Model identification and forecasting under structural break : three essays in macro econometrics /
Sims (1980) proposed the vector autoregressive IVARI model as a viable alternative to large-scale structural econometric models. This modeling technique has been widely used in empirical research because of its simplicity and ability to generate relatively accurate forecasts of economic variables. H...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
1998.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=733038401&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | Sims (1980) proposed the vector autoregressive IVARI model as a viable alternative to large-scale structural econometric models. This modeling technique has been widely used in empirical research because of its simplicity and ability to generate relatively accurate forecasts of economic variables. However, some authors (Cooley and Leroy; Leamer; and Sargent) have argued that while VAR models are useful for forecasting, they are not appropriate for policy analysis. Recovering the structural parameters from an estimated VAR requires that some identifying restrictions be imposed. This study considers the application of graph theory as an alternative to the Choleski decomposition. The proposed approach is a modification of Belnarke's orthogonalization procedure that allows the researcher to impose more restrictions on the structural model. A linear causal model called a directed acyclic graph (DAG) is initially specified and the resulting contemporaneous causal structure of the errors is used in assigning causal flows in Belnanke's procedure. Using Sims' (1986) macro model, identifying restrictions from a data-determined DAG is contrasted with a subjective, theory-based identification. Furthermore, this methodology was also used to answer the question of whether monetary authorities respond to information embodied in an auction-type commodities market. Lastly, this study address the question of whether imposing the restriction of cointegration always improve long range forecasts. Out-of-sample forecasts from error based on directed graphs produced more plausible inferences on the interactions among the macroeconomic variables under analysis. Moreover, empirical evidence also supports both passive and active monetary policy. Lastly, forecast results from alternative models indicate that VAR in levels specification is sometimes appropriate for modeling cointegrated data and that imposing cointegration restrictions on a VAR model does not always produce improved long-range forecasts. |
|---|---|
| Item Description: | Vita. |
| Physical Description: | xv, 197 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilm Inc. |
| Bibliography: | Includes bibliographical references: pages 77-81. |