The purpose of this paper is to investigate the impact of shipyard and shipowner heterogeneity on the price formation for individual newbuilding contracts.
The model controls for the shipbuilding market cycle, input costs, firm size, yard experience and contract-specific variables and captures the impact of yard and owner heterogeneity in fixed-effects regressions. The data sample contains contract information on 3,759 tankers, bulkers and container vessels constructed at 77 shipyards between 1990 and 2014.
Although the newbuilding price benchmarks (market conditions) and gross domestic product per capita (salary costs) are influential covariates, the main conclusion is that shipyards and, particularly, shipowners play an influential role on the US$ per Compensated Gross Tonnage price level in individual contracts.
The paper represents the first study of the impact of buyer and seller heterogeneity at the micro level in the shipbuilding market.
Adland, R., Norland, K. and Sætrevik, E. (2017), "The impact of shipyard and shipowner heterogeneity on contracting prices in the newbuilding market", Maritime Business Review, Vol. 2 No. 2, pp. 58-78. https://doi.org/10.1108/MABR-11-2016-0031
Emerald Publishing Limited
Copyright © 2017, Pacific Star Group Education Foundation
Shipbuilding is a truly global industry, with about 30 countries – mostly in Asia or Europe – having a significant production of merchant vessels (Stopford, 2009), of which 18 countries are considered major shipbuilders according to Clarkson Research (2016). In aggregate, the industry satisfies the conditions for a perfectly competitive market (Stopford, 2009, Alizadeh and Nomikos, 2009, Bertram, 2003):
both buyers and sellers are price takers;
the number of firms is large;
no barriers to entry;
products are identical;
complete market information; and
sellers are profit-maximizing entities.
However, at the micro level – i.e. the matching of a specific shipyard to build a ship for a specific owner – we postulate that heterogeneity will start to play a role. First, as discussed in Wijnolst et al. (2009), depending on how we define the shipyard “product”, ships are arguably not identical. Clearly, individual ships of the same category and size can be very different in terms of detailed technical specifications and quality. Even if we redefine the “product” to be shipyard building capacity – as most yards are willing to build any vessel from a large catalogue of designs – some heterogeneity will remain in terms of delivery times and shipyard experience with the ship type. Second, while the global number of shipyards is large, they are highly heterogeneous in terms of dock size, experience, expertise and the range of designs offered, and this may impact the number and type of orders they attract. For instance, Sauerhoff (2014) finds that shipyard competency – as determined by the variables experience, market expertise, cooperation with suppliers and external exchange of information – affects the yard’s ability to secure new contracts. Stott (1995) finds that a number of attributes related to a vessel, such as ease of maintenance and operation, fuel consumption, speed, safety and delivery conditions had a positive impact on the probability of attracting orders for the shipbuilder. However, he also reports that shipowners are reluctant to pay a significant premium for such features.
The influence of buyers and sellers on price formation is a new area of research within maritime economics. In the only existing study, Adland et al. (2016a) find that the identity of charterers and shipowners influence fixture rates even in the highly competitive spot freight market for large bulk carriers and tankers. The impact of buyer and seller heterogeneity at the micro level (i.e. individual contracts) in the shipbuilding market has not yet been investigated. The purpose and contribution of this paper to the literature is to model the price formation for individual newbuilding contracts and, specifically, assess the impact of yard and shipowner heterogeneity while controlling for other key variables such as shipyard experience and delivery lead time.
The remainder of this paper is structured as follows. In Section 2 we review the relevant shipping and shipbuilding literature. Section 3 presents our empirical methodology to consider buyer–seller heterogeneity. Section 4 presents the data and descriptive statistics. Section 5 presents our empirical results. Section 6 concludes with recommendations for future research.
2. Literature review
The literature on the shipbuilding industry can be broadly divided in two streams:
competitiveness and strategic analysis; and
the formation of newbuilding prices at the macro level.
Jiang et al. (2013) state that shipyard competitiveness is often assessed based on internal factors, such as costs. However, since shipbuilding is exposed to general market conditions and governmental interference, the authors introduce profit rate as a means of measuring competitiveness, and thus account for both internal and external factors. They find that China is the most competitive shipbuilding nation in the tanker and bulker markets, ahead of South Korea and Japan. Moreover, they find that China’s competitiveness stems from its lower cost base, while South Korea and Japan rely on a positive deviation from the market price. Importantly for our case, this suggests that a degree of price differentiation is present in the market. Jiang and Strandenes (2011) report similar results and state that China’s competitiveness stems mainly from lower labour and equipment costs, with steel costs being similar across the three major shipbuilding nations.
Much research has been devoted to the modelling of equilibrium prices within the four shipping markets – newbuilding, freight, second-hand vessels and demolition (Stopford, 2009). Beenstock (1985) assumes that the price of a new vessel perfectly reflects the second-hand price of a comparable ship at the time of delivery. Beenstock and Vergottis (1989a, 1989b) relax the strict assumption of perfect correlation between newbuild and second-hand prices and estimate an aggregated econometric model in which freight rates, newbuilding and second-hand prices, among other variables, are jointly and dynamically determined. Strandenes (1984) proposes that the price of a second-hand ship must be equal to the present value of the ship’s earnings. Strandenes (1986) argues that newbuilding prices are set based on the expected present value of future earnings, while second-hand prices are given by the weighted average of short- and long-term profits. Tsolakis et al. (2003) conclude that the main drivers for second-hand prices are the newbuilding price and time charter rates, although these variables are affecting distinct vessel types and sizes differently. Using a Vector Error Correction Model framework, Adland et al. (2006) test an equilibrium relationship between newbuilding prices, second-hand prices and freight rates in the 2003-2005 drybulk boom, and find that the second-hand market was closely co-integrated with the newbuilding and freight markets and behaved in accordance with maritime economic theory.
