Abstract
Purpose
Bulk shipping mostly facilitates the smooth flow of raw materials around the globe. Regardless, forecasting a bulk shipbuilding orderbook is a seldom researched domain in the academic arena. This study aims to pioneer an econophysics approach coupled with an autoregressive data analysis technique for bulk shipbuilding order forecasting.
Design/methodology/approach
By offering an innovative forecasting method, this study provides a comprehensive but straightforward econophysics approach to forecast new shipbuilding order of bulk carrier. The model has been evaluated through autoregressive integrated moving average analysis, and the outcome indicates a relatively stable good fit.
Findings
The outcomes of the econophysics model indicate a relatively stable good fit. Although relevant maritime data and its quality need to be improved, the flexibility in refining the predictive variables ensure the robustness of this econophysics-based forecasting model.
Originality/value
By offering an innovative forecasting method, this study provides a comprehensive but straightforward econophysics approach to forecast new shipbuilding order of bulk carrier. The research result helps shipping investors make decision in a capital-intensive and uncertainty-prone environment.
Keywords
Citation
Sakalayen, Q.M.H., Duru, O. and Hirata, E. (2021), "An econophysics approach to forecast bulk shipbuilding orderbook: an application of Newton’s law of gravitation", Maritime Business Review, Vol. 6 No. 3, pp. 234-255. https://doi.org/10.1108/MABR-03-2020-0019
Publisher
:Emerald Publishing Limited
Copyright © 2020, Pacific Star Group Education Foundation. Licensed re-use rights only.
1. Introduction
The new bulk shipbuilding order injects fluidity in the bulk shipping market, provides insight into the global economy, and remains remains an important indicator to decision making in shipping investments. The research on newbuilding order forecasting in the academic arena is scarce. Commercial newbuilding order forecasts are also either inaccessible and unclear or complex with outcomes that frequently go off beam. The absence of effective but flexible forecasting methodology that could manage the intricacy of the newbuilding domain seems the plausible cause of research scarcity in bulk shipbuilding order forecasting. The earlier research that addresses the bulk shipbuilding order forecasting advocate supply and demand model (Nielsen et al., 1982), ARIMA-based fleet renewal decision model (Yang et al., 2019), system dynamics model (Wada et al., 2018), time-series analysis (Stopford, 2001; Chen et al., 2014), vessel-based logit model (Alizadeh et al., 2016) and judgmental approach (Duru and Yoshida, 2009; Ariel, 1989). These models reflect on the high volatility of the complex bulk shipbuilding orderbook with a small number of parameters. Consequently, the outcomes of the forecasting models either go astray or remain less explainable. Econophysics approach of modelling can be useful with the flexibility to address this intricacy of the bulk shipbuilding orderbook forecasting.
Econophysics is a hybrid field that houses the strengths of both economics and physics and creates a bridge over a volatile and complex scenario (Chen and Li, 2012; Mantegna, 2016). It is a marriage between social and physical sciences (Schinckus, 2010) that provides a completely new avenue to address multifarious business environments, where careful assumptions of economics and empirical trends of physics resonate together providing meaningful elucidation. Though influenced by diverse constructs and conjectures, shipping is yet to embrace the econophysics approach of inference.
However, there are successful instances of applying econophysics approach in fields that include business volatility and stock markets, economic value and growth, economic and financial time series, behavioural finance, corporation financial stability, distribution and interactions of economic entities, market structure and financial risks (Chen and Li, 2012; Chakraborti et al., 2011; Huang, 2015; Guedes et al., 2019; Schinckus and Jovanovic, 2013; Zapart, 2015; McCauley, 2004; Meng et al., 2016; Rickles, 2007; Zhong et al., 2019). Mainly the concepts of physics such as Bernoulli’s equation, Newton’s law of gravitation, Brownian motion, Schrodinger equation, Bose-Einstein distribution, Gaussian function, Fourier transformation, and Heisenberg’s uncertainty principle have been adopted to naturalise the econophysics models (Donmez and Sen, 2018; Meng et al., 2016; Zhang and Huang, 2010; Cotfas, 2013; Pedram, 2012; Mantegna, 2016; Kusmartsev, 2011; Agustini et al., 2018; Hsu, 2010; Wang and Pei, 2015). This study illustrates a flexible econophysics model that can manage numerous relevant constructs of bulk shipbuilding order forecasting by using Newton’s law of gravity.
Previously, Newton’s law of gravity has been used in international trade to model bilateral trading flow and efficiency (Tinbergen, 1962; Abidin et al., 2013; Bialynicka-Birula, 2015), and also in the regional study to investigate regional integration effect (Darku, 2009). In this study, Newton’s law of gravitation has been naturalised for a bulk shipbuilding order forecasting model by developing predictors or variables with concerned constructs informed through the literature review on trends and factors of shipbuilding. Not all factors have been accommodated into the predictors because of either the relational distance with predictors or the non-availability of data. The developed model has been evaluated and trained through the auto-regressive integrated moving average (ARIMA) analysis technique of the SPSS software.
In the rest of the paper, Section 2 presents the literature review and sets the ground for developing predictors. Section 3 elaborates on the econophysics methodology. Section 4 illustrates data analysis and discusses the bulk shipbuilding forecast outcomes of this study. Section 5 is the conclusion.
2. Literature review
2.1 Trends of shipbuilding industry
Booms and bursts are common in the shipbuilding industry. China, South Korea and Japan are now the leading shipbuilding nations in the world. In each shipbuilding segment, these countries are competing with each other trying to optimise on their shipbuilding capabilities and deliveries. The bulk shipbuilding dominates in the newbuilding orderbook. Bulk shipping mainly serves the global resources sector; as a result, the bulk newbuilding order indicates the global production and economic dynamism. Figures 1 and 2 present segment-wise newbuilding order trends and newbuilding deliveries of the leading shipbuilding nations respectively.
The bulk shipping freight task is a dominating part in the growth of international seaborne freight transportation task (in tonne-km) that indicates the demand-side of shipping capacity (UNCTAD, 2018a). The global economic growth affects trade. Trade derives international seaborne-freight transportation task, where bulk shipping remains critical. However, the evolution of bulk fleet capacity presents the supply-side that is determined by subtracting the demolition figure from the summation of existing fleet and new delivery figures. Therefore, ship demolition and new delivery dynamics become crucial in meeting the seaborne-trade derived freight-transportation demand.
Capacity oversupply also persists in shipping industry. Figure 3 illustrates the global shipbuilding delivery, demolition and seaborne-trade trends. One of the consequences of shipping capacity oversupply is the low shipping freight-rate that mainly fluctuates based on the demand of international seaborne-freight transportation task and the supply of the vessel capacity. Nevertheless, the prospect of favourable future freight-rate drives the newbuilding order (Stopford, 2009).
The growth of shipbuilding is also explainable through newbuilding ship price and scrap price. The newbuilding prices in the dry bulk sector are related to the current and expected freight rates and global economic growth (Beenstock, 1985). Firstly, the growth of world trade and heavy industries boost newbuilding market (Imai, 2008); secondly, the uncertainty in global economic growth and freight-rate volatility affect the shipbuilding market; finally, the adoption of systematised technology and innovation provide leverage to newbuilding (Lim et al., 2017). The factors concerning these drivers require a closer look for shipbuilding order forecasting study.
