. Source: Compiled by authors from vale.com, samarco.com. Terminal name
Pier
Shipper
No. Berths
Latitude
Longtitude
Max DWT
Max LOA
Max draught
Load rate (MTPH)
Ponta da Maderia
Pier I Pier II Pier III
Vale
2
2.57
44.38
420,000 155,000 220,000
345/280
25
16,000 8000 8000
Ponta Ubu
Westside Pier Eastside Pier
Samarco
2
20.78
40.58
250,000 150,000
308 240
15.5 16.8
10,000 7000
CSN Terminal
Pier 401
Ferteco
2
22.87
43.77
n/a
17.3
5000
Tubarao
Pier 1 Pier 2
Vale Vale
2 2
20.28 20.28
40.25 40.25
200,000 365,000
285/320 350
17 20
7000 16,000
Vale Vale
1 1
22.87 23.00
43.77 44.02
230,000 300,000
Itaguai GIT
18 15.2
Table 2 Descriptive statistics for cargo size by port (tonnes). Source: LBH Group, compiled by authors.
CSN GIT Itaguai PDM Tubarao Ubu
Obs
Mean
Std. Dev.
Min
Max
965 1114 834 2745 3165 992
156,482 186,669 149,624 197,710 169,826 119,638
19,772 55,285 34,450 73,401 81,096 54,189
39,285 29,220 27,500 20,052 8000 29,998
198,236 285,902 274,380 395,384 395,373 204,350
4.2. Macroeconomic data As China is by far the largest importer of Brazilian iron ore, we use the Baltic Capesize Index C3 for the route Tubarão to Qingdao as reference for the freight market. As per Baltic Exchange (2015) standard route assumptions, vessels are delivered for the trip in the Antwerp-Rotterdam-Amsterdam (ARA) area and then sail in ballast to Tubarão for loading. Assuming that vessels are fixed upon departing the last discharge port, we match each laden voyage with the Baltic spot freight rate 21 days prior to the loading port ETA5 in order to reflect freight market conditions at the time of the fixture. We use the 380cst Rotterdam heavy fuel price at the same time as a reference for fuel costs. Both the spot freight rate and fuel price series were obtained from Clarkson Research (2016). We base our proxy for the value of the cargo on the 62%Fe-content spot iron ore index as constructed by The Steel Index (TSI) and obtained from Bloomberg. The TSI 62% spot iron ore price index is defined as the price per tonne of iron ore delivered in Tianjin, China, on a Cost, Insurance and Freight (CIF) basis. In order to avoid accounting for the freight component twice in our model, we calculate the Free On Board (FOB) price by subtracting the above freight rate from the CIF price and using the FOB price in our estimations. Similarly to the freight rate, we match each cargo with the iron ore price 21 days prior to the time of arrival at the load port anchorage in Brazil in order to reflect market conditions at the time of sale.6 Fig. 2 shows the development in our macro variables during the sample period. 4.3. Capacity utilization Fig. 3 shows the kernel density of the utilization, using bandwidth 0.725% and the Gaussian kernel function. We see that the majority of voyages are performed with fairly high capacity utilization (>85%), with a peak around 96%. Indeed, the average capacity utilization across the sample is 92%. The observations with very low capacity utilization may be either a result of vessels loading at multiple ports in Brazil, making it appear as if the cargo intake at each port is unusually low, or errors in the data entry process at the port agent. For instance, the cargo size or DWT may be wrongly entered, or a vessel name may not be unique to a certain size vessel. Fig. 4 shows a scatter plot of vessel capacity utilization vs. DWT. We see a clustering of vessel sizes into roughly four main DWT groups: Panamax and smaller (<100,000 DWT), Capesizes (<250,000 DWT), iron ore carriers and Valemaxes (>400,000 DWT). We note the presence of some ‘‘mini-Capes” around 115,000 DWT size but choose to include these with their bigger Capesize cousins. Table 3 summarizes the utilization statistics by size group. Both Fig. 4 and Table 3 confirm that smaller vessels (Panamax and smaller) have substantially lower utilization on average. 5 This includes 17 days sailing in ballast from Rotterdam plus an assumed average of 4 days between fixture and delivery as per Baltic Exchange (2015) definitions. 6 The sale of the cargo will probably have taken place even earlier than the fixture of the ship though this would be unavailable private information.
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
8
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
$/tonne
$/tonne 800
200 180
700
160 600 140 500
120
400
100 80
300
60 200 40 100
20 0 21/10/08 21/10/09 21/10/10 21/10/11 21/10/12 21/10/13 21/10/14 iron ore (LH)
freight (LH)
0
bunker price (RH)
10 0
5
Density
15
20
Fig. 2. Freight rate, bunker price and iron ore price development. Source: Clarkson Research (2016), Bloomberg.
.2
.4
.6
.8
1
utilization Fig. 3. Kernel density of vessel utilization.
Fig. 4. Scatter plot of utilization vs. DWT.