According to Stopford (2009), newbuilding prices are set by the number of slots available at shipyards in a given timeframe, and the number of vessels demanded for the same period. The demand for new orders is influenced by freight rates, prices for second-hand vessels, market sentiment, availability of credit and liquidity. Supply is affected by current capacity, shipyard costs, exchange rates and government subsidies. Despite this straightforward market mechanism, part of the literature suggests that newbuilding prices show too low volatility compared to second-hand prices and long-term freight rates. Zannetos (1966) argues that this is due to market imperfections and externalities, such as over-capacity and production smoothing incentives. This view is shared by Strandenes (2010) who suggests that one of the causes is strong labour unions in shipbuilding. Dikos (2004) suggests that an alternative explanation can be found within the framework of a functioning competitive equilibrium if the actions of shipyards are interpreted as a result of exercising real options to offer capacity under uncertainty. Adland and Jia (2015) argue that an order for a newbuilding can be interpreted as a forward contract with a typically downward-sloping price curve (i.e. there is a premium for early delivery). Since lead times and payment schedules are correlated with market conditions, this leads to an expected lower volatility of newbuilding prices. Haddal and Knudsen (1996) investigate whether the global shipbuilding market constitutes a single market by evaluating correlations between newbuilding prices for various ship types.
To our knowledge, no authors have investigated price formation in the shipbuilding market at the micro (individual contract) level and so our paper is the first attempt at filling this gap in the literature. However, such studies are common in the empirical literature on freight rate formation, with Tamvakis (1995), Tamvakis and Thanopoulou (2000), Mokia and Dinwoodie (2002), Alizadeh and Talley (2011a, 2011b), Köhn and Thanopoulou (2011), Agnolucci et al. (2014) and Adland et al. (2016a, 2016b), all estimating models on the determinants of the freight rate for individual contracts (fixtures) in both the spot and timecharter bulk shipping markets.
3.1 Dependent variable
The contracting price for an individual vessel is dependent on many technical specifications, for instance, the engine type and maker, the number and make of discharge pumps for tankers and the anti-fouling paint. Developing a model that accounts for all possible differences between individual vessels is clearly not a workable solution, mainly because the number of potential variables is very large, restricting the degrees of freedom, but also because not all vessel specifications are observable to the researcher. Additionally, models would have to differ across ship types to account for very different designs and equipment types. To obtain at a parsimonious model that is applicable across ship types, we here proxy the sophistication of a vessel by its Compensated Gross Tonnage (CGT) value. The CGT unit of measurement was introduced by Organisation for Economic Co-operation and Development as an indicator of the work amount required to build a ship and is calculated based on the following formula, where A and B are factors specific to the various ship types (OECD, 2007):
For contract prices to be comparable, we inflate all numbers to 2014 levels and convert all currencies into US$ based on the exchange rate at the contract date. Our dependent variable, the US$/CGT price for contract i is thus defined as:
While CGT is widely accepted as the best unit of shipyard production (Bertram, 2003), it is not without flaws. The CGT value will assume a standard ship for each vessel category; thus, Stopford (2009) points out that the CGT measure has decreasing value the more sophisticated or complex a ship is. Bertram (2003) makes a similar point when stating that differences in equipment can shift the number of required man-hours, while leaving the CGT value unchanged.
Our independent variables are summarized in Table I below, with further explanations in subsequent paragraphs. To avoid having to perform partial analyses by shipping segment, we purposely choose variables that are generic and meaningful all possible vessel types and sizes. That is, we avoid, type-specific variables that are applicable only to a certain type of vessels, e.g. the number of cranes for small bulkers or pump capacity for tankers. We acknowledge that in doing so we may omit variables that are important for the price formation within individual segments. However, as we are here primarily looking to isolate the market-wide effect of buyer and seller heterogeneity on newbuilding prices, and since any individual yard and owner is likely to build several types of ships throughout our sample, this is a necessary trade-off.
3.2 Firm-specific variables
3.2.1 Owner size.
Clarksons (2016a) categorizes ship-owning firms by fleet size: single ship, very small (2-5), small (6-10), medium (11-20), large (21-50), very large (51-100) and extra-large (100+). We assign dummy variables for each category mentioned, in addition to one for firms of unknown size. We note that both shipyard and shipowner sizes are a snapshot at the time of contract signing, and the same firm can thus be a different size for different observations. This is important, as we are implicitly able to account for the impact of organic growth or merger and acquisition activity over time. Porter (1979) postulates that buyers can affect the price they pay to their suppliers if the firm is of a certain size, or purchases in large volumes (i.e. a negative relationship between owner size and price paid). These two criteria are likely to be highly correlated in the shipbuilding market, as larger shipowners are more likely to contract new vessels more frequently for fleet renewal purposes.
3.2.2 Shipyard size.
Clarksons (2016b) classifies shipyard size based on the size of the order book in millions of CGT: very small (<0.049), small (0.049 < 0.01), medium (0.1 < 0.49), large (0.49 < 1) and mega (>1). These groups make up our shipyard size dummy variables, in addition to one for yards of unknown size. Porter (1979) suggests that larger producers are able to influence the price, particularly if producers are more concentrated than buyers, as in shipbuilding. Larger shipyards could also potentially obtain higher prices because of the flexibility that comes with size. These yards are presumably capable of competing for even the largest vessels, while this might not be feasible for smaller yards. Hence, large shipbuilders are possibly operating in segments with fewer competitors, as well as having a greater variety of contracts to bid on. On the other hand, larger shipyards might be better able to utilize economies of scale, pushing costs down. This is notably the case if the shipbuilder receives several orders for identical vessels (Stopford, 2009). Whether yard size affects shipbuilding prices in a positive or negative manner is thus dependent on which of these effects is dominant.
3.2.3 Shipyard experience.
Sauerhoff (2014) finds that practical experience affects a shipyard’s ability to secure contracts. We therefore postulate that shipyard experience at the signing of a contract, defined here as the contracting year less the first year of delivery for a yard, has a positive impact on price.
3.3 Contract-specific variables
3.3.1 Delivery time.
A vessel with shorter delivery time will have a higher value to the shipowner in present value terms than an equal vessel with longer delivery time (Adland et al., 2006, Adland and Jia, 2015). Stott (1995) finds partial empirical evidence for this hypothesis, as shipbuilders with superior delivery times are more likely to attract orders. Bertram (2003) states that differences in delivery times are likely to be captured in contract prices, which is also the point of Adland and Jia (2015). Following the theory presented in the latter paper, we postulate a negative impact on price from the delivery time, defined here as the build date less the contracting date (months).