2.2 Factors impacting shipbuilding
Shipbuilding order forecasting is a complex, cumbersome and less researched area. It is because the growth of the industry is nonlinear and is affected by various volatile and exogenous factors such as international political events and the strategy of shipbuilding nations. The confidential characteristics of the shipbuilding market are another reason for limited research (Charemza and Gronicki, 1981). However, determining an appropriate timing for investment in new shipbuilding is regarded as a mysterious quality (Goulielmos and Psifia, 2009), where forecasting in the short-term can play a critical role.
Historically, international politics and implementation of international regulations affect the shipbuilding industry. In the past, the shipbuilding industry was influenced by political intervention, pre-war demand, post-war redevelopment due to wartime losses, Suez Canal closure, shortage of shipbuilding capacity (dockyard shortage), government credit and incentive schemes, continuous recession (1920-1942 period), 1980s depressions etc. (Stopford, 2009). The governments’ protective views and over-optimism bring in subsidies that also distort the shipbuilding industry frequently. Ship-owners’ unconstructive attitudes towards global trade growth cause over-supply that eventually impact shipbuilding industry (Stopford, 1987; Chou and Chang, 2004). In the present time, economic competitiveness, new management system innovative technology become critical for newbuilding order attraction (Zheng et al., 2013; Vishnevskiy et al., 2017; Jha, 2016). From ship’s operational perspective, the ship speed has an impact on newbuilding; for instance, a slow speed generates more supply requirement for the same quantity of seaborne-freight transportation task and vice-versa. Speed reduction also assists balancing oversupply rather than accelerating newbuilding order. However, the ultimate competition in newbuilding largely depends on the level of demand and supply volatility, shipbuilding price, ship-owners’ attitudes towards demand volatility, and shipbuilding nations’ strategy (Shin and Lim, 2014).
A consistent shipbuilding decision requires a well-thought forward planning to meet a situation where a continuous demand may exist. Kavussanos and Alizadeh (2002) illustrate that current market rates and long-term equilibrium rates influence newbuilding, particularly in the dry-bulk market. Irrespective of ship-sizes and freight-rate volatility, influential factors for newbuilding order include exchange rate volatility, shipyard capacity change, and shipbuilding cost volatility (Dai et al., 2015). However, the quality data on these factors are either not available or inaccessible. Existing data quality is improving though (Halim et al., 2019). For bulk shipbuilding order forecasting model, it is important to focus on the balancing factors of demand and supply sides of the market that forms the core part of the newbuilding activity (Stopford, 2009). In other words, global economic growth, freight-rate volatility and systematised technology play a crucial role in shipbuilding decisions for ship-owners. Adopting new technology in shipbuilding decision is leveraged through the demolition market. For demolition, the ships of age 20 years and above are always regarded as best candidates (Stopford and Barton, 1986; BRS Group, 2019). The scrap price leverages the newbuilding decision. Hence, the ship demolition market appears as a balancing element between the supply and demand that ultimately drives the market equilibrium and freight-rate level (Karlis and Polemis, 2016). The newbuilding price index affects the shipbuilding orders. The banking system that provides a letter of credit for shipbuilding also focuses on the newbuilding price index. Hence, the ratio of scrap price to newbuilding price index is directly related to the newbuilding order. The higher the ratio the higher the probability of newbuilding order. Various challenges in shipbuilding also form an “uncertain part” that always looms in the forecasting of newbuilding order. In econophysics, uncertainty is mostly dealt with a coherent manner appreciating other factors and quality data on the factors. The uncertainty reduces also with the improvement of the models (Chen, 2017; Dionisio et al., 2006; Schinckus, 2009). As such, applying econophysics approach to forecasting not only adds value as a pioneering tool to shipbuilding forecasting but also provides an insight on the “uncertainty part” in the newbuilding order forecasting.
Overall, the force that drives the shipbuilding order forms through the constructs such as freight rate, international seaborne freight task, scrap price, newbuilding price index, the share of scrapable ships of age 20 and above, existing fleet, new ship delivery and ship demolition. This forms a reasonable set of factors or constructs to develop the predictors of this study; other factors have not been accommodated into the predictors mainly either due to the relational distance to predictors or non-availability of quality data. In context of the above discussion, the next section illustrates the methodology of this research.
3. Econophysics methodology
The application of the theories of physics in economics or business is rapidly evolving and proves efficient in many complex market scenario (Jovanovic and Schinckus, 2016; McCauley, 2004; Schinckus and Jovanovic, 2013; Zapart, 2015). The new shipbuilding market is a complex dynamics where shipbuilding order is gravitated by the force between two critical masses: the “existing shipping market prospect” and ship’s age-adjusted “future shipping market prospect” (Stopford, 2009; Bruce and Garrard, 1999; UNCTAD, 2018a, Steidl et al., 2018). This force can indicate the thrust for new shipbuilding order that resembles the force of gravity in physics. With this notion, this study assumes an econophysics approach of forecasting new bulk shipbuilding order by adopting and naturalising Newton’s law of gravitation [equation (1)]. Figure 4 presents a logical framework of this methodology.
The developed model is then investigated within a data set between the years 2008 and 2017. This period is chosen as all required data of the variables and constructs (that form the independent variables for the econophysics model) are available and easily accessible. The easily accessible data enhances the convenience of utilizing the model. However, the multivariate ARIMA analysis is adopted to examine the forecasting model. The multivariate ARIMA has been chosen as it captures a set of various standard temporal constructs in the time series data (Gujarati, 2003; Tabachnick and Fidell, 2007; Duke University, 2018) which is the case for this shipbuilding order forecasting through econophysics approach. In other words, the ARIMA statistical analysis technique well suits with the complexity of the new shipbuilding market. The rest of this section depicts the development of the model and illustrates the data.
Newton’s law of gravitation in physics is expressed as follows:
F → denotes the force between the two objects M1 and M2;
M1 → indicates the mass of one object;
M2 → indicates the mass of the other object;
d → refers to the distance between the two objects; and
G → indicates the gravitational constant.
Resembling Newton’s law of gravitation, the force for the new bulk shipbuilding order is framed as follows:
Bfr → indicates the existing bulk shipping market prospect;
Bfp → indicates the future bulk shipping market prospect;
SbOB → refers to the force between Bfr and Bfp and indicates the new shipbuilding order;
dB → refers to the distance between Bfr and Bfp indicated by the fleet evolution; and
UB → indicates the uncertainty constant for the concerned newbuilding shipping market.
For simplicity, replacing dB2 by dBsq and assuming the value of UB in equation (3) as 1 (binary one) to recognise the presence of uncertainty (which later will be evolved as the unexplained variance of the dependent variable during ARIMA analysis) equation (3) is expressed as follows:
The orientation and constructs of the dependent variable (SbOB) and relevant independent variables (Bfr, Bfp and dBsq) of this econophysics model for bulk shipbuilding forecasting in equation (3) have been further elaborated in Table 1.
Based on the proposed variables and constructs of the variables presented in Table 1, the equation (3) finally takes the following overarching form:
where:
BBDI → indicates bulk freight rate;
BIST → indicates the “main bulk” international seaborne freight task;
BTWA → indicates the share of bulk carriers of age 20 and above;
BSCRP → indicates bulk carrier average scrap price;
BNBPI → indicates bulk carrier newbuilding price index;
BFLT → indicates the bulk carrier existing fleet;
BNDEL → indicates bulk carrier new delivery; and
BDEMO → indicates bulk carrier demolition.