We note that there is some non-linearity in the mean-utilization-DWT relationship, with particularly the sub-Capsize segment but also the very large vessels showing lower utilization than the standard Capesize segment (around 180,000 DWT).
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
9
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx Table 3 Descriptive statistics for capacity utilization by size group and port. DWT range
Obs
Mean
Std. Dev.
Min
Max
0–99,999 100,000–249,999 250,000–399,999 400,000+
1589 6405 1652 169
0.801 0.944 0.936 0.934
0.175 0.069 0.074 0.079
0.285 0.153 0.027 0.650
1.00 1.00 1.00 0.99
Port name
Obs
Mean
Std. Dev.
Min
Max
CSN GIT Itaguai PDM Tubarao Ubu
965 1114 834 2745 3165 992
0.916 0.954 0.924 0.945 0.904 0.860
0.081 0.058 0.075 0.080 0.131 0.149
0.230 0.388 0.153 0.140 0.027 0.250
1.00 1.00 1.00 1.00 1.00 1.00
Table 4 Regression results. Dependent variable: utilization % (1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
Adj. R-squared VIF
0.104
0.042
0.232
0.236 1.39
0.242 1.49
0.238 1.380
0.241 1.530
0.256 1.54
0.261 1.44
0.264 1.41
Constant
0.84 (201.78) [0.00]
0.89 (399.22) [0.00]
0.80 (182.94) [0.00]
0.77 (132.30) [0.00]
0.83 (112.51) [0.00]
0.80 (108.30) [0.00]
0.85 (154.41) [0.00]
0.77 (175.89) [0.00]
0.77 (104.92) [0.00]
0.79 (109.86) [0.00]
DWT (tonne) (’000000)
0.46 (23.69) [0.00]
0.14 (32.17) [0.00]
0.14 (32.47) [0.00]
0.14 (32.43) [0.00]
0.14 (32.26) [0.00]
0.14 (32.31) [0.00]
0.12 (23.61) [0.00]
0.12 (23.81) [0.00]
0.12 (24.01) [0.00]
DWT_3 (tonne) 250,000–399,999t
0.14 (28.56) [0.00]
0.14 (28.91) [0.00]
0.14 (29.17) [0.00]
0.14 (28.89) [0.00]
0.14 (29.05) [0.00]
0.11 (19.80) [0.00]
0.11 (20.16) [0.00]
0.11 (20.41) [0.00]
DWT_4 (tonne) 400,000+t
0.13 (17.91) [0.00]
0.14 (18.40) [0.00]
0.14 (18.79) [0.00]
0.14 (18.27) [0.00]
0.14 (18.57) [0.00]
0.11 (13.78) [0.00]
0.11 (14.16) [0.00]
0.12 (14.68) [0.00]
1.08 (7.51) [0.00]
0.69 (4.54) [0.00]
0.75 (4.88) [0.00]
0.73 (4.92) [0.00]
0.68 (4.85) [0.00]
DWT^2 (tonne) (’E+12)
0.74 (18.44) [0.00]
DWT_group_dummy (<100,000t benchmark) DWT_2 (tonne) 100,000–249,999t
Freight rate (C3 route $/tonne) (’000) Bunker price (380cst Rotterdam $/tonne) (’000)
[0.08] (7.49) [0.00]
Iron ore price (62% FOB $/tonne) (’000)
0.03 (0.66) [0.51]
[0.08] (7.67) [0.00] 0.17 (6.08) [0.00]
[0.07] (9.50) [0.00]
0.07 (1.90) [0.06]
0.14 (4.96) [0.00]
Distance (nautical mile) (’000)
5.97E03 (13.46) [0.00]
5.14E04 (13.03) [0.00]
5.71E03 (12.81) [0.00]
Winter_dummy
3.51E04 (0.18) [0.85]
5.14E04 (0.27) [0.79]
1.20E03 (0.63) [0.53]
Note: Figures in () are t-stats; figures in [ ] are the p- values of H0: zero coefficients.