3.3.2 Top speed and gross tonnage.
As CGT is only an approximate measure of the work content of a ship of a certain type (Bertram, 2003), we introduce two variables – design speed (knots) and vessel size as measured by Gross Tonnes (GT) – as proxies for the scope of work for individual vessels within a vessel type category. For instance, higher-speed vessels may have more slender hull form or larger engine installation. We acknowledge that these are imperfect measures, but they are commonly available and comparable across contracts for different vessel types.
3.4 Macro variables
3.4.1 Newbuilding price index.
In a competitive shipbuilding industry (Stott, 1995; Stopford, 2009), we expect that a large part of the contracting price for a newbuilding is explained by the shipbuilding market cycle, which we can consider as exogeneous to our micro-level model. To isolate this effect, we introduce an index of estimated competitive market prices as an explanatory variable. Clarksons (2016b) provides a monthly price index (US$/CGT) for a standard vessel built at a top-tier competitive shipyard. There are three separate indices covering bulk carriers, fully cellular container (FCC) vessels and tankers, which we match with the corresponding vessel types in the data set.
3.4.2 Delivery time and freight rate interaction variable.
As established by Adland et al. (2006) and Adland and Jia (2015), the value of shorter delivery time is partly a function of freight rates (the opportunity cost). To account for this effect, we introduce an interaction variable between delivery time and a freight rate factor variable. The factor variable is constructed such that values at or below the 33rd percentile make up our low freight rates scenario, values between the 33rd and 66th percentile make up our normal scenario and values above the 66th percentile make up the high freight rates scenario. The freight rate here represents the one-year timecharter rate for each of the vessel types at the time of contract signing. As the value of early delivery is increasing in freight rates (Adland et al., 2006; Adland and Jia, 2015), we expect a greater negative impact on US$/CGT from the interaction variable in a high freight rate scenario than in a low-rate environment. To avoid specification errors in the model that might arise from leaving out variables included in an interaction variable (Wooldridge, 2013, p. 191), freight rates and delivery time are incorporated as stand-alone variables. The positive effect of freight rates on shipbuilding demand and prices is well understood (Strandenes, 2010).
3.4.3 Gross domestic product per capita.
Costs are the single most important factor in determining the competitive position of a shipyard (Bertram, 2003). Furthermore, nearly half of the costs incurred during the construction of a standardized bulk carrier are related to overheads and labour (Stopford, 2009). Obtaining reliable time series for wages is tricky, particularly for China which is a dominant player in the industry. We therefore introduce gross domestic product (GDP) per capita as proxy for wages, as recommended by ILO (2008). Further empirical support is found by Rodrik (1999) who demonstrates significant effects from GDP per capita on wages. This variable will also to some extent capture other cost elements specific to the country of manufacture that are hard to include by other means. We expect a positive impact on price from GDP per capita.
3.4.4 Oil and steel prices.
Steel is the most important of all materials used in shipbuilding, amounting to 17 per cent of construction costs for a standard bulker (Stopford, 2009), and is proxied here by the global average price for hot rolled coil steel at the time of contract signing. The cost of energy in production is proxied by the average price of crude (i.e. a basket of dated Brent, West Texas Intermediate (WTI) and Dubai crude oil). Both cost drivers expectedly have a positive impact on US$ per CGT.
3.5 Omitted variables
3.5.1 Currency exchange rates.
Currency exchange rates affect the relative competitiveness of shipyards as nearly all shipbuilding contracts are quoted in US$, while shipbuilders have most of their costs in local currency (Wijnolst et al., 2009). However, currency effects will eventually materialize in current market prices, and so the newbuilding price indices should indirectly capture this element.
3.5.2 Government subsidies.
Shipbuilding is a labour-intensive and export-oriented industry with spill-over effects on the domestic economy, making it the repeated beneficiary of various forms of government aid in pretty much any nation with shipbuilding output of some size (Strandenes, 2010). These subsidies are known to distort pricing mechanisms in the market, as the true production costs might not be covered by the shipyard (Jon, 2010). However, subsidies are hard to quantify, in part because not all aid is given in the form of direct monetary support. However, these effects will be at least partly captured by either shipyard fixed effects – for subsidies such as beneficial financing that remain relatively constant over time – or by the current competitive market prices for more short-term aid.
3.6 Econometric approach
When analysing data with repeated interactions of entities, in our case shipyards and shipowners, fixed- or random-effects models are suitable. However, only fixed-effects models allow estimates to account for heterogeneity, as they adjust for time-invariant unobserved effects that vary across entities, and/or constant time-varying effects across entities (Wooldridge, 2013, p. 477). Let the dependent variable, the US$/CGT price for contract i, be denoted by Ui. Furthermore, Gi is a vessels’ gross tonnage, Di is the delivery time, Ei is a measure of yard experience, Yi,y is the dummy variable for shipyard size, Oi,o is the dummy variable for owner size and θk represents the coefficients for the size dummy variables. We first propose three purely microeconomic model specifications, starting with only firm size effects:
In the second specification, we control for shipyard characteristics, excluding yard size, as this is rather constant and explained by yard fixed effects. The γ represents fixed-effect coefficients, while δ represents dummy variables generated for the respective yards and owners:
In the third specification, we have a two-way fixed-effect model controlling for both shipyard and owner fixed effects:
As we want to isolate firm fixed effects, we also introduce a second set of specifications that include macroeconomic variables to control for time-varying effects. First, we control for firm sizes [Equation (6)], second, we introduce shipyard fixed effects [Equation (7)] and thirdly, we include both yard and owner fixed effects [Equation (8)]. We here introduce an interaction variable between delivery time and freight rates, with τ being the coefficient and Fi representing dummies for low, medium or high freight rate level. Additionally, our macroeconomic specifications include It representing the competitive market price for average ships, St representing the steel price, Pt is the oil price and Gt is the shipyard nations’ GDP per capita:
We do not look at specific shipyard–owner relationships (match effects), as repeated transactions between a yard and an owner occur very rarely in our sample and is unlikely to cause distinct effects. All regressions are done with a clustered sandwich estimator to correct standard errors in the presence of heteroscedasticity (Cameron and Miller, 2015).
