«Price Formation of Dry Bulk Carriers in the Chinese Shipbuilding Industry Liping Jiang December 2010 © University of Southern Denmark, Esbjerg and ...»
One method of removing such multicollinearity in regression is to use Principal Component Regression (PCR) approach (Jolliffe 2002). PCR actually is the combination of Principal Component Analysis (PCA) and multiple linear regression techniques. The central idea of PCA is to transform a number of possibly correlated variables into a smaller number of uncorrelated principal components. Small amount of components can be extracted and contain the most of the information of original data (Jolliffe 2002). The component Ci is given by Ci i1 * Z ( X 1 ) i 2 * Z ( X 2 ) ik * Z ( X k )(i 1, k ) (1) Where Ci stands for the i th principal component, Z ( X k ) stands for the k th standardized independent variable, ik stands for the component score coefficient of k th standardized independent variable on i th principal component.
5.2. Principal Component Regression
By using PCA, the extracted components become ideal to use as predictors in a regression equation since they have optimized spatial patterns and removed the possible multicollinearity (Al-Alawi 2007). Thus, Principal Component Regression can be expressed as, Z (Y ) 1C1 2C2 i Ci (i 1, p k ) (2) Where Z (Y ) is the standardized dependent variable, Ci stands for the i th principal component, i is the i th standardized coefficient of the standardized principal component regression equation. The component equation (1) will be applied to the equation (2) and then the standardized linear regression equation will be yielded after sorting out all the X i variables. Thus, Z (Y ) b1 * Z ( X 1 ) b2 * Z ( X 2 ) bk * Z ( X k ) (3) Where, bk is the partial regression coefficient of principal component regression equation and then bk has to be changed to unstandardized coefficient (Liu et al.
6. Results6.1. Principal Component Analysis
SPSS 16.0 is used to run the Principal Component Analysis first and then the Principal Component Regression.
In our case, the KMO test result is 0.544 which normally should be larger than 0.6. It indicates there are some limits by using PCA to fully represent our data and it will be improved in the future study. The PCA involves several areas as follows.
First, five components are extracted from independent variables. Then varimax rotation technique is used to maximize the loading of a variable on one component and to produce a ranked series of factors (Al-Alawi et al. 2005). We select the first three as principal components which together account for 95.838% of total variance. The Table 1 reports the result of component extraction, varimax rotation and component score coefficient.
6.2. Principal Component Regression The three principal components subsequently used as independent variables in a multiple regression model. Furthermore, three dummies are included to examine the impact of the dry bulk carrier’s vessel type. Since all dummies have to be jointly included or exclude in the model, ‘enter’ regression method is chosen for the dummy group and ‘stepwise’ method is chosen for the component group. The PCR result is summarized in Table 2.
The hypotheses are confirmed by the significantly positive signs of five coefficients. According to the standardized coefficient model, it indicates the most important role time charter rate plays in determining the dry bulk carrier prices, followed by shipbuilding cost index, price-cost margin, shipbuilding capacity utilization and credit rate in decreasing importance. For example, a 10% increase in time charter rate will make price of dry bulk carrier rise by 4.07%.
This is in accordance with the reality that shipbuilding and shipping markets are tightly connected. For dry bulk carrier in particular, values of cargo transportation are lower than those of tankers and containers. Therefore, it is assumed that higher the time charter rate for dry bulk carrier, the higher return on investment for ship owners, and as a result, ship owners will be more willing to invest in dry bulk carrier with higher prices. For large shipyards that are able to switch their production from one vessel type to another, the building of a dry bulk carrier is cheaper and hence is a shipyard’s last resort in its strive to maximize revenue (Haralambides et al. 2005). The prices of dry bulk carrier may be able to depend on the market demand to a greater degree.
Shipbuilding cost index, constructed by three major cost components in the Chinese shipbuilding industry, is found to be the second most important factor in determining the prices of dry bulk carrier. This result is the further proof of the fact that world shipbuilding centers have always been relocated to the lower cost countries.
The price cost margin impact on the price level of dry bulk carrier to a moderate degree. The answer could be that China is becoming a major shipbuilding nation for dry bulk carrier in recent years and has held more than 37% market share in terms of new orders since 2006. With regard to China’s big market power in dry bulk sector, market competitions from other shipbuilding nations have a certain degree of influence on the Chinese prices.
The shipbuilding capacity utilization is found to have a significant but smaller effect on price in our study. One reason is that capacity utilization has been keeping as high as 93% to 100% since the start of modern Chinese shipbuilding industry. Another reason is due to the limitation of our dataset mentioned before. Most of the data concentrate on year 2006-2008 during which was the market prosperity and therefore capacity utilization kept extremely high, even 100%. The fluctuation of ship export credit rate has a significant but the least effect. It confirms that the Chinese shipbuilding industry has less relied on the government subsidies after transforming from a central planning system to a free market launched by the COSTIND (State Commission of Science and Technology for National Defense Industry) in 1999. Major shipyards are prompted to be more market-oriented and to take part in the world competition by improving technology and quality. At the same time, all coefficients of dummy variables are significant and prove the systematically different prices of four types of dry bulk carrier. The bigger ship has a higher price level and the intercept for Capesize, Panamax, Handymax and Handysize in unstandardized coefficient models are 21.08, 8.952, 4.097 and 1.111 respectively.
