University of Piraeus (UNIPI)
Κριτική ανάλυση παραδοσιακών μεθόδων χρηματοδότησης ναυτιλιακών επενδύσεων και προτάσεις καινοτομίας
Abstract
dc:descriptionForecasting accurately freight rates was the dream of all generations of Maritime economists. The purpose of this thesis is to use an indeed very large number of data for shipping, i.e. 428 (discrete) monthly observations (January 1968 to August 2003), concerning the index of dry cargo trip charter to “predict” its last 11 known months (within and outside the sample). My effort falls into the use of nonlinear tools (Chaos theory). In the thesis, by using different tools, such as the Rescaled Range Analysis, the BDS Statistic, the Jarque-Bera test and the Power Spectrum, I have shown that both trip and time charter dry indexes not only have long term memory and aperiodic cycles but also are not independently and identically distributed, showing also heavy kurtosis and skewness. Moreover, I proved that freight rates attractor has a correlation dimension slightly over 4 (>4) which is a fractal attractor and thus chaotic. Moreover, the system’s correlation dimension will be taken as equal to next integer (slightly over 4) i.e. 5. This means that the trip freight market 1968-2003 was a low dimensional (less than 10) system, which allows us for short term-although very important - predictions. Having calculated the Largest Lyapunov Exponent to the 428 monthly freight rates with: a time delay 1, embedding dimension m=10 (as D>4+, given also that ), I found that the is equal to +0.094 [ ]. Thus I have already shown that the trip charter dry index has a fractal dimension and a positive Lyapunov exponent, which means that, is created by a nonlinear deterministic dynamic system. The invariant characteristics of the dynamic system have been preserved in the reconstructed phase space where our time series was a part. Now the suitable prediction model remains to be specified. For a time series if chaotic, we pay the price of the inability of the long term predictability, unless we know the initial (starting conditions) conditions, which we are not. The maximal Lyapunov exponent on the other hand specifies our prediction limits. In my analysis, for forecasting the last 11 observations inside and outside the sample, I used the OLS and KDE non-linear models, that gave me a very small (EOLS=0.134 and EKDE=0.132). This means that the usage of those models was a successful one.
Degree
thesis:*- Grantor dc:publisher
- University of Piraeus (UNIPI)
- Year dc:date
- 2006
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Psifia, Elpiniki-Maria
- Ψηφία, Ελπινίκη-Μαρία
Subjects
dc:subject × 20- BDS στατιστική
- Αντεπιμονή
- Αιτιοκρατική ( ντετερμινιστική ) δυναμική
- Ανακατασκευασμένος χώρος φάσεων
- Διάσταση εμβύθισης
- ΔΥΝΑΜΙΚΟ ΣΥΣΤΗΜΑ
- Εκθέτης lyapunov
- Ελκυστής
- Antipersistence
- Attractors
- Balloon payment
- Deterministic
- Dynamic systems
- Embedding dimension
- Hurst exponent
- Lyapunov exponents
- Social Sciences
- Economics and Business
- Κοινωνικές Επιστήμες
- Οικονομικά και Επιχειρήσεις
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/26185