An article authored by Assistant Professor Raffaella Giacomini and Prof. Clive W.J. Granger, "Aggregation of Space-Time Processes", appears as the lead article in the Journal of Econometrics' festschrift in honor of Manfred Deistler (Vol. 118, issues 1-2). Giacomini's coauthor, Clive Granger, was one of the supervisors of her doctoral dissertation at the University of California-San Diego, and is a 2003 Nobel Laureate in Economic Science.
Their article deals with the issue of forecasting "spatially correlated" variables, such as time series observed over time and across a number of different regions. A natural question in this field is whether forecasting the individual series and aggregating them is more appropriate than forecasting the aggregate total. In the words of their paper's abstract,
"In this paper we compare the relative efficiency of different methods of forecasting the aggregate of spatially correlated variables. Small sample simulations confirm the asymptotic result that improved forecasting performance can be obtained by imposing a priori constraints on the amount of spatial correlation in the system. One way to do so is to aggregate forecasts from a space-time autoregressive model (Elements of Spatial Structure, Cambridge University Press, Cambridge, 1975), which offers a solution to the `curse of dimensionality' that arises when forecasting with VARs. We also show that ignoring spatial correlation, even when it is weak, leads to highly inaccurate forecasts. Finally, if the system satisfies a `poolability' condition, there is a benefit in forecasting the aggregate variable directly." (© Elsevier B.V.)
The full text of a draft version of their paper is available in PDF format from the Boston College Economics Working Papers series as paper no. 582.
19 Nov 2003