Overall, the time charter rate and shipbuilding cost have the greatest effect of all variables on the determination of dry bulk carrier price. It is argued in the previous studies (Haralambides et al. 2005; Jin 1993) that shipbuilding is a supply- and cost-driven industry instead of demand driven and cost is a decisive factor of world newbuilding prices. Our results indicate that time charter rate has a higher impact on China’s dry bulk prices than the shipbuilding cost. A reason behind this may be the ‘China Factor’. Under the strong economy growth, China becomes one of the world’s largest energy consumers and has greatest influence on the bulk trade. The demand for iron ore promotes the dry bulk shipping market dramatically. In the meanwhile, China has become a net coal importer since 2009. This not only increases the seaborne trade for Chinese domestic use, but also forces Japan and South Korea, who imported coal from China before, to turn to farther distance Australia to quest for resources. On a global scale, ‘China Factor’ gives a great push to the freight rate boost in bulk shipping market, and consequently stimulates ship owner’s desire to expand the dry bulk fleet and brings up the prices of dry bulk carrier. Therefore time charter rate plays a greater role than shipbuilding cost in determining the prices of Chinese dry bulk carrier. What else can be implied from model is that international competition in shipbuilding industry is more closely tied to the dry bulk carrier prices than other two domestic factors. Therefore, changes of nationally industrial policy or capacity utilization may not affect the price of Chinese dry bulk carrier as much as it would in the case of competition degree in world shipbuilding.
6.3. Simulations and Discussions
Previously we test the price formation of dry bulk carrier. In this section, the model is used to carry out few simulations in order to understand the likely price behaviour under various exogenous variable changes. Here we choose time charter rate and shipbuilding cost index as the impact variables and fix all other variables in simulations. It is because time charter rate is the most important price determinant and shipbuilding cost is the core competence of the Chinese shipyards. The reference year chosen is 2009 and twelve scenarios are tested against the baseline scenario in Table 3. Because of the American financial crisis, the time charter rate in year 2009 was already at historically low level. We now assume that time charter rate will change at four different levels, namely -50%, 50%, 100% and 150%, compared to 2009. Chinese shipbuilding industry is facing with heavy pressure about rising cost, for this reason cost index will change at 0, 50%,100% levels compared to 2009. Numbers given below are percentage changes over year 2009 for time charter rate, cost index and prices.
There is clearly a large increase in prices for all types under S2 to S12 when compared to baseline scenario. In S1 to S3, time charter rate levels are the same but different levels of cost increase makes significant changes to prices. The higher the cost is, the higher the price change is. All price changes in S4 to S6, S7 to S9 and S10 to S12 share the same characteristics. Under the same cost increase, for example S2, S5, S8 and S11, shipbuilding prices will continue to rise if there are favourable conditions in time charter rate. In most of scenarios, expect for four, smaller ships will have larger price changes. But there is hardly any difference of price changes for all types in S1, S4, S7 and S10 under which only time charter rate changes compared to baseline scenario. We also notice that there is a big price gap between S3 and S4, S6 and S7, and S9 and S10 for all types. It means lower time charter rates combined with higher costs can also lead to higher prices.
7. Conclusions In this study, we propose a price formation model for dry bulk carrier in the Chinese shipbuilding industry, thus filling a gap in the literature of world shipbuilding prices. We utilize the actual vessel prices quoted by major Chinese shipyards and propose a new approach PCR for price modelling. In particular, the theoretical relationship between price and determinants is discussed by including generic market factors as well as Chinese elements. Several conclusions have been drawn. First, time charter rate is the most important determinants.
While increase in shipbuilding cost, price-cost margins, shipbuilding capacity utilization and credit rate have positive effects in decreasing importance. Despite the common perception that shipbuilding industry is cost driven, we conclude the highest impact of time charter rate on the Chinese price of dry bulk carrier attributes to the ‘China Factor’. The competitive indicator we constructed also influences the price moderately. The dynamic price simulations indicate that there is a likely big increase in the prices as a result of growth in time charter rate and shipbuilding cost. Particularly, smaller vessels tend to have larger price increase under same condition. Second, PCR approach is proved to be a good way of overcoming the problem of multicollinearity in maritime economic field. Dummies in the model successfully serve to distinguish the systematic difference of vessel types. Third, investors and policy makers in shipbuilding industry can benefit from applying the model when making decisions. It also has implications for emerging shipbuilding nations who start development from dry bulk carrier sector and will enter the arena of major shipbuilding nations.
8. References Al-Alawi, S.M., Abdul-Wahab, S.A. and Bakheit, C.S. (2008). Combining principal component regression and artificial neural networks for more accurate predictions of ground-level ozone. Enviromental Modelling & Software 23, pp.