Our data sample contains newbuilding contracts extracted from Clarksons (2016a), covering vessels built in the period between 1970 and 2014 that are still trading. We note that while the database contains some newbuilding contract information on 91,112 vessels, only 7,604 observations included a usable contract price. The observations contain information regarding vessel name, contract parties, contract and delivery dates, carrying capacity [deadweight (DWT) and GT], CGT and vessel type. Information specific to the contract parties is also included, such as name, size of shipyard or shipowner and year of first delivery from the shipyard. These variables enable us to calculate delivery times, yard experience, contract price in US$ per CGT, in addition to creating firm size dummies. Contracts stated in currencies other than US$ are converted based on exchange rates at the time of contract signing. We inflate prices to 2014 values by using the US consumer price index (CPI). Although the USA is not a major player in either shipping or shipbuilding, its domestic inflation measure appears to be the most widely used to obtain real prices in similar studies (Jiang et al., 2013; Akram, 2009; Lizardo and Mollick, 2010).
We remove observations with missing data for at least one of our variables. Furthermore, we remove all vessel types other than tankers, bulkers and container vessels (FCCs) – for two reasons. First, the remaining three vessel types all have a large number of observations. Second, they have relatively low standard deviation of US$/CGT values, reflecting a high degree of standardization. Also, specialized vessels, such as offshore service vessels, will show a greater variation in equipment levels. As CGT does not explicitly account for variations in technical specifications within a vessel category, our model will be less suitable in these cases. The final data set includes 3,759 observations for vessels built between 1990 and 2014, constructed for 835 different shipowners by 77 different shipyards located in 11 countries. We acknowledge that the low sample size relative to the true population may distort the outcome of our analysis, for instance, if certain types of contracts are more likely to have their value publicly known. Furthermore, the database only includes vessels in the current fleet, and one could imagine that vessels from certain (poor) yards are more inclined to be scrapped early. As an indication of whether our sample is representative, Figure 1 compares the US$/CGT annual average values observed in our data set to the price indices reported by Clarksons (2016b). The observation that the two move closely in tandem gives us confidence that our data sample provides a satisfactory representation of the shipbuilding market. The larger spread in values found in the data set is expected, simply because there is one observation per vessel – as opposed to one per month for the index.
Macroeconomic data were collected from several sources. Clarksons (2016b) provided newbuilding price indices and freight rates. Oil prices were extracted from the International Monetary Fund’s (2016) commodity database, while steel prices were collected through the World Bank Commodity Database (2016a) supplemented with missing data points from Bloomberg (2016). Currency exchange rates and GDP per capita were obtained from the World Bank WDI database (2016b). All macroeconomic price variables were inflated in line with contract prices.
Presenting data for each individual shipowner and shipyard is impractical due to the large number of firms. Therefore, Table II presents yard experience and delivery times by country rather than by yard. This aggregation seems reasonable because individual yards within a nation are likely to have emerged in the same period and share other common traits (Stopford, 2009). Yard experience appears affected by established shipyards appearing frequently throughout our sample period, as there is one observation per newbuilding contract. Chinese shipyards exhibit the greatest spread in terms of experience, as expected due to the industry’s relatively recent emergence in the country. We observe outliers in delivery times, as a ship rarely arrives as early as four months, or as late as 96 months, after contract signing. However, none of the values are implausible, and could be caused by the state of the shipping market and size of the orderbook (Adland and Jia, 2015; Adland et al., 2006).
Table III reports owner and builder frequency for the ten largest players, with the remainder grouped as “other”. Market concentration among the top ten shipbuilders varies across segments, from a 42.4 per cent market share for bulkers, to 79.8 per cent for tankers and 84.5 per cent for FCCs. This reflects the lower technological complexity of bulk vessels, and the generally larger size of tankers and FCCs – presumably favouring established shipbuilders of some size (Stopford, 2009). In terms of shipowner concentration, we find that the top ten owners represent just above 20 per cent for tankers and bulkers and 39.5 per cent of the market for FCCs. COSCO, Maersk and China Shipping are the most frequent owners in the data set as a whole, which seems reasonable, as they are among the world’s largest ship operators across segments (Fan et al., 2011).
Table IV presents the descriptive statistics for our macroeconomic variables. We note the impact of the shipping and commodity supercycle in the middle of the last decade, which has caused great volatility in our sample.
5. Empirical results
5.1 Testing for multicollinearity
The Variance Inflation Factor (VIF) test is a common tool to examine the risk of multicollinearity. A VIF value of 10 is usually set as a warning level, above which determinants are said to carry risk of multicollinearity in regression estimates, though such a cut-off value is necessarily arbitrary (Wooldridge, 2013, p. 94). In Table V, the only variables displaying high VIF values are the interaction variables and freight rates. Freight rates appear both in the interaction variable and separately as control variables, resulting in four values surpassing the threshold value. However, it is generally accepted that high VIF values for variables constituting an interaction variable safely can be ignored (Allison, 2012). No other values even remotely close to 10 are observed and, thus, we conclude that the remainder of determinants show no risk of multicollinearity.
5.2 Microeconomic regressions
Table VI shows the regression results for our set of microeconomic specifications, grouped by bulk carriers, FCCs and tankers to isolate segment-specific effects. For firm sizes, we use mega yards and medium owners as base levels to test our hypotheses. Specifications without fixed effects are estimated using ordinary least squares (OLS). One should keep in mind when assessing results that these estimates do not account for time effects or current market conditions.
The basic OLS specifications with only firm- and contract-specific variables result in an R2 of only 17-20 per cent across segments, but this increases to 34-42 per cent once we add shipyard fixed effects, and as much as 60-78 per cent once we add both shipyard and shipowner fixed effects. This indicates that time-invariant unobservable and observable characteristics of both yards and, especially, owners are important in determining contract prices. Heterogeneity across yards could, for instance, be related to specialization premiums, bargaining power or superior ship designs. For owners, it could reflect differing market timing ability, where some owners look at the newbuilding market as a source of asset play profits, while others take a more strategic long-term view of fleet renewal.
