X.Zhang et al.Physica A 415 (2014)43-53 49 New-building 6 New-building ship market ship market Second hand Freight market Second hand ship market ship market Freight market Shipping stock Shipping stock market market c New-building ship market Second hand ship market Freight narket Shipping stock market Fig.3.Granger causality relationships(at the 10%level of statistical significance)among the monthly log returns of the new-building ship market.second ship sale purchase market,freight market and shipping stock market over three sample periods:(a)prior to crisis (January 2003-December 2006).(b) during crisis (January 2007-December 2010)and(c)post-crisis (January 2011-June 2013)and marks with arrows the relationships that are significant at the 10%level.Single-headed arrows indicate uni-directional causal relationships.and double-headed arrows indicate bi-directional causal relationships. New-building b New-building ship market ship market Second-hand Freight Second-hand Freight ship market ship market Shipping Shipping stock market stock market New-building ship narket Second hand Freight ship market market Shipping stock market Fig.4.Brownian distance correlation relationships(at the 10%level of statistical significance)among the monthly log returns of the new-building ship market,second ship sale purchase market,freight market and shipping stock market over three sample periods:(a)prior to crisis (anuary 2003-December 2006).(b)during crisis (January 2007-December 2010)and (c)post-crisis (January 2011-June 2013).Single-headed arrows indicate uni- directional causal relationships,and double-headed arrows indicate bi-directional causal relationships.Red arrows indicate that feedback relationships are all significant when time lag I 1,2 and 3:green arrows indicate that feedback relationships are significant when time lag I=1 and 2;black arrows indicate that feedback relationships are significant when time lag I=1.(For interpretation of the references to color in this figure legend,the reader is referred to the web version of this article.) 5.3.Dynamic causality index To fully appreciate the impact of causal relationships in the shipping industry,finally we apply the dynamic causality index(DCI)defined such that an increase in the DCI indicates a higher level of system interconnectedness.We first apply a Granger causality test based on the new ship market containing a total of 12 time series to explore its interdependence with the second-hand ship market,the freight market,and shipping stock market prices.Based on this DCI definition,we use moving 24-month windows of monthly return prices from January 2003 to December 2012.Fig.5(a)shows the DCl ofX. Zhang et al. / Physica A 415 (2014) 43–53 49 Fig. 3. Granger causality relationships (at the 10% level of statistical significance) among the monthly log returns of the new-building ship market, second ship sale & purchase market, freight market and shipping stock market over three sample periods: (a) prior to crisis (January 2003–December 2006), (b) during crisis (January 2007–December 2010) and (c) post-crisis (January 2011–June 2013) and marks with arrows the relationships that are significant at the 10% level. Single-headed arrows indicate uni-directional causal relationships, and double-headed arrows indicate bi-directional causal relationships. Fig. 4. Brownian distance correlation relationships (at the 10% level of statistical significance) among the monthly log returns of the new-building ship market, second ship sale & purchase market, freight market and shipping stock market over three sample periods: (a) prior to crisis (January 2003–December 2006), (b) during crisis (January 2007–December 2010) and (c) post-crisis (January 2011–June 2013). Single-headed arrows indicate uni￾directional causal relationships, and double-headed arrows indicate bi-directional causal relationships. Red arrows indicate that feedback relationships are all significant when time lag l = 1, 2 and 3; green arrows indicate that feedback relationships are significant when time lag l = 1 and 2; black arrows indicate that feedback relationships are significant when time lag l = 1. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 5.3. Dynamic causality index To fully appreciate the impact of causal relationships in the shipping industry, finally we apply the dynamic causality index (DCI) defined such that an increase in the DCI indicates a higher level of system interconnectedness. We first apply a Granger causality test based on the new ship market containing a total of 12 time series to explore its interdependence with the second-hand ship market, the freight market, and shipping stock market prices. Based on this DCI definition, we use moving 24-month windows of monthly return prices from January 2003 to December 2012. Fig. 5 (a) shows the DCI of