ESTIMATING CO-INTEGRATING REGRESSION IN TIME SERIES

Fikry Mahmoud, Rania

doi:10.21608/fjard.2019.190383

ESTIMATING CO-INTEGRATING REGRESSION IN TIME SERIES

Document Type : Research articles.

Author

Rania Fikry Mahmoud

Cent. Lab. for Design and Stat. Analysis Res., ARC.

10.21608/fjard.2019.190383

Abstract

The study investigated the equilibrium relation, on the long run,
between the local production (GDP) and the index number for the wholesale
prices during the period 1986-2016 and identifying their integration degree so that
some methods could be used to estimate and compare the regression of cointegration to create accurate estimations with a high degree of confidence.
Prediction of greater precision and the study of the unit root test for the stability of
time series show that the time series for both the index number of wholesale
prices and the local production (GDP) are unstable at its level at the first
difference while it was stable at the second difference. The latent root
maximization test showed that there was at least one vector of co-integration.
Three methods of joint integration were used in comparison with the ordinary
least squares method namely: the modified least squares method, the method of
regression of the common integers and the lower dynamic squares method. The
method of dynamic least squares was the best way to estimate the regression of
co-integration since it gave the best results in interpreting the relationship between
the two variables.
The study recommends further studies in which this finding could be
confirmed when using a larger number of independent variables as well as
directing research to apply these methods when studying and analyzing the
behavior of time series. Also, conducting further research in the field of time
series stability, causal relationships and methods of co-integration regression on
the short and long term.

Main Subjects