A time series test for probabilistic causality /
| Main Author: | |
|---|---|
| Other Authors: | , , , |
| Format: | Thesis Book |
| Language: | English |
| Published: |
1989.
|
| Subjects: | |
| Online Access: | ProQuest, Abstract Link to OAKTrust copy |
| Abstract: | A post-sample forecasting method is proposed for testing probabilistic or Granger's definition of full causality between two economic variables. The method is applied to the question of causality between the live cattle futures market and Texas-Oklahoma slaughter cattle cash prices. A time series method is employed and yields implications regarding market informational efficiencies and their dynamic interrelationships. The cattle futures market's effect upon price uncertainty in the cattle cash market is examined within a forecast context. The two price series were observed over an 14-year period, 1973 through 1986. This period was partitioned into six sub-periods, which allowed testing Granger's third axiom of causality, as well as observing the evolution of the two markets' informational efficiencies. Two different statistical loss functions were used to identify the models in order to test the robustness of the results. Univariate and bivariate prequential forecasting systems in each period were subjected to two tests of probabilistic forecast performance. Live cattle futures prices were shown to Granger-cause slaughter cattle cash prices with some evidence of feedback. The causal inferences drawn from the post-sample forecast results were found to be robust with respect to model order and invariant with respect to time. A positive dynamic relationship between the two markets was observed with the cash market showing particular sensitivity to shocks in the futures market. The cattle futures market and especially the cattle cash market were found to be informationally inefficient. Addition of futures prices to the information set of a univariate cash price forecast model resulted in relatively tighter forecast distributions. These distributions were judged superior to the univariate distributions using a calibration criterion. From this it may be inferred that information contained in past and present cattle futures prices can be used to reduce future cattle cash price uncertainty. |
|---|---|
| Item Description: | Typescript (photocopy). Vita. "Major subject: Agricultural Economics." |
| Physical Description: | xvi, 219 leaves : illustrations ; 29 cm |
| Bibliography: | Includes bibliographical references. |