294 X.Sun et aL Economic Modelling 42 (2014)287-295 80 frequency component of the BDTI is smoother than that of the WTI, Lowf of BDTI -60 and exhibits an obvious fluctuation cycle with a periodicity of about 506 days. 40 In conclusion,the relevance between the BDTI and the WTl is mainly 20 accounted for by residual and low-frequency components.This implies that there are medium and long-term correlations between freight rates and oil prices.The BDTI is an index that reflects the trend and fluctua- tion of international tanker freight rates.In order to conduct a more in-depth analysis,tanker freight rates for all sizes of major crude oil shipping routes shown with the unit of dollar per barrel in Fig.10 are considered. According to OPEC's Annual Statistics Bulletin 2012,Middle East had more than 54%and Latin America about 23%of oil reserves of the world at the end of 2011.The routes from the Gulf to the West,from the Gulf to the East and from the Caribbean to US Atlantic Coast(USAC)cover oil -800 transportation from the above two major oil-producing regions to oil- 1998.08.032000.08.162002.09.012004.09.222007.04.122009.06.162011.08.08 consuming countries.The trend of the above freight rates is similar to Fig 9.Low-frequency components of the BDTI and WTL that of oil price.Specifically,the Caribbean/USAC as a regional market is correlated with oil price at a low level with a coefficient of 0.25.The Gulf/West and Gulf/East routes with longer shipping distance are more international than the Caribbean/USAC.Freight rates of the Gulf/ West route have the highest correlation with oil price and the next is During the period of low oil price,the long-term correlation is the Gulf/East route.The correlation coefficients of these two routes are positive while it is negative during the period of high oil price.One 0.62 and 0.58,respectively.Taking Fig.10 together with Fig.9,the possible explanation might be that tanker freight rates are not only performance of annual tanker freight rates is similar to that of the affected by costs that are mainly determined by oil price,but also influ- low-frequency components.Moreover,the correlations between oil enced by tanker supply and demand.As shown in Fig.8,the tanker fleet price and freight rates for major routes are close to those of low- was stable and it decreased a little during 1988 to 2002.According to the frequency components of BDTI and WTl:Pearson's coefficient is 0.68 statistics of OPEC,"the supply of tanker fleet recovered to the level of in the period of low oil price and 0.52 in the period of high oil price. 1998 in 2003,and after that the tanker fleet quickly increased until 2009,displaying high volatility during the period.The identification of 5.Conclusion and future work the year 2003 is in accord with analysis of the turning point in Fig.6. Oil price and tanker freight in the first period are mainly set accord- Uncertainty of tanker shipping market is correlated to volatility of ing to the basic economic law of supply and demand.Oil price changes the oil market,and this makes necessary to explore the inherent dy- around the equilibrium price due to the relatively stable supply and namic relationship between freight rates and oil prices.Taking the demand.Meanwhile,the tanker supply and demand are also relatively time-dependent features into account when modeling freight rates stable.Under the circumstances,an increase in oil price tends to this paper tries to investigate the dynamic relationship between tanker increase fuel cost and then pushes the tanker freight rate up.This freight rates and oil prices in the perspective of multiscale relevance, leads to a positive correlation.While in the second period,more and which is different from extant literature.Most importantly,it provides more non-economic factors affect the price and changes in freight a novel framework to investigate the inherent dynamic relationship be- rates,such as speculation,hedging and geopolitics,besides the law of tween two markets. supply and demand.Price rise has been the major characteristic of the Empirical results show that tanker freight rates and oil prices exhibit oil market because of the growing demand and limited supplies as oil different multiscale properties with economic meaning which can be is not a renewable resource.So oil price may move away from the identified as the long-term trend,medium-term pattern in low frequen- equilibrium price to a very large extent.In the meantime,although oil cy and short-term fluctuation in high frequency.Interestingly,it is price still affects fuel cost,it is no longer as important a factor as in the found that there is a turning point of relevance structure in June 2003. first period.Tanker freight rates are also influenced by other factors in which divided the whole sample into two sub-periods,high oil price pe- which new capacity plays an important role.This explains why most riod and low oil price period.Correlation between the two residuals of of the correlations in the second period are a little lower than in the the BDTI and WTI is extremely high and almost close to 1.00,but in to- first period.More and more newly built tankers are continuously put tally opposite direction in the two sub-periods.However,correlations into service because of the growing oil demand,and this can pull freight between medium-term patterns in low frequency are both high and rates down due to market competition. positive.This implies that it is necessary and rational to consider the dy- Also,Fig.9 shows that the second highest correlation is between namic relationship in multiscales under the relevance structure. low-frequency components which account for the medium-term In fact,the multiscale relevance between tanker freight rates and oil pattern of original BDTI and WTI.When the low-frequency component prices is our focus in the paper.In this context,the paper does not in- of oil price exhibits bigger fluctuations,low-frequency component of volve the following issues:the supply adjustment cycle,determinants the BDTI also experiences similar volatility.As mentioned above,the of the multiscale relevance,time frequency of the original data and fore- low-frequency component describes the structural pattern of original casting the freight rates when considering multiscale relevance.Inevita- time series.These two medium-term patterns show a similar trend, bly,our future work will center on these issues. for example,the fluctuation marked by the shadow.The low- Acknowledgments The authors gratefully acknowledge the financial support from National Natural Science Foundation of China (Nos.71003091, The data is available at http://www.opec.org/library// 71373009 and 71133005)the program of Youth Innovation Promotion interactive/current/FileZ/Main-Dateien/Section4html. Association supported by Chinese Academy of Sciences(CAS),and KeyDuring the period of low oil price, the long-term correlation is positive while it