forecasting with large scale econometric systems
abstract (introduction): large scale econometric systems cause special problems for estimation, forecasting, and control. common feature of large scale systems is a hierarchical structure which recommends to make use of decentralized procedures in system applications. the paper investigates the char...| Link(s) zu Dokument(en): | IHS Publikation |
|---|---|
| 1. Verfasser: | |
| Format: | IHS Series NonPeerReviewed |
| Sprache: | Englisch |
| Veröffentlicht: |
institut fuer hoehere studien
1976
|
| Zusammenfassung: | abstract (introduction): large scale econometric systems cause special problems for estimation, forecasting, and control. common feature of large scale systems is a hierarchical structure which recommends to make use of decentralized procedures in system applications. the paper investigates the characteristic features of hierarchical systems and focuses on decentralized solution and forecasting methods. the concepts developed serve as a starting point for other system applications like the investigation of stochastic system properties or optimal control experiments. all results stress the importance of stochastic aspects in the analysis of large scale systems. hierarchical systems are characterized by a number of subsystems which are connected via interaction channels. due to these interactions a forecast produced by the individual subsystems in general will not meet some overall system constraints. in case of a world trade model, e.g., the import forecasts of the individual countriesbased on their assumptions about export demand might not add up to the volume of world trade calculated either by adding up country imports or adding up country exports. therefore the role of a coordinator has to be defined to take into account the interactions between the subsystems. attempts to formulate a statistical forecasting theory for hierarchical systems indicate that the advantages of a decentralized decision process in a forecasting procedure depend on the information structure of the system. in case of an improper information exchange either no forecast is found which meets all system restrictions or the coordinator "wipes out" the subsystem forecasts. based on kalman filter theory suggestions are made for the information exchange between the subsystems and a coordinator in a forecasting exercise. explicitely the relative precision of the subsystem forecasts and the coordinator is taken into account. a number of potential applications become evident for the link system - aworldwide research project to study the international transmission of economic fluctuations by linking national econometric models. the basic research strategy of link stresses the importance of the model builders' knowledge in the process of model specification and model application. the theory of decentralized solution procedures for large scale systems provides us with a better understanding of the present information exchange pattern used within the link system and with proposals for improve |
|---|