In this proposed model, the relationships of the dependent variable, SbOB (the new bulk shipbuilding order) with the independent variables (Bfr, Bfp and dBsq) play a significant role in naturalizing with the econophysics approach (in this case, the use of Newton’s law of gravitation). Further elaborating on this naturalising aspect, it is to state that the new bulk shipbuilding order sits at the crossroad between the demand and supply. In other words, it balances two loops such as the supply-side capacity adjustment loop and the demand-side capacity utilisation loop (Randers and Göluke, 2007). The increase in shipping capacity (supply or fleet evolution) can decrease capacity utilisation efforts or demand meeting efforts (in this case, new shipbuilding order). The fleet evolution also creates the distance (dB) between the “existing bulk shipping market prospect” and “future bulk shipping market prospect”. Hence, inversely proportional relation exists between the fleet evolution [i.e. (existing ships + new deliveries – ship demolitions), later squared and expressed as dBsq] and the new bulk shipbuilding order (SbOB).
On one hand, the conjugate momentum of ascending or descending freight rate and international seaborne freight task increases or decreases the new shipbuilding order prospect respectively (DSF, 2018; Stopford, 2009). The international seaborne freight task is a direct reflection of the global economic growth in maritime business, while the global economic growth increases global trade and in-turn the international seaborne freight task, which is a derived demand of the trade (UNCTAD, 2018b). Hence, the “existing shipping market prospect (Bfr)” as a product of the “freight rate” (in this case the BDI) and “international seaborne freight task” is directly proportional to the new shipbuilding order (SbOB). On the other hand, new shipbuilding ordering activity not only focuses on future shipping market prospect but also regarded as a replenish activity of ship demolition. The older vessels are first considered for ship demolition. In any ship category, the percentage share of the ships of “age twenty and above” is sensitive to demolition particularly in the context of emissions reduction related emerging technology adoption possibility. However, this age range (twenty and above) covers greater and nearly exhaustive list of probable bulk scrap candidates. Besides this, the right time for demolition is when the scrap price is higher. Hence, the “ratio between the average scrap price and new shipbuilding price indices” with a multiplier of “twenty and above age ships” captures the “prospect of future new bulk shipbuilding market (Bfp)” and is directly proportional to the new shipbuilding order (SbOB).
However, forecasting new shipbuilding order is complex and involves constructs that can be significantly influenced by the behaviour of the market players. This attitude or behaviour is sometimes unexpected and not predictable (Stopford, 2009) and may arise because of different viewpoints of the market players on international regulatory, business and operational environment. Therefore, the unpredictable part of the new shipbuilding order forecasting may appear as an unexplained variance of the dependent variable. The data for this study is collected from various sources such as the Clarkson data, the UNCTAD’s reports and database, OECD reports and the reports of the French-based brokers house the Barry Rogliano Salles (BRS) Group. The data of Clarkson, UNCTAD, OECD and BRS have been used because they are very much maritime focused and the data references are adjusted every year based on the availability of actual data of immediate previous years rather than depending on estimated data. The UNCTAD, OECD and BRS data and their analyses are also in most cases based on Clarkson database that in-turn ensures homogeneity of the data sources and enhances the generalizability and acceptability of the data.
The data for this study has been gathered over the period from 2008 to 2017. This period (2008-2017) has been chosen because of the optimal availability of quality data on all required constructs of the models for bulk carrier newbuilding forecasting. Though yearly data is prevalent and easily accessible rather than monthly data, monthly data is more essential to comprehend the insights of the market dynamics through statistical analysis. Based on the availability, the monthly data collection has been prioritised and later in case of absence of monthly data the yearly data has been interpolated over the selected period in various ways based on the characteristics of the data. Because of the interpolation of the yearly data, which is a frequent practice in forecasting (Zhao et al., 2019; Tokumitsu et al., 2015) and even in simulation and data generation (Li et al., 2020), the outliers in the data set have been eroded automatically before starting the analysis. Aligning or removing outliers as a logical pre-processing step of data analysis brings in benefits by avoiding a few bad apples that may spoil the entire bushel (Cant and Xu, 2020; Shah and Patil, 2019; Lyutikova and Shmatova, 2020).
The monthly data have been used finally for the ARIMA analysis for bulk new shipbuilding order forecasting. The ARIMA analysis model has been used as it has the capability of explaining a given irregularly patterned time series based on its own lags (Ohyver and Pudjihastuti, 2018; Duan and Zhang, 2020), which is the case for several time series used in this study. Due to the vastness of the monthly data, the yearly data have been presented in Table 2 to provide a glimpse of the data requirement of this study. Usually, the data are demonstrated in different units in different reports. For the sake of analysis and to provide an even footing for the model, the data in this study have been either synchronised in similar units or created indices through normalisation, taking logarithm or converted in percentage share. The data shows a moderate cyclical nature of about 36 months or 3 years that has been declared during preparing the ARIMA analysis technique through the SPSS.
4. Data analysis and discussion
In this multivariate ARIMA analysis, there are three predictors or independent variables (Bfr, Bfp and dBsq) for one dependent or outcome variable [the new bulk shipbuilding order (SbOB)]. During fitting the analysis model for multivariate ARIMA through the SPSS software, the econophysics model of this study (see equation (4) has been ensured by assigning the variables as numerators (for Bfr and Bfp) and denominator (for dBsq). The analysis has been completed in four stages such as 1) Checking data assumption and identification of models, 2) model estimation and selection, 3) diagnosis and validation 4) forecasting and discussion.
4.1 Checking data assumption and identification of models
Before data analysis, the normality and stationarity of the data have been checked and found that log data of the variables are required to ensure normality (Figure 5 and Table 3). Figure 5 shows a strong dependence of variability of the bulk shipbuilding order and indicates the necessity of data transformation, for example, logarithmic transformation to stabilise the variance. Table 4 depicts the detail normality tests of the variables.
As an assumption of the ARIMA analysis, the stationarity of the dependent variable needs to be ensured (Gujarati, 2003; Tabachnick and Fidell, 2007), which has been taken care of in this study by taking the first difference of the dependent variable (SbOB) during performing the ARIMA analysis through the SPSS (Statistical Package for the Social Sciences) software version 25. The requirement of taking the 1st difference of the dependent variable during ARIMA analysis of this study (SbOB) is also evident from the descriptive statistics as shown in Table 4.
As a common rule for the non-normal actual dependent variable, it is important to ensure more than twenty records for each independent variable (Tabachnick and Fidell, 2007). In this study, five years (2008-2012) monthly data have been used as a training data set. Later, eight years of monthly data (2008-2015) have been used as the testing data set. This testing data set has been emphasised for model selection and estimation as it ensures adequate data points such as 96 records for each series. The monthly data from 2016-2017 is used during the validation stage. Finally, ten years of monthly data (2008-2017) have used for predicted values and future forecasting. The detail of the ARIMA analysis in four stages is provided below.
4.2 Model estimation and selection
The autocorrelation function (ACF) and partial autocorrelation function (PACF) have been generated on the testing data set to estimate the tentative ARIMA models. Figure 6 shows the correlograms of ACF and PACF of the log transformed bulk shipbuilding orders (Log of SbOB) at the first non-seasonal differencing. Although the Lag spikes are within the 95% confidence level, some spikes are still showing that significant information is contained and there is a tendency of correlations.
Based on the significant closest lag spikes both in the ACF and PACF correlograms (i.e. Lag1,3,9 and 12), the ARIMA models (1,1,3), (3,1,1), (1,1,9), (9,1,1), (3,1,9) (9,1,3), (12,1,1), (12,1,3), (12,1,9) and (12,1,12) are selected tentatively and run. As usual, the distance lag spikes are deliberately ignored initially (such as Lag24) to avoid model over-fitting but later considered to achieve a stingy or parsimonious model. The outcome criteria (indicated in column 1 in Table 5) of the tentative models are summarised in Table 5. The expectations of the criteria values of the models are indicated in Column 2 in Table 5. Mainly the values on the criteria such as the Stationary R-squared (the indicator of the variance captured by the dependent variable), Normalized BIC (Bayesian Information Criteria) Index (designed to choose between models as its lower value is better) Mean Absolute Error (MAE, the lower value is better) are used to select a model.