is negative during the period of high oil price. One possible explanation might be that tanker freight rates are not only affected by costs that are mainly determined by oil price, but also influ￾enced by tanker supply and demand. As shown in Fig. 8, the tanker fleet was stable and it decreased a little during 1988 to 2002. According to the statistics of OPEC,4 the supply of tanker fleet recovered to the level of 1998 in 2003, and after that the tanker fleet quickly increased until 2009, displaying high volatility during the period. The identification of the year 2003 is in accord with analysis of the turning point in Fig. 6. Oil price and tanker freight in the first period are mainly set accord￾ing to the basic economic law of supply and demand. Oil price changes around the equilibrium price due to the relatively stable supply and demand. Meanwhile, the tanker supply and demand are also relatively stable. Under the circumstances, an increase in oil price tends to increase fuel cost and then pushes the tanker freight rate up. This leads to a positive correlation. While in the second period, more and more non-economic factors affect the price and changes in freight rates, such as speculation, hedging and geopolitics, besides the law of supply and demand. Price rise has been the major characteristic of the oil market because of the growing demand and limited supplies as oil is not a renewable resource. So oil price may move away from the equilibrium price to a very large extent. In the meantime, although oil price still affects fuel cost, it is no longer as important a factor as in the first period. Tanker freight rates are also influenced by other factors in which new capacity plays an important role. This explains why most of the correlations in the second period are a little lower than in the first period. More and more newly built tankers are continuously put into service because of the growing oil demand, and this can pull freight rates down due to market competition. Also, Fig. 9 shows that the second highest correlation is between low-frequency components which account for the medium-term pattern of original BDTI and WTI. When the low-frequency component of oil price exhibits bigger fluctuations, low-frequency component of the BDTI also experiences similar volatility. As mentioned above, the low-frequency component describes the structural pattern of original time series. These two medium-term patterns show a similar trend, for example, the fluctuation marked by the shadow. The low￾frequency component of the BDTI is smoother than that of the WTI, and exhibits an obvious fluctuation cycle with a periodicity of about 506 days. In conclusion, the relevance between the BDTI and the WTI is mainly accounted for by residual and low-frequency components. This implies that there are medium and long-term correlations between freight rates and oil prices. The BDTI is an index that reflects the trend and fluctua￾tion of international tanker freight rates. In order to conduct a more in-depth analysis, tanker freight rates for all sizes of major crude oil shipping routes shown with the unit of dollar per barrel in Fig. 10 are considered. According to OPEC's Annual Statistics Bulletin 2012, Middle East had more than 54%, and Latin America about 23% of oil reserves of the world at the end of 2011. The routes from the Gulf to the West, from the Gulf to the East and from the Caribbean to US Atlantic Coast (USAC) cover oil transportation from the above two major oil-producing regions to oil￾consuming countries. The trend of the above freight rates is similar to that of oil price. Specifically, the Caribbean/USAC as a regional market is correlated with oil price at a low level with a coefficient of 0.25. The Gulf/West and Gulf/East routes with longer shipping distance are more international than the Caribbean/USAC. Freight rates of the Gulf/ West route have the highest correlation with oil price and the next is the Gulf/East route. The correlation coefficients of these two routes are 0.62 and 0.58, respectively. Taking Fig. 10 together with Fig. 9, the performance of annual tanker freight rates is similar to that of the low-frequency components. Moreover, the correlations between oil price and freight rates for major routes are close to those of low￾frequency components of BDTI and WTI: Pearson's coefficient is 0.68 in the period of low oil price and 0.52 in the period of high oil price. 5. Conclusion and future work Uncertainty of tanker shipping market is correlated to volatility of the oil market, and this makes necessary to explore the inherent dy￾namic relationship between freight rates and oil prices. Taking the time-dependent features into account when modeling freight rates, this paper tries to investigate the dynamic relationship between tanker freight rates and oil prices in the perspective of multiscale relevance, which is different from extant literature. Most importantly, it provides a novel framework to investigate the inherent dynamic relationship be￾tween two markets. Empirical results show that tanker freight rates and oil prices exhibit different multiscale properties with economic meaning which can be identified as the long-term trend, medium-term pattern in low frequen￾cy and short-term fluctuation in high frequency. Interestingly, it is found that there is a turning point of relevance structure in June 2003, which divided the whole sample into two sub-periods, high oil price pe￾riod and low oil price period. Correlation between the two residuals of the BDTI and WTI is extremely high and almost close to 1.00, but in to￾tally opposite direction in the two sub-periods. However, correlations between medium-term patterns in low frequency are both high and positive. This implies that it is necessary and rational to consider the dy￾namic relationship in multiscales under the relevance structure. In fact, the multiscale relevance between tanker freight rates and oil prices is our focus in the paper. In this context, the paper does not in￾volve the following issues: the supply adjustment cycle, determinants of the multiscale relevance, time frequency of the original data and fore￾casting the freight rates when considering multiscale relevance. Inevita￾bly, our future work will center on these issues. Acknowledgments The authors gratefully acknowledge the financial support from National Natural Science Foundation of China (Nos. 71003091, 71373009 and 71133005) the program of Youth Innovation Promotion Association supported by Chinese Academy of Sciences (CAS), and Key 4 The data is available at http://www.opec.org/library/Annual%20Statistical%20Bulletin/ interactive/current/FileZ/Main-Dateien/Section4.html. -800 -400 0 400 800 -40 -20 0 20 40 60 80 Lowf of BDTI Lowf of WTI 1998.08.03 2000.08.16 2002.09.01 2004.09.22 2007.04.12 2009.06.16 2011.08.08 Fig. 9. Low-frequency components of the BDTI and WTI. 294 X. Sun et al. / Economic Modelling 42 (2014) 287–295