| dc.description.abstract | In this work the exchange rate variable through time series models, short-term memory and long memory, is comparing the results from the regression to determine which the best alternative is.
Are studied through regression analysis theories that attempt to explain exchange rate movements, such as: purchasing power parity, the monetary approach to exchange rate, monetary approach to the balance of payments, balanced portfolio models, BEER model and the traditional model of flows (FEER).The paper presents a detailed analysis of each model and presented evidence that validate the quality of the results, the investigation concluded that cutting models Keynesian monetary and present problems of autocorrelation and heterosedasticity, which from the econometric point of view invalidates any study and inference is made from them, so they are not useful for explaining the behavior of the peso dollar exchange rate in the selected period.
The analysis of time series behavior of the exchange rate was based on the Box-Jenkins methodology, which corresponds to short memory ARIMA models that were released by Box and Jenkins (1976) and contributed to Tim Bollerslev GARCH (1986). The study was extended to test the long memory models released by Granger (1980) and Hosking (1981) are ARFIMA or FARIMA and FIGARCH, the latter by Ballie, Bollerslev and Mikkelsen (1996) and found interesting results are illustrated in detail.
For long memory models are used tests for the same, which are generally accepted as the coefficient of Hurst (1956), Geweke and Porte-Hudak (1982) traditionally called GPH test and last testlocal Whittle likelihood estimators.
The results obtained in the regression models suggest that the exchange rate is better explained by models that accept the variance is not constant over time, these being the GARCH models and FIGARCH. The best result was given by the fractionally integrated model, provided that all assumptions and constraints of the methodology. | |
