A Generalized Dynamic Factor Model for the U.S. Port Sector

Jason Angelopoulos, Costas I. Chlomoudis


Although rarely available ports produce a polymorphic set of timely available monthly import, export, transport and labor utilization series, providing frequent snapshots of freight volumes either as being transferred between modes, or trans-shipped to secondary destinations. Utilizing monthly inflows and outflows of several cargo types as well as cruise passenger volumes from U.S ports, we: a) demonstrate the potential and the added value of information carried by common factors shaped by ports with respect to outlining the underlying forces of a national economy and b) provide competitive forecasts of disaggregate trade series from single ports (such as, e.g. outgoing or incoming TEUs) by exploiting factor dynamics, We test this concept in the context of Forni et al. (2005) one-sided generalized dynamic factor model, exploring the links between ports and the driving factors of the U.S. economy, as these are captured through its common and idiosyncratic components. Our model, employing 192 series from 31 major port complexes -covering 84.4% of TEUs and 60.1% of the dry bulk volume between 2005 and 2012-, displays a promising forecasting performance for individual ports and aggregate economic indicators versus benchmark models at 4-7 months ahead and explains a high fraction of the US GDP and Industrial Production indices variance.

JEL Classification: C38, F47, L99


dynamic factor models, U.S ports, trade, forecasting

Full Text:


η δικτυακή πύλη της ευρωπαϊκής ένωσης ψηφιακή ελλάδα ΕΣΠΑ 2007-2013