There is a positive and highly significant effect of delivery time across segments and specifications. While this is contrary to our expectations, this can here be explained by the fact that we do not account separately for the market cycle. Consequently, as delivery time and market prices are positively correlated (Adland et al., 2006; Adland and Jia, 2015), what this variable picks up is simply the effect of higher newbuilding demand increasing prices and filling up the orderbook, thus lengthening delivery times. The impact of design speed, where significant, is positive as expected. The impact of vessel size (GT) is positive and highly significant for tankers, in accordance with our expectations, but negative for container vessels. This could well be an artefact related to the different impact of vessel size (GT) in the CGT calculations (ref. Table A1 in the Appendix) for the various vessel types.
We note that shipyards of all sizes obtain lower prices on a US$/CGT basis than the mega yards, though apart from this observation the trend with regards to yard size is not clear. This indicates the possible presence of a two-tier market where the behemoth yard groups in the Far East have a degree of pricing power. While an alternative explanation could be that these yards are capable of building vessels towards the more sophisticated end of the spectrum for each vessel type, and that such effects are not properly captured by our US$/CGT price measure, we would expect this to be picked up in part by our vessel size (GT) and speed variables.
Broadly speaking, shipyard experience has the predicted positive impact on prices in the tanker segment, negative for bulkers and mixed for FCC (depending on whether fixed yard effects are included). Keeping in mind that the market cycle is not accounted for in this set of specifications, and knowing that a lot of the new greenfield yards in China were set up to build basic bulkers during the 2003-2008 shipping boom, we suggest that this relationship is simply reflecting the combination of high bulker contracting prices and low/no experience of certain yards at the time. This explanation may also hold for parts of the FCC segment.
In Table VII, our second set of model specifications adds macroeconomic variables, including an interaction variable between freight rates and delivery time that is assessing the impact of high/low rate deviations. Discussions of the two variables separately are irrelevant, as they appear merely to avoid specification errors when we introduce the interaction variable.
The key difference when including benchmark newbuilding prices in the regression is that even the basic OLS specification, with firm size effects only, has a relatively high explanatory power ranging from 55 per cent for tankers to 74 per cent for bulkers. Adding yard fixed effects increases the R2 to the 70-80 per cent range, increasing further to 82-95 per cent for both yard and owner fixed effects. Again, therefore, fixed yard and owner effects play a considerable role in the price formation for individual newbuilding contracts. Moreover, owner heterogeneity remains more important than yard heterogeneity in the explanation of contract price variation.
As expected, both the price benchmark and GDP/capita (as a proxy for wages) show a positive and highly significant relationship with the US$/CGT pricing across specifications and market segments. The coefficients for steel prices, where significant, are also positive as expected. The statistically significant estimates for yard experience are now positive across segments, which supports our earlier explanation that the observed negative values for the microeconomic specifications were caused primarily by a failure to account for the addition of new yard capacity at high price levels. The impact of delivery time is now negative for tankers, as predicted by theory, though still positive for container vessels. We can only speculate that in a market with oligopolistic competition such as the container market, there may be a strategic value of occupying newbuilding slots further out in time in strong markets to prevent competitors from gaining market share.
Where significant, all yard sizes display negative price deviations when compared to mega yards, confirming the earlier findings. However, there are no other clear trends in contracting prices with regards to neither yard size nor owner size. The remainder of our variables either show similar behaviour as in the microeconomic regression (speed and GT) or show mixed results with regards to signs across specifications, making any conclusions difficult.
6. Concluding remarks
In this paper we have proposed and estimated a new model for price formation in the newbuilding market where shipowner and shipyard heterogeneity is accounted for. Using data on individual contracts in the tanker, bulker and container segments, our empirical methodology relies on the estimation of fixed-effect models. Although the newbuilding price benchmarks (market conditions) and GDP/capita (salary costs) are influential covariates, our main conclusion is that shipyards and, particularly, shipowners play an influential role on the US$/CGT price level in individual contracts. Additionally, steel prices and yard experience have a positive impact on prices. Delivery times have the expected negative impact on prices for tankers, but positive for container vessels, which we suggest relate to the possible strategic value of newbuilding slots in an oligopolistic market.
From a practical point of view, our proposed model and methodology can in principle be used to benchmark the performance of individual yards and owners in negotiations for newbuilding contracts. Similarly, knowledge about the effect of yard identity, size and experience on contracting prices can assist shipowners in the selection of which yards to approach.
We acknowledge that the lack of observable contract prices for such a large share of the trading world fleet means we are analysing only a relatively small sample and cannot be sure of unbiasedness. Another limitation of our study is that our empirical framework is dedicated to the measurement of time-invariant owner and yard effects rather than to their explanation. There are also limitations in our data that are difficult to rectify, for instance, the fact that only relatively few contracts per year have a known value.
While shipbuilding productivity almost certainly has increased over the course of our sample, we note that the impact on prices of such macro effects is embedded in our market price index, which reflects the US$/CGT price of a standard vessel built at a “first class” yard at the time. As such, this variable incorporates both an element of technological change and the “size creep” that we can observe in the fleet.
Future research should focus on developing the ideas presented here further, particularly in identifying variables that can shed further light on the sources of price variation due to yard and owner heterogeneity.