The (12,1,1) ARIMA model has been selected as most appropriate to perform the diagnostic. This model has been selected as appropriate because it meets all expectation (column 2 in Table 5) particularly the normalized-BIC is the lowest with significance level 0.068 (>0.05).
4.3 Diagnosis and validation
For diagnostic purpose, the residual analysis of the selected appropriate model (12,1,1) has been performed and shown as correlograms in Figure 7. The correlograms on noise residuals are decaying within the 95% confidence level; however, lag24 indicates that there may be some information remaining to capture. From the initial correlograms on the Log of SbOB (Figure 6), it is also evident that the distant lag24 both in ACF and PACF may have potential to capture the remaining information which ultimately necessitates adjusting for the most appropriate model by investigating additional models with lag24. Emphasising on the lag24, the additional ARIMA models (12,1,24), (24,1,12), (24,1,9), (24,1,3) and (24,1,1) are selected for further diagnostics. This process ultimately ensures avoiding over-fitting and reaching to a most parsimonious or stingy model. Running these additional models and comparing the criteria values with the appropriate model selected earlier (i.e. model 12,1,1) provide the best model.
As presented in Table 6, the further diagnostic outcomes show that (12,1,1) is still the best model because of its generated Ljung-Box statistics with significance level >0.05 (Sen et al., 2016) and its lowest normalized-BIC value. The other additional models mainly could not generate the required Ljung-Box statistics. The best model (12,1,1) is then run on a different set of data (Monthly data of 2016–2017) for the validation.
The model statistics and a summary of the validation run are shown in Tables 7 and 8. The model captured about 65.6% variance of the dependent variable through the three predictor variables. Therefore, the uncertainty appeared as the unexplained variance of about 34.4%. The Ljung-Box stat is reasonable with the significance level >0.05. The ACF and PACF of the noise residual from the dependent variable are also decaying (or flat) within the 95% confidence level as expected for a good model (shown earlier in Figure 7).
As the model (12,1,1) well-performed in the validation stage, this model is then used for bulk shipbuilding forecasting with the ten years’ monthly data (2008-2017).
4.4 Forecasting and discussion
At the forecasting stage, the model (12,1,1) captures 73.3% variance of the dependent variable through the three predictor (independent) variables in the study. The Ljung-Box statistics is also good with the significance level >0.05, as demonstrated by other empirical study (Sen et al., 2016). The model statistics and summary of the forecasting run are shown in Tables 9 and 10.
Overall, the analysis of this study presents a good fit for an ARIMA model. The ARIMA (12,1,1) model performs well for forecasting the new bulk shipbuilding order (Figure 8). The three independent variables formed by using the econophysics approach (i.e. the law of gravitation) support capturing about 65-73% variance of the dependent variable (i.e. the new bulk shipbuilding order) through the ARIMA (12,1,1) model (Tables 5, 6 and 9). On the one hand, this model can provide a useful tool for deciding on the new bulk shipbuilding order and help investment related risk management. On the other hand, it is needed to be very cautious as the UCL (Upper Confidence Level) and LCL (Lower Confidence Level) values and the analyses of errors on the predicted results are very wide (see columns 4, 5, 6 and 8 of Table 11); though the errors in the percentage of the predicted values to the original values (actual bulk new shipbuilding order) appears to be low (see column 7 of Table 11). Taking into account of the UCL and LCL values for illustrating this ARIMA analysis elevates the quality of the outcomes (Mohamadi et al., 2011).
The unexplained variance of about 27-35% might have a reflection of the constant of the model (UB) (see Equation (2) and 3) that initially assumed to be a binary one (1) to recognise the uncertainty part in new bulk shipbuilding order forecasting. This uncertainty is due to the fact that shipbuilding order forecasting becomes a complex issue that plausibly arises from international regulatory, business and ship operation environment that eventually form a variable for behavioural pattern of shipping market players. This unpredictable pattern is very evident in the outcome of future forecast on the bulk shipbuilding order through the ARIMA (12,1,1) model based on the SbOB monthly data series (2008-2017) [Figure 9 and Table 12]. For instance, the future predicted value of bulk shipbuilding order in this study for 2018 is 18.33 M dwt with the UCL and LCL values are 52.89 and 6.9 M dwt respectively, whereas an already published report shows that the actual bulk shipbuilding order in 2018 was 48.1 M dwt (BRS Group, 2019). This variable order placement may have occurred due to the changing international regulatory, trade and climate-related operational dynamics in the shipping industry which has different meanings to different investors. However, new bulk shipbuilding order in the 1st half of 2019 is about 73% down in comparison to the 1st half of 2018 (Watkins, 2019); this falling trend seems apparently reflected in the future forecasting for 2019 (see Table 12).
The study reveals an important limitation of the quality of maritime data, which is improving gradually though (World Bank Group, 2019). The lack of quality data may have an impact on the forecasting results. It is found during the construction of the independent (predictive) variables through econophysics approach that most of the publicly available reports used the Clarkson database in various ways; the presentation of these data are either not clear or not comparable even among different reports of the same organisation. Mostly the yearly data were available. However, the actual monthly data could reflect a better insight of the industry.
The lack of reliable required data was another limitation, for example, the absence of reliable bulk carrier’s speed related data which could improve the distance variable (dBsq) of this study. In drawing resemblance to the law of gravity, the incorporation of bulk carrier’s monthly speed data as a multiplier could better reflect the distance (dBsq) between the “existing bulk shipping market prospect (Bfr)” and the “future bulk shipping market prospect (Bfp)”. This improvement is noticeable while lower speed creates demand for more shipping capacity and vice versa. However, the improvement of the independent variables may bring more reliable forecasting results in future through this model. This inherent opportunity to improve variables eventually ensures the robustness of the econophysics-based ARIMA (12,1,1) model. Another limitation of this study is related to the emissions reduction efforts from international shipping. On the one hand, the efforts of reducing emissions from shipping would reduce the demand of resources such as coal consumption, which in turn may impact the bulk shipbuilding sector. On the other hand, the requirement of automation and change of the propulsion system may increase new bulk shipbuilding order. Though these two diverse impacts scenario may have a balancing or nullifying impact on the bulk shipbuilding order, this has not been captured in this forecasting study, as there were lack of relevant data.
5. Conclusion
An effective shipbuilding order forecasting is a trailblazing task that relates to many internal and external factors. The literature review of this study illustrates the constructs of the shipbuilding market and reveals a moderate cyclical nature in the bulk shipbuilding order forecasting. The manifestation of the moderate cyclical nature of shipbuilding (within the period 2008–2017) has been informed in preparing the ARIMA technique to evaluate the econophysics-based bulk shipbuilding order forecasting model of this study. Among the main three segments of shipbuilding, bulk shipbuilding order forecasting is worth doing as it is a relatively good indicator of global resources trade that reflects on global production dynamics and provides an insight into the global economy.
An innovative method of forecasting can enrich forecasting study as well as can provide an opportunity to compare with the other forecasting methods’ outcomes. In this context, an econophysics approach has been pioneered in this study to develop a bulk shipbuilding order forecasting model. The flexibility of the econophysics approach is also well suited to complex shipbuilding forecasting where addressing a large number of constructs requires to be accommodated.
