Directed graphs, model specifications, and forecast encompassing : an analysis of U.S. meat demand /
This research offers a comparison of forecasting ability of both static and dynamic models with an application to U.S. meat demand. Four static demand systems are considered: Almost Ideal Demand System (AIDS), Rotterdam, Directed Graph Model (DGM) and Asymptotic Ideal Model (A1M) together with tw...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
2000.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=727712341&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | This research offers a comparison of forecasting ability of both static and dynamic models with an application to U.S. meat demand. Four static demand systems are considered: Almost Ideal Demand System (AIDS), Rotterdam, Directed Graph Model (DGM) and Asymptotic Ideal Model (A1M) together with two time series models: Autoregressive Distributed Lag model (ADL) and Vector Error Correction Model (VECM). In general, the dynamic models forecast better than the static models; among the four static models, the DGM model performs best, followed by the AIDS model. The VECM model is the best single method, while composite forecasts are more accurate than forecasts from any individual model. These results are based on thirty-two one-step ahead and thirty-one two-steps ahead recursive forecasts of 1990:1 through 1997:4 consumption of beef, poultry and pork. Evidence from directed graphs on innovations from three VECM models shows changes in beef, poultry and pork consumption cause changes in their prices, not vice versa. This lends support to the use of inverse demand systems, against the more popularized quantity-dependent demand systems. The hypothesis of forecast encompassing is considered in both series versus series and system versus system tests. The VECM model encompasses the AIDS model in terms of all three individual forecast series. But the VECM as a whole system does not encompass the AIDS system. This dissertation provides the first evidence on whether imposition of homogeneity restrictions on a cointegration space can improve the forecast accuracy of a VECM model. Such restrictions improve the forecast performance of both VECM models on beef and poultry consumption, but lead to larger mean square forecast errors in the VECM model on pork consumption. A possible explanation for this result is that time series properties are less evident in pork consumption series than in the other meat series. |
|---|---|
| Item Description: | Vita. "Major Subject: Agricultural Economics". |
| Physical Description: | x, 150 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilm Inc. |
| Bibliography: | Includes bibliographical references (leaves 89-107). |