Predicted effects on US$/CGT from explanatory variables
|Variable||Unit||Predicted effect on US$/CGT price|
|Delivery time * Freight rate interaction||Negative|
|Newbuilding price index||US$/CGT||Positive|
|GDP per capita||US$||Positive|
Shipyard experience and delivery time (pr. shipyard nation)
|Country||Observation||No. of yards||Yard experience (years)||Delivery times (months)|
Top ten shipyards and owners in the data sample
|Builder yard||Contracts||(%)||Cumulative||Owner (buyer)||Contracts||(%)||Cumulative|
|A. Bulk market (1,354 contracts)|
|Jiangnan SY Group||75||5.54||5.54||COSCO Group||56||4.14||4.14|
|Sinopacific||71||5.24||10.78||China Shipping (H.K.)||54||3.99||8.12|
|Hyundai HI||66||4.87||15.66||Pan Ocean||29||2.14||10.27|
|STX SB||64||4.73||20.38||Eagle Bulk Shipping||27||1.99||12.26|
|Oshima SB Co||64||4.73||25.11||Genco Shipping and Trading||22||1.62||13.88|
|Shanghai Waigaoqiao||55||4.06||29.17||Grieg Star||20||1.48||15.36|
|CSC Jinling Shipyard||48||3.55||32.72||U.Ming Marine Transport||18||1.33||16.69|
|Hyundai Mipo||45||3.32||36.04||Sino Shipping Group||18||1.33||18.02|
|Hudong Zhonghua||43||3.18||39.22||Jinhui Shipping and Transportation||18||1.33||19.35|
|B. FCC market (1,239 contracts)|
|Hyundai HI||287||23.16||23.16||Maersk Company||91||7.34||7.34|
|Samsung HI||189||15.25||38.42||COSCO Group||60||4.84||12.19|
|Daewoo (DSME)||154||12.43||50.85||CSC Group||58||4.68||16.87|
|Hanjin HI (Yeongdo)||70||5.65||62.47||OOCL||45||3.63||24.29|
|Jiangsu New YZJ||48||3.87||70.70||APL||41||3.31||30.91|
|Hyundai Mipo||41||3.31||74.01||Seaspan Corporation||40||3.23||34.14|
|Shanghai Shipyard||39||3.15||77.16||Rickmers Reederei||33||2.66||36.80|
|Jiangnan SY Group||33||2.66||79.82||Hapag.Lloyd Cont||33||2.66||39.47|
|C. Tanker market (1,166 contracts)|
|Hyundai HI||197||16.90||16.90||China Shipping (H.K.)||49||4.20||4.20|
|Hyundai Mipo||154||13.21||30.10||Scorpio Group||45||3.86||8.06|
|Samsung HI||143||12.26||42.37||Teekay Tankers||38||3.26||11.32|
|Dalian Shipbuilding||108||9.26||62.01||Nat Iranian Tanker||25||2.14||16.12|
|STX SB||91||7.80||69.81||TORM A/S||24||2.06||18.18|
|Guangzhou SY Intl||72||6.17||75.99||AET Tanker||23||1.97||20.15|
|SPP Sacheon SY||38||3.26||79.25||Maersk Company||21||1.80||21.96|
|Shanghai Waigaoqiao||33||2.83||82.08||BW Maritime||21||1.80||23.76|
|Brodosplit||28||2.40||84.48||Minerva Ship Mngt||20||1.72||25.47|
Descriptive statistics of macroeconomic variables
|Freight rates (US$/day)|
|Market prices (US$/CGT)|
Variance inflation factor (VIF) test of multicollinearity
|Low freight rate||14.15|
|High freight rates||13.63|
|Delivery time × Low freight rates||12.62|
|Delivery time × High freight rates||15.03|
|GDP per capita||1.230|
Microeconomic regression results
|Constant||354.7 (438.1)||1,357*** (423.0)||1,669*** (563.2)||1,296*** (184.4)||1,574*** (193.7)||1,755*** (339.9)||1,545*** (266.5)||877.1*** (316.2)||560.0 (419.8)|
|GT||0.000143 (0.000483)||0.000120 (0.000611)||0.00178* (0.00102)||−0.00364*** (0.000317)||−0.00345*** (0.000365)||−0.00304*** (0.000429)||0.00430*** (0.000399)||0.00292*** (0.000541)||0.00205*** (0.000681)|
|Speed||107.4*** (31.10)||35.88 (29.11)||−2.160 (38.92)||40.99*** (7.742)||17.57** (7.706)||25.64** (9.977)||7.575 (17.78)||3.916 (19.59)||31.15 (30.42)|
|Delivery time||16.29*** (1.242)||14.66*** (1.343)||8.910*** (2.021)||15.16*** (1.101)||19.61*** (1.128)||18.10*** (1.159)||10.67*** (1.531)||16.12*** (1.772)||12.24*** (2.945)|
|Experience||−1.076 (1.189)||−13.18*** (2.858)||−7.598* (4.567)||3.180*** (1.200)||−6.689*** (2.494)||−2.515 (2.773)||5.148*** (1.224)||8.114*** (2.163)||6.725** (3.295)|
|L-Yard||−93.64** (37.00)||−132.1** (61.25)||−58.90 (65.66)|
|M-Yard||−91.69** (42.98)||−135.7*** (49.45)||−51.20 (49.69)|
|S-Yard||−82.43 (59.56)||−137.5** (65.42)||−262.4*** (76.53)|
|VS-Yard||53.06 (116.7)||−656.3*** (94.42)||−113.9 (112.2)|
|XL-Owner||80.20 (115.9)||7.060 (116.4)||176.8*** (50.13)||177.8*** (44.65)||150.5* (79.50)||170.8** (79.26)|
|VL-Owner||112.6** (43.67)||67.41 (44.22)||85.43** (42.37)||124.7*** (37.06)||−20.29 (46.84)||−26.54 (46.98)|
|L-Owner||−36.38 (39.68)||−40.52 (40.76)||86.99* (48.79)||125.3*** (44.96)||77.70 (47.97)||66.53 (45.24)|
|S-Owner||−130.9*** (46.26)||−140.9*** (48.12)||131.2* (69.60)||135.8** (67.81)||103.6** (49.22)||168.3*** (49.44)|
|VS-Owner||93.26** (46.07)||20.25 (47.66)||184.6* (94.39)||99.81 (76.34)||108.3* (61.52)||137.9** (61.77)|
Robust standard errors in parentheses;
statistically significant at a 10% level;
statistically significant at a 5% level;
statistically significant at a 1% level; XL = extra large, VL = very large, L = large, M = medium, S = small, VS = very small
Macroeconomic regression results
|Constant||−1,671*** (325.5)||−921.2*** (335.9)||152.0 (328.5)||−13.09 (153.8)||−361.3** (163.9)||−365.9 (249.2)||−412.9* (228.6)||−11.86 (242.8)||−170.5 (325.8)|