The outcomes of the econophysics model indicate a relatively stable good fit. Although relevant maritime data and its quality need to be improved, the flexibility in refining the predictive variables ensure the robustness of this econophysics-based forecasting model. However, the uncertainty in the external environment also looms as a critical factor in shipbuilding forecasting that may necessitate investors to keep a cautious look at various exogenous factors such as global maritime regulatory environment, business environment ship’s operational environment. It is also worthwhile to state that the interpretation and predictive power of exogenous factors may vary considerably from investors to investors that may inspire in using this flexible econophysics-based forecasting model for bulk shipbuilding orderbook.
Figures
Naturalizing the variables of econophysics model for new bulk shipbuilding order forecasting
| Variables of the proposed econophysics model, |
Proposed constructs of the variables in bulk shipbuilding forecasting model |
|---|---|
| Bfr → the existing bulk shipping market prospect | It can be measured by the product of ‘bulk freight rate (BBDI)[the Baltic Dry Index (BDI) is used]’ and the ‘international seaborne main bulk freight task (BIST)’. The data sources are the Clarkson database for BBDI and UNCTAD report (analysis based on Clarkson database) for BIST |
| Bfp → the future bulk shipping market prospect | It can be measured by the product of the percentage share of bulk ships age of twenty and above (BTWA)’ and ‘the ratio between the average bulk ship scrap price (BSCRP) and shipping newbuilding price index (BNBPI)’. The data sources include the Clarkson database for BNBPI and UNCTAD report (analysis based on Clarkson database) for BTWA and the report of the French-based brokers house the Barry Rogliano Salles (BRS) Group (analysis based on Clarkson database) for BSCRP |
| dBsq → the square of the distance between ‘Bfr’ and ‘Bfp | It is measured by the ‘fleet evolution in the bulk shipping market’. The fleet evolution of the bulk market is further determined by the summing ‘existing bulk fleet (BFLT)’ with the ‘new delivered bulk fleet (BNDEL)’ and then subtracting the ‘bulk fleet demolition (BDEMO)’ from the summation. The data sources for all three constructs (BFLT, BNDEL ,BDEMO) are several reports of the French-based brokers house the Barry Rogliano Salles (BRS) Group (analysis of which are based on Clarkson database) |
| SbOB → the force between ‘Bfr’ and ‘Bfp’ that indicates bulk carrier new shipbuilding order | This is the target or dependent variable that will be forecasted in this study for new bulk shipbuilding order. The data sources for this variable (SbOB) include several reports of the French-based brokers house the Barry Rogliano Salles (BRS) Group (analysis based on Clarkson database) |
The yearly data on the fundamental constituents of the variables on bulk shipbuilding forecasting
| Year | SbOB (in M dwt) |
BFLT (in M dwt) |
BNDEL (in M dwt) |
BDEMO (in M dwt) |
BBDI (Index) |
BIST (in Bill Ton-Miles) |
BTWA (in % of total dwt) |
BSCRP (yearly average in $US/ Long Ton) |
BNBPI (Index) |
|---|---|---|---|---|---|---|---|---|---|
| 2008 | 85.4 | 391.13 | 28.9 | 0.48 | 6344.98 | 10476 | 30.30 | 452.5 | 191 |
| 2009 | 23.4 | 418.36 | 22.47 | 1.37 | 2612.47 | 11006 | 28.80 | 275.4 | 162 |
| 2010 | 88.5 | 456.62 | 79.55 | 0.76 | 2759.71 | 12336 | 27.40 | 390.4 | 150 |
| 2011 | 39.7 | 532.04 | 53 | 2.5 | 1546.61 | 13019 | 23.50 | 484.6 | 140 |
| 2012 | 24 | 623.01 | 54.24 | 3.54 | 923.67 | 14099 | 17.60 | 426.3 | 135 |
| 2013 | 75.4 | 686.64 | 34.55 | 2.3 | 1213.88 | 14764 | 11.00 | 398.8 | 125 |
| 2014 | 57 | 728.32 | 26.72 | 1.66 | 1104.17 | 15828 | 9.65 | 431.3 | 122 |
| 2015 | 33.8 | 761.78 | 26.76 | 2.89 | 712.66 | 15897 | 8.99 | 335.6 | 115 |
| 2016 | 16.3 | 779.29 | 25.93 | 3.04 | 675.81 | 16314 | 8.04 | 254.2 | 121 |
| 2017 | 39.2 | 795.52 | 21.05 | 1.43 | 1152.61 | 17217 | 7.01 | 354 | 125 |
Indication of the requirement of achieving normality and stationarity of the variable parameters
| Skewness | Kurtosis | Mean (Std. Error) |
Variance | SD | |
|---|---|---|---|---|---|
| SbOB | 0.752 | −0.405 | 4.451 (0.261) | 6.532 | 2.558 |
| Bfr | 2.210 | 6.158 | 25967904.519 (2280497.876) | 499264373995425.200 | 22344224.623 |
| Bfp | 2.476 | 6.962 | 63.595 (6.069) | 3536.063 | 59.465 |
| dBsq | 1.140 | 0.520 | 389081.537 (31302.547) | 94065544849.836 | 306701.068 |
Necessity of ensuring normality of the bulk shipbuilding forecasting variables through log transformation [based on monthly ‘testing data set’ (2008-2015)]
| Test → ↓ Variables |
Kolmogorov–Smirnova | Shapiro–Wilk | ||||
|---|---|---|---|---|---|---|
| Statistics | df | Sig. | Statistics | df | Sig. | |
| SbOB | 0.118 | 96 | 0.002 | 0.922 | 96 | 0.000 |
| Log of SbOB | 0.068 | 96 | 0.200* | 0.974 | 96 | 0.054 |
| Bfr | 0.193 | 96 | 0.000 | 0.767 | 96 | 0.000 |
| Log of Bfr | 0.085 | 96 | 0.083 | 0.980 | 96 | 0.159 |
| Bfp | 0.207 | 96 | 0.000 | 0.721 | 96 | 0.000 |
| Log of Bfp | 0.061 | 96 | 0.200* | 0.980 | 96 | 0.153 |
| dBsq | 0.151 | 96 | 0.000 | 0.877 | 96 | 0.000 |
| Log of dBsq | 0.050 | 96 | 0.200* | 0.981 | 96 | 0.170 |
aLilliefors Significance Correction.
