The main purpose of this dissertation, after providing an extensive review with respect to the major characteristics of maritime industry, is to examine the extent of which the freight rates affect the stock performance of shipping companies. In order to do so, we conducted an empirical analysis by regressing the returns of Dow Jones Global Shipping Index (R_DJGSH), which comprises the 25 highest ranked shipping companies by indicated annual yield, on the returns of Baltic Dry Index (R_BDI), a major freight rate indicator. We expected that the higher the price of freights the better the stock performance of the shipping companies. The results that came up after the analysis confirmed the aforementioned positive relationship. Later on, we tried to examine whether the addition of one more independent variable, this of the returns of Crude Oil Index (R_CLC1), will provide a better interpretation to our investigation. Crude Oil is added due to the fact that it is one of most important factors when evaluating the transportation costs of a shipping voyage; It is estimated that almost half of the total cost of a shipping voyage is attributed to fuel costs, thus, the fluctuations of crude oil are likely to have a major impact on the profitability of shipping companies. As expected, the addition of R_CLC1 enhanced significantly the predicting ability of our model. The data have been collected in a time period between September 2006 and September 2016 on a weekly basis and were extracted from Thomson Reuters “Eikon”. We ran econometric tests for stationarity for all the three variables, by using unit root tests, i.e., Dickey Fuller (DF), Augmented Dickey Fuller (ADF) and the stationarity test Kwiatkowski–Phillips–Schmidt–Shin (KPSS). Further on, we estimated simple and multiple linear regression models, and tested the significance of the independent variables by using the Stepwise selection method. Finally, we tested if the five principal assumptions for the use Ordinary Least Squares (OLS) estimators are valid in our regression model and, in particular, we ran statistical tests in order to check the existence of heteroscedasticity, autocorrelation and normality among the residuals, as well as the existence of multicollinearity between the explanatory variables. All the tests have been run in the statistical package of “E-Views”. The conclusion is that both the R_BDI and R_CLC1 are statistically significant, the regression model meets all the assumptions and explains sufficiently the total variability of R_DJGSH at a percent of 20%.
Collections
Show Collections
