SIMULATION STUDY ON THE COMPARISON OF ERROR CORRECTION MODEL AND AUTOREGRESSIVE DISTRIBUTED LAG MODEL FOR NON-NORMALLY DISTRIBUTED DATA

Abstract

This study investigates simulation study of the comparison between Error Correction Model
(ECM) and Autoregressive Distributed Lag (ARDL) Model on non-normally distributed data.
The study utilized both simulated data with 5000 iterations and the real life data of inflation rate
and unemployment rate in Nigeria from a period of 1981 to 2016. The Augmented-Dickey Fuller
(ADF) test revealed that inflation and unemployment are stationary at first difference while
cointegration test and bound testing revealed an existence of long run relationship between the
macroeconomic variables. The Root Mean Square Error (RMSE) of the simulated data of the two
models were compared and the normality test (Jarque-Bera test) shows that the variables are nonnormally
distributed. The study revealed that the ARDL model outperforms the ECM model for
both the real life data and simulated data and that there is evidence of both short and long run
relationship between inflation rate and unemployment rate in Nigeria. It also revealed that there
is causality relationship between inflation rate and unemployment. The study therefore
recommended based on the findings that the Central Bank of Nigeria should pursue monetary
policy that is consistent with the maintenance of a realistic and stable inflation rate in order to
reduce the unemployment rate in Nigeria and the government should formulate adequate
economic policies to stimulates and sustain the economic growth. In addition, the inflation rate
should adequately check to achieve viable economy.