*This is a lower bound of the true significance
The tentative models with their comparable criteria values
| Model selected → | Arima (1,1,3) | Arima (3,1,1) | Arima (1,1,9) | Arima (9,1,1) | Arima (3,1,9) | Arima (9,1,3) | Arima (12,1,1) | Arima (12,1,3) | Arima (12,1,9) | Arima (12,1,12) | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | (9) | (10) | (11) | (12) |
| ↓ Criteria | ↓Expectation | ||||||||||
| Stationary R-squared | Higher value | 0.187 | 0.225 | 0.443 | 0.316 | 0.528 | 0.594 | 0.726 | 0.728 | 0.774 | 0.785 |
| R-squared | Higher value | 0.479 | 0.503 | 0.643 | 0.561 | 0.697 | 0.740 | 0.824 | 0.825 | 0.855 | 0.862 |
| RMSE | Lower value | 0.199 | 0.194 | 0.171 | 0.190 | 0.159 | 0.148 | 0.122 | 0.124 | 0.118 | 0.117 |
| MAPE | Lower value | 56.790 | 59.993 | 46.597 | 47.446 | 38.377 | 38.614 | 30.554 | 30.619 | 30.021 | 29.915 |
| MaxAPE | Lower value | 2471.260 | 2881.671 | 1709.349 | 2027.033 | 1498.643 | 1708.299 | 1524.565 | 1531.075 | 1389.177 | 1363.536 |
| MAE | Lower value | 0.139 | 0.135 | 0.126 | 0.125 | 0.107 | 0.097 | 0.066 | 0.067 | 0.068 | 0.068 |
| MaxAE | Lower value | 0.617 | 0.649 | 0.369 | 0.704 | 0.519 | 0.493 | 0.459 | 0.458 | 0.383 | 0.360 |
| Normalized- BIC | Lower value | −2.701 | −2.749 | −2.716 | −2.512 | −2.761 | −2.912 | −3.242 | −3.127 | −2.940 | −2.802 |
| Ljung-Box Statistics (Sig.) |
Lower value (Sig. should be > (0.05) |
71.371 (0.000) |
57.422 (0.000) |
48.015 (0.000) |
65.728 (0.000) |
49.334 (0.000) |
57.920 (0.000) |
10.283 (0.068) |
11.947 (0.008) |
− ( − ) |
− ( − ) |
| Constant Estimate (Sig.) |
Lower value (Sig. should be > (0.05) |
−0.480 (0.515) |
−0.591 (0.363) |
−0.746 (0.160) |
−0.696 (0.306) |
−0.650 (0.292) |
−0.648 (0.253) |
−0.003 (0.995) |
−0.003 (0.994) |
−0.322 (0.416) |
−0.659 (0.072) |
The outcome criteria values of the further diagnostic models
| Model selected → | Arima (12,1,1) | Arima (12,1,24) | Arima (24,1,12) | Arima (24,1,9) | Arima (24,1,3) | Arima (24,1,1) | |
|---|---|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) |
| ↓ Criteria | Expectation | ||||||
| Stationary R-squired | Higher value | 0.726 | 0.770 | 0.782 | 0.809 | 0.795 | 0.798 |
| R-squared | Higher value | 0.824 | 0.852 | 0.860 | 0.878 | 0.868 | 0.871 |
| RMSE | Lower value | 0.122 | 0.135 | 0.131 | 0.119 | 0.117 | 0.115 |
| MAPE | Lower value | 30.554 | 30.949 | 29.986 | 26.598 | 26.582 | 27.565 |
| MaxAPE | Lower value | 1524.565 | 1407.358 | 1412.571 | 1200.166 | 1193.314 | 1245.573 |
| MAE | Lower value | 0.066 | 0.073 | 0.069 | 0.064 | 0.066 | 0.065 |
| MaxAE | Lower value | 0.459 | 0.314 | 0.374 | 0.322 | 0.346 | 0.352 |
| Normalized- BIC | Lower value | −3.242 | −1.949 | −2.001 | −2.336 | −2.655 | −2.799 |
| Ljung-Box Statistics (Sig.) |
Lower value (Sig. should be > 0.05) |
10.283 (0.068) |
− ( − ) |
− ( − ) |
− ( − ) |
− ( − ) |
− ( − ) |
| Constant Estimate (Sig.) |
Lower value (Sig. should be > 0.05) |
−0.003 (0.995) |
−0.664 (0.208) |
−0.665 (0.163) |
−0.636 (0.051) |
−0.669 (0.013) |
−0.316 (0.175) |
Model statistics on the validation data set (monthly 2016-2017)
| Model Statistics | ||||||
|---|---|---|---|---|---|---|
| Model | No. of Predictors |
Model Fit statistics | Ljung-Box Q(18) | |||
| Stationary R-squared |
Normalized BIC | Statistics | DF | Sig. | ||
| Log of SbOB_Model(12, 1, 1) | 3 | 0.656 | 0.660 | 7.111 | 5 | 0.213 |
Model summary on the validation data set (monthly 2016-2017)
| Model Summary | |
|---|---|
| Fit statistics | Mean |
| Stationary R-squared | 0.656 |
| R-squared | 0.708 |
| RMSE | 0.356 |
| MAPE | 107.391 |
| MaxAPE | 1081.509 |
| MAE | 0.094 |
| MaxAE | 0.287 |
| Normalized BIC | 0.660 |
| Ljung-Box Statistics (Sig.) | 7.111 (0.213) |
| Constant Estimate (Sig.) | −2.198 (0.806) |
Model statistics on the forecasting data set (monthly 2008-2017)
| Model Statistics | ||||||
|---|---|---|---|---|---|---|
| Model | No. of Predictors |
Model Fit statistics | Ljung-Box Q(18) | |||
| Stationary R-squared |
Normalized BIC | Statistics | DF | Sig. | ||
| Log of SbOB_Model(12,1,1) | 3 | 0.733 | −3.427 | 8.677 | 5 | 0.123 |
Model summary on the forecasting data set (monthly 2008-2017)
| Model summary | |
|---|---|
| Fit statistics | Mean |
| Stationary R-squared | 0.733 |
| R-squared | 0.841 |
| RMSE | 0.121 |
| MAPE | 33.344 |
| MaxAPE | 1732.568 |
| MAE | 0.066 |
| MaxAE | 0.417 |
| Normalized BIC | −3.427 |
| Ljung-Box Statistics (Sig.) | 8.677 (0.123) |
| Constant Estimate (Sig.) | −0.222 (0.571) |
The fit values of ARIMA (12, 1, 1) model run on the forecasting data set (monthly 2008-2017)
| Year | Original | Predicted | LCL* | UCL** | Error | Error of Predicted values to Original values in (%) |
Probable error of Predicted values in (%) |
|---|---|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) |
| 2008 | 85.40 | 85.63 | 32.21 | 228.37 | 11.01 | −0.27 | 12.85 |
| 2009 | 23.40 | 27.32 | 16.70 | 44.94 | 12.49 | −16.75 | 45.71 |
| 2010 | 88.50 | 81.56 | 49.83 | 134.06 | 12.57 | 7.84 | 15.42 |
| 2011 | 39.70 | 42.63 | 25.97 | 69.90 | 11.44 | −7.39 | 26.83 |
| 2012 | 24.00 | 24.38 | 14.86 | 39.99 | 13.29 | −1.59 | 54.51 |
| 2013 | 75.40 | 69.53 | 42.37 | 114.04 | 13.84 | 7.78 | 19.91 |
| 2014 | 57.00 | 57.72 | 35.31 | 95.04 | 11.26 | −1.26 | 19.51 |
| 2015 | 33.80 | 34.31 | 20.93 | 56.34 | 11.68 | −1.50 | 34.05 |
| 2016 | 16.30 | 16.99 | 10.30 | 27.73 | 12.28 | −4.25 | 72.26 |
| 2017 | 39.20 | 35.44 | 21.58 | 57.92 | 11.34 | 9.59 | 31.99 |
*Lower confidence level value;
**Upper confidence level value
Future forecast for new bulk shipbuilding in ARIMA (12, 1, 1) model run on the SbOB monthly data (2008-2017)
| Year | Future forecast for New Bulk shipbuilding order (in M dwt) |
UCL** (in M dwt) |
LCL* (in M dwt) |
Available actual New Bulk Shipbuilding order (in M dwt) |
|---|---|---|---|---|
| 2018 | 18.33 | 52.89 | 6.90 | 46.4 (Close to UCL) |
| 2019 | 27.83 | 120.56 | 6.43 | 24.6 (Close to future forecast) |
| 2020 | 17.69 | 100.80 | 3.14 | – |
| 2021 | 21.57 | 149.96 | 3.11 | – |
| 2022 | 16.28 | 137.60 | 1.93 | – |
*Lower confidence level value;
**Upper confidence level value
References
Abidin, I.S.Z., Bakar, N.A.A. and Sahlan, R. (2013), “The determinants of exports between Malaysia and the OIC member countries: a gravity model approach”, Procedia Economics and Finance, Vol. 5, pp. 12-19.