|GT||−0.000275 (0.000290)||−0.000653* (0.000365)||−0.000249 (0.000477)||−0.000616*** (0.000227)||−0.00104*** (0.000263)||−0.000995*** (0.000296)||0.00347*** (0.000271)||0.00177*** (0.000314)||0.00144*** (0.000431)|
|Speed||97.61*** (23.09)||52.07** (23.09)||−5.723 (20.21)||22.81*** (4.892)||20.01*** (6.175)||29.34*** (8.218)||33.51** (13.63)||9.051 (13.85)||9.820 (22.19)|
|Delivery time||−0.452 (1.233)||1.043 (1.352)||−0.404 (2.144)||3.771*** (0.990)||4.922*** (1.043)||4.940*** (1.146)||−4.639** (2.167)||−3.058* (1.814)||−5.973*** (2.187)|
|L-Rates||−84.42 (86.94)||−56.51 (82.26)||52.43 (103.7)||−166.0** (73.92)||−217.2*** (80.42)||−309.1*** (90.55)||−154.7* (92.01)||−157.2* (86.77)||−262.8*** (100.4)|
|H-Rates||−32.59 (70.72)||16.96 (73.50)||−25.92 (92.87)||−107.5 (83.32)||−123.6 (83.02)||−73.43 (86.53)||−486.3*** (86.08)||−396.3*** (81.44)||−402.9*** (112.3)|
|L-Rates*Delivery time||7.781** (3.254)||4.863* (2.931)||0.461 (3.254)||0.628 (2.391)||3.326 (2.650)||7.804** (3.171)||3.118 (3.029)||2.434 (3.101)||9.307*** (3.535)|
|H-Rates*Delivery time||−3.061* (1.680)||−4.396** (1.739)||−3.195 (2.460)||3.658* (1.944)||3.776* (2.002)||3.065 (2.073)||10.66*** (2.609)||8.085*** (2.442)||10.39*** (3.553)|
|Experience||4.246*** (0.672)||−0.110 (1.456)||−1.365 (1.751)||1.069 (0.824)||3.376** (1.583)||0.965 (1.689)||9.016*** (0.986)||1.598 (1.290)||2.403 (1.931)|
|Index||0.969*** (0.0336)||0.930*** (0.0341)||0.971*** (0.0511)||0.668*** (0.0419)||0.668*** (0.0402)||0.662*** (0.0460)||0.805*** (0.0450)||0.735*** (0.0458)||0.825*** (0.0657)|
|Steel price||0.390*** (0.101)||0.264** (0.106)||−0.0786 (0.187)||0.392** (0.155)||0.220 (0.164)||0.263 (0.193)||0.357** (0.149)||0.0182 (0.135)||−0.00750 (0.202)|
|Oil price||−1.097** (0.476)||−0.799 (0.582)||0.678 (0.968)||−0.403 (0.845)||−0.743 (0.915)||−0.230 (1.056)||−4.719*** (0.812)||−1.422* (0.787)||−0.943 (1.228)|
|GDP/capita||0.00305*** (0.000573)||0.00839*** (0.00176)||0.0106*** (0.00250)||0.00760*** (0.000929)||0.0176*** (0.00285)||0.0159*** (0.00354)||0.00231** (0.000959)||0.0290*** (0.00398)||0.0203*** (0.00614)|
|L-Yard||−38.24* (20.06)||−89.89*** (31.63)||11.87 (45.59)|
|M-Yard||−6.475 (26.61)||−52.42* (31.56)||−44.02 (36.68)|
|S-Yard||12.75 (36.70)||−71.32* (41.25)||−193.2*** (72.62)|
|VS-Yard||−90.80** (38.13)||−303.6*** (68.63)||46.02 (69.27)|
|XL-Owner||−30.64 (62.55)||−8.469 (62.83)||82.11*** (26.84)||113.4*** (26.79)||79.86 (53.23)||27.53 (54.15)|
|VL-Owner||29.70 (22.53)||34.38 (21.76)||56.12** (24.32)||61.34** (23.98)||−51.71 (34.85)||−53.59 (34.53)|
|L-Owner||25.93 (21.88)||34.85* (20.39)||55.96* (30.34)||100.3*** (32.25)||64.42* (36.77)||52.01 (32.90)|
|S-Owner||−7.228 (23.47)||8.805 (24.27)||64.84* (33.61)||111.4*** (33.21)||−25.75 (38.44)||12.40 (36.95)|
|VS-Owner||53.01** (26.52)||39.48 (28.60)||−35.30 (69.39)||−56.77 (60.78)||−42.83 (50.85)||−41.80 (45.50)|
Robust standard errors in parentheses;
statistically significant at a 10 per cent level;
statistically significant at a 5 per cent level;
statistically significant at a 1 per cent level; XL = extra large, VL = very large, L = large, M = medium, S = small, VS = very small
CGT coefficients by vessel type
|General cargo ships||27||0.64|
|Full container carrier||19||0.68|
Source: OECD (2007)
Huber-White Sandwich estimator in STATA.
Adland, R. and Jia, H. (2015), “Shipping market integration: the case of sticky newbuilding prices”, Maritime Economics and Logistics, Vol. 17 No. 4, pp. 389-398.
Adland, R., Cariou, P. and Wolff, F.-C. (2016a), “The influence of charterers and owners on bulk shipping freight rates”, Transportation Research Part E, Vol. 86, pp. 69-82.
Adland, R., Jia, H. and Strandenes, S. (2006), “Asset bubbles in shipping? An analysis of recent history in the drybulk market”, Maritime Economics and Logistics, Vol. 8 No. 3, pp. 223-233.
Adland, R., Alger, H., Banyte, J. and Jia, H. (2016b), “Does fuel efficiency pay? Empirical evidence from the drybulk timecharter market revisited”, Transportation Research Part A, Vol. 95, pp. 1-12.
Agnolucci, P., Smith, T. and Rehmatulla, N. (2014), “Energy efficiency and time charter rates: energy efficiency savings recovered by ship owners in the Panamax market”, Transportation Research Part A, Vol. 66, pp. 173-184.
Akram, Q.F. (2009), “Commodity prices, interest rates and the dollar”, Energy Economics, Vol. 31 No. 6, pp. 838-851.
Alizadeh A.H. and Talley, W.K (2011a), “Vessel and voyage determinants of tanker freight rates and contract times”, Transport Policy, Vol. 18 No. 5, pp. 665-675.
Alizadeh A.H. and Talley, W.K. (2011b), “Microeconomic determinants of dry bulk shipping freight rates and contract times”, Transportation, Vol. 38 No. 3, pp. 561-579.
Alizadeh, A. and Nomikos, N.K. (2009), Shipping Derivatives and Risk Management, Palgrave Macmillan, London.
Allison, P. (2012), “When can you safely ignore multicollinearity”, Statistical Horizons, 10 September.
