Three essays in time series models of futures markets /
We demonstrated the usefulness of probability forecast evaluations in gaining greater understanding into the nature of information flows in the futures markets. Time series models (ARIMA, VAR and ECM) were used to derive probability forecasts. Probability forecast evaluations were made based on the...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
1999.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=730340291&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | We demonstrated the usefulness of probability forecast evaluations in gaining greater understanding into the nature of information flows in the futures markets. Time series models (ARIMA, VAR and ECM) were used to derive probability forecasts. Probability forecast evaluations were made based on the Brier score. Yates decomposition was used to gain an understanding of the relative forecasting abilities of these models. We applied these methods in two related studies: to understand the effect of large speculators' positions on nearby futures price in cotton and testing for market efficiency of Indian black pepper futures market. Further, we studied the information flows among on-call sales, purchases, cash and futures price in cotton. Granger causality is demonstrated between large speculators' net positions and futures prices in the case of cotton futures markets. The improvement in the forecasts of bivariate models, which use the information coming from the net positions of the large speculators show higher covariance and lower scatter and bias terms. Probability forecast evaluations were used in order to assess the univariate, VAR and VEC models forecasting performance for one to ten steps-ahead forecasts in the case of Indian black pepper futures market. In cash price models, VAR and VEC models gave lower Brier scores (performed better). Improvements come from the improved slope (better sorting capability) and reduced scatter (lower conditional variance) and bias (calibrated well in-the-large). VEC model performed better in long-range forecasting since it took into account the long-run relationships. Futures prices of black pepper are weakly exogenous and the univariate model performed better than VAR and VEC models in forecasting. The Indian domestic black pepper futures market was efficient. The cash market apparently uses the price signals generated in the futures market. Dynamic relationships between the on-call sales, purchases, cash and futures price are studied in an error correction model in the cotton market. Forecast error variance decompositions and impulse response functions were applied to the non-recursive ordering obtained through the directed graph algorithm. A11 four variables entered long run, constipating relationships. Both cash and futures price respond positively to the volumes of on-call sales and purchases. |
|---|---|
| Item Description: | Vita. "Major Subject: Agricultural Economics". |
| Physical Description: | xiii, 166 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilm Inc. |
| Bibliography: | Includes bibliographical references (leaves 79-86). |