Agustini, W.F., Ika Restu, A. and Endah, R.M.P. (2018), “Stock price prediction using geometric Brownian motion”, Journal of Physics: Conference Series, Vol. 974 No. 1, p. 12047.
Alizadeh, A.H., Strandenes, S.P. and Thanopoulou, H. (2016), “Capacity retirement in the dry bulk market: a vessel based logit model”, Transportation Research Part E: Logistics and Transportation Review, Vol. 92, pp. 28-42.
Ariel, A. (1989), “Delphi forecast of the dry bulk shipping industry in the year 2000”, Maritime Policy and Management, Vol. 16 No. 4, pp. 305-336.
Beenstock, M. (1985), “A theory of ship prices”, Maritime Policy and Management, Vol. 12 No. 3, pp. 215-225.
Bialynicka-Birula, J. (2015), “Modelling international trade in art – modified gravity approach”, Procedia Economics and Finance, Vol. 30, pp. 91-99.
BRS Group (2017), Annual Review and Archives, Barry Rogliano S alles (BRS) Brokers, Paris.
BRS Group (2018), Annual Review and Archives, Barry Rogliano Salles (BRS) Brokers, Paris.
BRS Group (2019), Annual Review and Archives, Barry Rogliano Salles (BRS) Brokers, Paris.
Bruce, G. and Garrard, I., (1999), The Business of Shipbuilding, Routledge, London.
Cant, J.S. and Xu, Y. (2020), “One bad apple spoils the whole bushel: the neural basis of outlier processing”, NeuroImage, Vol. 211, p. 116629.
Chakraborti, A., Toke, I.M., Patriarca, M. and Abergel, F. (2011), “Econophysics review: I. Empirical facts”, Quantitative Finance, Vol. 11 No. 7, pp. 991-1012.
Charemza, W. and Gronicki, M. (1981), “An econometric model of world shipping and shipbuilding”, Maritime Policy and Management, Vol. 8 No. 1, pp. 21-30, doi: 10.1080/03088838100000019.
Chen, J.M., (2017), Risk and Uncertainty. Econophysics and Capital Asset Pricing: Splitting the Atom of Systematic Risk, Springer International Publishing, Cham.
Chen, S.-H. and Li, S.-P. (2012), “Econophysics: bridges over a turbulent current”, International Review of Financial Analysis, Vol. 23, pp. 1-10.
Chen, S., Meersman, H., Voorde, E.V.D. and Frouws, K., (2014), Modelling and Forecasting in Dry Bulk Shipping, Informa Law from Routledge, London and New York, NY.
Chou, C.-C. and Chang, P.-L. (2004), “Core competence and competitive strategy of the Taiwan shipbuilding industry: a resource-based approach”, Maritime Policy and Management, Vol. 31 No. 2, pp. 125-137.
Cotfas, L.-A. (2013), “A finite-dimensional quantum model for the stock market”, Physica A: Statistical Mechanics and Its Applications, Vol. 392 No. 2, pp. 371-380.
Dai, L., Hu, H., Chen, F. and Zheng, J. (2015), “The dynamics between newbuilding ship price volatility and freight volatility in dry bulk shipping market”, International Journal of Shipping and Transport Logistics, Vol. 7 No. 4, pp. 393-406.
Darku, A.B. (2009), “The gravity model and the test for the regional integration effect: the case of tanzania”, The Journal of Developing Areas, Vol. 43, pp. 25-44.
Dionisio, A., Menezes, R. and Mendes, D.A. (2006), “An econophysics approach to analyse uncertainty in financial markets:an application to the Portuguese stock market”, The European Physical Journal B, Vol. 50 Nos 1/2, pp. 161-164.
Donmez, C.C. and Sen, D. (2018), “Desirability index with sector analysis for investment decisions by Bernoulli theorem: a case of FTSE-100”, International Journal of Computational Physics Series, Vol. 1 No. 1, pp. 151-160.
DSF (2018), Shipping Market Review, November 2018, Danish Ship Finance (DSF), Copenhagen, Denmark.
Duan, X. and Zhang, X. (2020), “ARIMA modelling and forecasting of irregularly patterned COVID-19 outbreaks using Japanese and South Korean data”, Data in Brief, Vol. 31, p. 105779.
Duke University (2018), ARIMA Models with Regressors, Duke University, NC, available at: https://people.duke.edu/∼rnau/arimreg.htm (accessed 22 June 2020).
Duru, O. and Yoshida, S. (2009), “Judgmental forecasting in the dry bulk shipping business: statistical vs Judgmental approach”, The Asian Journal of Shipping and Logistics, Vol. 25 No. 2, pp. 189-217.
Goulielmos, A.M. and Psifia, M.-E. (2009), “Forecasting weekly freight rates for one-year time charter 65 000 dwt bulk carrier, 1989-2008, using nonlinear methods”, Maritime Policy and Management, Vol. 36 No. 5, pp. 411-436.
Guedes, E.F., Ferreira, P., Dionísio, A. and Zebende, G.F. (2019), “An econophysics approach to study the effect of BREXIT referendum on European union stock markets”, Physica A: Statistical Mechanics and Its Applications, Vol. 523, pp. 1175-1182.
Gujarati, D.N., (2003), Basic Econometrics, McGraw-Hill, New York, NY.
Halim, R.A., Smith, T. and Englert, D. (2019), “Understanding the economic impacts of greenhouse gas mitigation policies on shipping”, Policy Research Working Paper: WPS8695, The World Bank.
Hsu, L.-C. (2010), “A genetic algorithm based nonlinear grey Bernoulli model for output forecasting in integrated circuit industry”, Expert Systems with Applications, Vol. 37 No. 6, pp. 4318-4323.
Huang, J.P. (2015), “Experimental econophysics: complexity, self-organization emergent properties”, Physics Reports, Vol. 564, pp. 1-55.
Imai, S. (2008), “Recent progress and future trends for shipbuilding steel”, Welding International, Vol. 22 No. 11, pp. 755-761.
Jha, S.K. (2016), “Emerging technologies: Impact on shipbuilding”, Maritime Affairs: Journal of the National Maritime Foundation of India, Vol. 12 No. 2, pp. 78-88.
Jovanovic, F. and Schinckus, C. (2016), “Breaking down the barriers between econophysics and financial economics”, International Review of Financial Analysis, Vol. 47, pp. 256-266.
Karlis, T. and Polemis, D. (2016), “Ship demolition activity: a monetary flow process approach”, Scientific Journal of Maritime Research, Vol. 30, pp. 128-132.
Kavussanos, M.G. and Alizadeh, A.H. (2002), “Efficient pricing of ships in the dry bulk sector of the shipping industry”, Maritime Policy and Management, Vol. 29 No. 3, pp. 303-330, doi: 10.1080/03088830210132588.
Kusmartsev, F.V. (2011), “Statistical mechanics of economics I”, Physics Letters A, Vol. 375 No. 6, pp. 966-973.
Li, M., Koopman, S.J., Lit, R. and Petrova, D. (2020), “Long-term forecasting of El Niño events via dynamic factor simulations”, Journal of Econometrics, Vol. 214 No. 1, pp. 46-66.