Beenstock, M. (1985), “A theory of ship prices”, Maritime Policy and Management, Vol. 12 No. 3, pp. 215-225.
Beenstock, M. and Vergottis, A. (1989a), “An econometric model of the world market for dry cargo freight and shipping”, Applied Economics, Vol. 21 No. 3, pp. 339-356.
Beenstock, M. and Vergottis, A. (1989b), “An econometric model of the world tanker market”, Journal of Transport Economics and Policy, Vol. 23 No. 3, pp. 263-280.
Bertram, V. (2003), “Strategic control of productivity and other competitiveness parameters”, Proceedings of the Institution of Mechanical Engineers, Part M: Journal of Engineering for the Maritime Environment, Vol. 217 No. 2, pp. 61-70.
Bloomberg (2016), Historical Prices for HRC1 (Hot Rolled Coilsheet).
Cameron, A.C. and Miller, D.L. (2015), “A practitioner’s guide to cluster-robust inference”, Journal of Human Resources, Vol. 50 No. 2, pp. 317-372.
Clarkson Research (2016), World Shipyard Monitor.
Clarksons (2016a), available at: http://wfr.clarksons.net/ (accessed 10 February 2016).
Clarksons (2016b), available at: https://sin.clarksons.net (accessed 1 April 2016).
Dikos, G. (2004), “New building prices: demand inelastic or perfectly competitive?”, Maritime Economics and Logistics, Vol. 6 No. 4, pp. 312-321.
Fan, L., Koehler, M.M. and Wilson, W.W. (2011), “Intermodalism and new trade flows”, The Blackwell Companion to Maritime Economics, pp. 121-137.
Haddal, R. and Knudsen, K. (1996), “One shipbuilding market?”, SNF-Working Paper 30/96, SNF, Bergen.
International Labour Organization (ILO) (2008), Global Wage Report 2008/09, International Labour Organization.
International Monetary Fund (IMF) (2016), IMF-Primary Commodity Prices, The International Monetary Fund, available at: www.imf.org/external/np/res/commod/External_Data.xls (accessed 1 April 2016).
Jiang, L. and Strandenes, S.P. (2011), Assessing the Cost Competitiveness of China’s Shipbuilding Industry, Syddansk Universitet, Institut for Miljø- og Erhvervsøkonomi.
Jiang, L., Bastiansen, E. and Strandenes, S.P. (2013), “The international competitiveness of China’s shipbuilding industry”, Transportation Research Part E, Vol. 60, pp. 39-48.
Jon, J.S. (2010), “Government policies and the shipbuilding industry”, in Grammenos, C. (Ed.), The Handbook of Maritime Economics and Business, 2nd ed., Lloyd’s List, London.
Köhn, S. and Thanopoulou, H. (2011), “A GAM assessment of quality Premia in the drybulk timecharter market”, Transportation Research Part E, Vol. 47 No. 5, pp. 709-721.
Lizardo, R.A. and Mollick, A.V. (2010), “Oil price fluctuations and US dollar exchange rates”, Energy Economics, Vol. 32 No. 2, pp. 399-408.
Mokia, Z. and Dinwoodie, J. (2002), “Spatial aspects of tanker lay-times”, Journal of Transport Geography, Vol. 10 No. 1, pp. 39-49.
Organisation for Economic Co-operation and Development (OECD) (2007), Compensated Gross Ton (CGT) System, OECD Directorate for Science Technology and Industry.
Porter, M.E. (1979), “How competitive forces shape strategy”, Harvard Business Review, Vol. 57 No. 2, pp. 137-145.
Rodrik, D. (1999), “Democracies pay higher wages”, Quarterly Journal of Economics, Vol. 114 No. 3, pp. 707-738.
Sauerhoff, C. (2014), Competitive Differentiation Within the Shipbuilding Industry: The Importance of Competence in the Field of Services, Springer Gabler, Verlag.
Stopford, M. (2009), Maritime Economics, Routledge, London.
Stott, P. (1995), “Marketing strategy for merchant shipbuilders”, Journal of Ship Production, Vol. 11 No. 3, pp. 149-158.
Strandenes, S. (1984), “Price determination in the time-charter and second-hand markets”, Center for Applied Research, Norwegian School of Economics and Business Administration, Working Paper MU, 6.
Strandenes, S.P. (1986), NORSHIP: A Simulation Model for Bulk Shipping Markets, Norwegian School of Economics and Business Administration.
Strandenes, S. (2010), “Economics of the markets for ships”, in Grammenos, C.T. (Ed.), The Handbook of Maritime Economics and Business, 2nd ed., Lloyd’s List, London.
Tamvakis, M.N. (1995), “An investigation into the existence of a two-tier spot freight market for crude oil tankers”, Maritime Policy and Management, Vol. 22 No. 1, pp. 81-90.
Tamvakis, M.N. and Thanopoulou, H.A. (2000), “Does quality pay? The case of the dry bulk market”, Transportation Research Part E, Vol. 36 No. 4, pp. 297-307.
Tsolakis, S., Cridland, C. and Haralambides, H. (2003), “Econometric modelling of second-hand ship prices”, Maritime Economics and Logistics, Vol. 5 No. 4, pp. 347-377.
Wijnolst, N., Wergeland, T. and Levander, K. (2009), Shipping Innovation, IOS Press, Amsterdam.
Wooldridge, J.M. (2013), Introductory Econometrics: A Modern Approach, South-Western Cengage Learning, Mason, OH.
World Bank (2016a), Commodity Markets, World Bank, available at: http://pubdocs.worldbank.org/pubdocs/publicdoc/2016/5/148131457048917308/CMO-Historical-Data-Monthly.xlsx (accessed 1 April 2016).
World Bank (2016b), World Developments Indicator Database, World Bank, available at: http://data.worldbank.org/data-catalog/world-development-indicators (accessed 1 April 2016).
Zannetos, Z.S. (1966), The Theory of Oil Tankship Rates: An Economic Analysis of Tankship Operations, MIT Press, Cambridge, MA.
Jiang, L. and Lauridsen, J.T. (2012), “Price formation of drybulk carriers in the Chinese shipbuilding industry”, Maritime Policy and Management, Vol. 39 No. 3, pp. 339-351.