5. Empirical results Table 4 shows the results for the various specifications of our multiple regression model. We note that the high statistical significance and estimated signs of all parameters are stable across specifications, which is encouraging for the validity of our
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
10
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
hypotheses and correct specification of our empirical model. Moreover, the VIF diagnostic test suggests the absence of multicollinearity in all cases. Of the three different relationships between vessel size (DWT) and capacity utilization (specifications 1–3), the size group dummies perform the best. Our results are similar to those found for tankers in Smith et al. (2013) in the sense that smaller vessels operate with lower average capacity utilization. However, the relationship is non-linear, with vessel sizes above 100,000 DWT showing only marginal changes. For the remainder of the specifications we use only the DWT group dummies as they have the highest explanatory power. In specification 4 we add the impact of the freight market. As expected, the spot freight rate loads positively and is highly statistically significant. From an economic point of view, a change in the spot freight rate from $30/tonne observed at recent peaks down to the $10/tonne at the cycle lows implies a reduction in the vessel capacity utilization of 2.2%-points, all else being equal. Given that the ship-specific variation in capacity utilization is relatively low (ref. the peaked distribution in Fig. 3), this is economically meaningful. When we add the fuel price and/or commodity price as explanatory variables (specifications 5 and 6), this effect is reduced somewhat. The bunker fuel price is also a highly significant determinant, and the iron ore price less so, though both conform to our a priori expectations. However, we note that the fuel price and commodity price appears to be picking up the same ‘commodity-boom’ effect, as inclusion of the former in the regression (specifications 5 and 7) makes the latter insignificant at standard levels of significance. The economic intuition for including the fuel price as a determinant, aside from it showing up in Eq. (4), is arguably less appealing than our other macro variables, particularly with low variation in voyage speeds. The distance variable is also highly significant and loads positively, which suggests that the alternative cost effect and/or port infrastructure constraints dominate the effect on fuel intake. In economic terms, the increase in utilization between a short and long voyage, say Rotterdam (4965 nautical miles) vs Qingdao (11,017 nautical miles), is approximately 3.6%points. Seasonality, at least how we have defined it here, does not seem to make a difference to vessel capacity utilization. 6. Concluding remarks We have shown that several macro- and microeconomic determinants, notably freight market conditions, fuel prices, vessel size, sailing distance and vessel size, influence the vessel capacity utilization of drybulk carriers. From a maritime economic perspective this is an important finding, as it means that real vessel earnings receive a ‘double blow’ during poor freight markets from both low spot rates and reduced capacity utilization. The latter effect results both from increased competition for sub-optimal cargo sizes and, likely, more efficient sailing patterns in times of overcapacity. In terms of maritime environmental policy, our results suggest that the emissions per tonne-mile of cargo transported are greater during low freight markets, due to the lower estimated utilization. Consequently, the typical assumption of static capacity utilization in the literature will lead to biases in estimates that are correlated with the freight market cycle. Our study is a first step to improve the accuracy and quality of national and global emission accounting for policy making along this important dimension. We acknowledge that there remains a large unexplained component in the observed vessel capacity utilization data. While some of this is likely due to measurement errors in the original dataset, future revisions of this work should try to further improve the explanatory power of the models herein. In particular, vessel age and hold structure should be considered. Older tonnage will expectedly have lower capacity utilization – both because it is less attractive on the spot market and therefore may have to bid for sub-optimal cargo sizes but also because of reduced structural strength. This is particularly a concern for heavy cargo such as iron ore, where the fatigue caused by numerous loading and discharge cycles over the life of the vessel can eventually cause a collapse of hull sections (Paik and Melchers, 2008). Similarly, specialised iron ore carriers and other bulk carrier vessels denoted as ‘‘strengthened for ore” will expectedly have a higher capacity utilization than vessels not built to this higher specification. This is in part to avoid failures in the ship structure, particularly during the loading process. However, ore carriers also have a different hold shape than standard bulk carriers, ensuring a higher centre of gravity and better sea keeping performance. It is also worth modeling the relationships between maximum vessel draughts and any draught restrictions in the loading or discharge port in detail, as this is known to sometimes be a limiting factor. The challenge here is that the final port of discharge will not always be known. Acknowledgements This research was partly funded by the Research Council of Norway’s SMARTRANS programme under the project CARGOMAP – mapping vessel behaviour and cargo flows (project #239104). We thank two anonymous referees for their valuable comments as well as participants at the International Forum for Shipping, Ports and Airports (IFSPA 2015) at the Hong Kong Polytechnic University for input on an earlier version of this paper. References Adland, R., 2012. Microeconomic modelling of the supply function in bulk shipping revisited. In: International Association of Maritime Economists (IAME) Annual Conference, Taipei. September 5–8. Adland, R., Jia, H., 2016. Dynamic speed choice in bulk shipping. Marit. Econ. Logist. http://dx.doi.org/10.1057/s41278-016-0002-3. Adland, R., Strandenes, S.P., 2007. A discrete-time stochastic partial equilibrium model of the spot freight market. J. Transp. Econ. Policy 41 (2), 189–218. Adland, R., Cariou, P., Wolff, F.-C., 2016. The influence of charterers and owners on bulk shipping freight rates. Transp. Res. Part E 86, 69–82.
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023
R. Adland et al. / Transportation Research Part A xxx (2016) xxx–xxx
11
Alizadeh, A.H., Talley, W.K., 2011a. Vessel and voyage determinants of tanker freight rates and contract times. Transp. Policy 18 (5), 665–675. Alizadeh, A.H., Talley, W.K., 2011b. Microeconomic determinants of dry bulk shipping freight rates and contract times. Transportation 38 (3), 561–579. Baltic Exchange, 2015. Manual for Panelists,
Please cite this article in press as: Adland, R., et al. The determinants of vessel capacity utilization: The case of Brazilian iron ore exports. Transport. Res. Part A (2016), http://dx.doi.org/10.1016/j.tra.2016.11.023