Lim, C., Kim, Y. and Lee, K. (2017), “Changes in industrial leadership and catch-up by latecomers in shipbuilding industry”, Asian Journal of Technology Innovation, Vol. 25 No. 1, pp. 61-78.
Lyutikova, L. and Shmatova, E. (2020), “Using a logical derivative to analyze data outlier”, Procedia Computer Science, Vol. 169, pp. 304-309.
Mccauley, J.L., (2004), Dynamics of Markets: Econophysics and Finance, Cambridge University Press, Cambridge.
Mantegna, R.N. (2016), “Some past and present challenges of econophysics”, The European Physical Journal Special Topics, Vol. 225 Nos 17/18, pp. 3261-3267.
Meng, X., Zhang, J.-W. and Guo, H. (2016), “Quantum Brownian motion model for the stock market”, Physica A: Statistical Mechanics and Its Applications, Vol. 452, pp. 281-288.
Mohamadi, M., Foumani, M. and Abbasi, B. (2011), “Process capability analysis in presence of autocorrelation”, Journal of Optimization in Industrial Engineering, Vol. 4, pp. 15-20.
Nielsen, K.S., Kristensen, N.E., Bastiansen, E. and Skytte, P. (1982), “Forecasting the market for ships”, Long Range Planning, Vol. 15 No. 4, pp. 70-75.
Ohyver, M. and Pudjihastuti, H. (2018), “A“rima model for forecasting the price of medium quality rice to anticipate price fluctuations”, Procedia Computer Science, Vol. 135, pp. 707-711.
Pedram, P. (2012), “The minimal length uncertainty and the quantum model for the stock market”, Physica A: Statistical Mechanics and Its Applications, Vol. 391 No. 5, pp. 2100-2105.
Randers, J. and Göluke, U. (2007), “Forecasting turning points in shipping freight rates: lessons from 30 years of practical effort”, System Dynamics Review, Vol. 23 Nos 2/3, pp. 253-284.
Rickles, D. (2007), “Econophysics for philosophers”, Studies in History and Philosophy of Science Part B: Studies in History and Philosophy of Modern Physics, Vol. 38 No. 4, pp. 948-978.
Schinckus, C. (2009), “Economic uncertainty and econophysics”, Physica A: Statistical Mechanics and Its Applications, Vol. 388 No. 20, pp. 4415-4423.
Schinckus, C. (2010), “Econophysics and economics: sister disciplines?”, American Journal of Physics, Vol. 78 No. 4, pp. 325-327.
Schinckus, C. and Jovanovic, F. (2013), “Towards a transdisciplinary econophysics”, Journal of Economic Methodology, Vol. 20 No. 2, pp. 164-183.
Sen, P., Roy, M. and Pal, P. (2016), “Application of ARIMA for forecasting energy consumption and GHG emission: a case study of an Indian pig iron manufacturing organization”, Energy, Vol. 116, pp. 1031-1038.
Shah, N.J. and Patil, H.A. (2019), “A novel approach to remove outliers for parallel voice conversion”, Computer Speech and Language, Vol. 58, pp. 127-152.
Shin, J. and Lim, Y.-M. (2014), “An empirical model of changing global competition in the shipbuilding industry”, Maritime Policy and Management, Vol. 41 No. 6, pp. 515-527.
Steidl, C., Daniel, L. and Yildiran, C., (2018), “Shipbuilding market developments”, Working Party on Shipbuilding, Organisation for Economic Co-operation and Development (OECD), Paris.
Stopford, R.M. (1987), “A new life for shipbuilding in the 1990s?”, Maritime Policy and Management, Vol. 14 No. 4, pp. 301-312.
Stopford, R.M., (2001), Forecasting the Dry Bulk, Tanker and Container Markets, Clarkson Research, Maritime Cyprus. Cyprus.
Stopford, M., (2009), Maritime Economics, Routledge, London.
Stopford, R.M. and Barton, J.R. (1986), “Economic problems of shipbuilding and the state”, Maritime Policy and Management, Vol. 13 No. 1, pp. 27-44, doi: 10.1080/03088838600000019.
Tabachnick, B.G. and Fidell, L.S., (2007), Using Multivariate Statistics, Pearson Education, Boston, MA.
Tinbergen, J., (1962), Shaping the World Economy: Suggestions for an International Economic Policy, Twentieth Century Fund, New York, NY.
Tokumitsu, M., Hasegawa, K. and Ishida, Y. (2015), “Toward resilient sensor networks with spatiotemporal interpolation of missing data: an example of space weather forecasting”, Procedia Computer Science, Vol. 60, pp. 1585-1594.
UNCTAD (2018a), “50 Years of review of Maritime transport, 1969-2018: Reflecting on the past, exploring the future”, Transport and Trade Facilitation Series No. 10. New York, NY and Geneva: United Nations Conference on Trade and Development (UNCTAD).
UNCTAD (2018b), “Review of Maritime transport”, Report by UNCTAD SecretariatUnited Nations Conference on Trade and Development (UNCTAD), New York, NY and Geneva.
Vishnevskiy, K., Karasev, O., Meissner, D., Razheva, A. and Klubova, M. (2017), “Technology foresight in asset intensive industries: the case of Russian shipbuilding”, Technological Forecasting and Social Change, Vol. 119, pp. 194-204.
Wada, Y., Hamada, K., Hirata, N., Seki, K. and Yamada, S. (2018), “A system dynamics model for shipbuilding demand forecasting”, Journal of Marine Science and Technology, Vol. 23 No. 2, pp. 236-252.
Wang, Z. and Pei, L. (2015), “A fourier residual modified Nash nonlinear grey Bernoulli model for forecasting the international trade of Chinese high-tech products”, Grey Systems: Theory and Application, Vol. 5 No. 2, pp. 165-177.
Watkins, O. (2019), “Record low newbuild orders placed in Q2, 2019”, The Maritime Executive, 6 July 2019.
Yang, Z., Jiang, Z., Notteboom, T. and Haralambides, H. (2019), “The impact of ship scrapping subsidies on fleet renewal decisions in dry bulk shipping”, Transportation Research Part E: Logistics and Transportation Review, Vol. 126, pp. 177-189.
Zapart, C.A. (2015), “Econophysics: a challenge to econometricians”, Physica A: Statistical Mechanics and Its Applications, Vol. 419, pp. 318-327.
Zhang, C. and Huang, L. (2010), “A quantum model for the stock market”, Physica A: Statistical Mechanics and Its Applications, Vol. 389 No. 24, pp. 5769-5775.
Zhao, X., Li, H., Ding, L. and Liu, M. (2019), “Research and application of a hybrid system based on interpolation for forecasting direct economic losses of marine disasters”, International Journal of Disaster Risk Reduction, Vol. 37, p. 101121.
Zheng, J., Hu, H. and Dai, L. (2013), “How would EEDI influence Chinese shipbuilding industry?”, Maritime Policy and Management, Vol. 40 No. 5, pp. 495-510, doi: 10.1080/03088839.2013.797121.
Zhong, G.-Y., Li, J.-C., Mei, D.-C. and Tang, N.-S. (2019), “An approach for measuring corporation financial stability by Econophysics and Bayesian method”, Physica A: Statistical Mechanics and Its Applications, Vol. 527, p. 121197.
Further reading
Broadbridge, S.A. (1965), “Technological progress and state support in the Japanese shipbuilding industry”, The Journal of Development Studies, Vol. 1 No. 2, pp. 142-175, doi: 10.1080/00220386508421151.